diff --git a/.gitattributes b/.gitattributes index bed0738c7eeb449bca98b5d2f33c89a1ee56349a..668f0cbf63a03c57c6e1c1bdc7dc53c92c3e1474 100644 --- a/.gitattributes +++ b/.gitattributes @@ -58,3 +58,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text # Video files - compressed *.mp4 filter=lfs diff=lfs merge=lfs -text *.webm filter=lfs diff=lfs merge=lfs -text +JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue0-gpu2.log filter=lfs diff=lfs merge=lfs -text +JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue1-gpu3.log filter=lfs diff=lfs merge=lfs -text +JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue2-gpu2.log filter=lfs diff=lfs merge=lfs -text +JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue3-gpu3.log filter=lfs diff=lfs merge=lfs -text diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/downstream_results/dcnv2_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/downstream_results/dcnv2_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..7796dbbfd568b2693739546ca09b6f38150f4885 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/downstream_results/dcnv2_credit_card_default_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "credit_card_default", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpbp_c49mn", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpbp_c49mn/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.45164215564727783, + "lr": 3.511174872983247e-05, + "credit_card_default_val_accuracy": 0.8092661499977112, + "credit_card_default_val_precision": 0.7886391878128052, + "credit_card_default_val_recall": 0.8092661499977112, + "credit_card_default_val_f1-score": 0.7855390310287476, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7485185265541077, + "credit_card_default_val_score": 0.8092661499977112 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4460753798484802, + "lr": 3.51117632817477e-05, + "credit_card_default_test_accuracy": 0.8136333227157593, + "credit_card_default_test_precision": 0.7959308624267578, + "credit_card_default_test_recall": 0.8136333227157593, + "credit_card_default_test_f1-score": 0.7935397028923035, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7620080709457397, + "credit_card_default_test_score": 0.8136333227157593 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.45115557312965393, + "lr": 8.856372005539015e-06, + "credit_card_default_val_accuracy": 0.8116334080696106, + "credit_card_default_val_precision": 0.7919310331344604, + "credit_card_default_val_recall": 0.8116334080696106, + "credit_card_default_val_f1-score": 0.787671685218811, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7514497637748718, + "credit_card_default_val_score": 0.8116334080696106 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4473575949668884, + "lr": 8.856375643517822e-06, + "credit_card_default_test_accuracy": 0.8158159852027893, + "credit_card_default_test_precision": 0.7988327741622925, + "credit_card_default_test_recall": 0.8158159852027893, + "credit_card_default_test_f1-score": 0.7944850921630859, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7607871294021606, + "credit_card_default_test_score": 0.8158159852027893 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4507676362991333, + "lr": 3.942649982491275e-06, + "credit_card_default_val_accuracy": 0.8109570741653442, + "credit_card_default_val_precision": 0.7910326719284058, + "credit_card_default_val_recall": 0.8109570741653442, + "credit_card_default_val_f1-score": 0.7879649996757507, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7503932118415833, + "credit_card_default_val_score": 0.8109570741653442 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44750547409057617, + "lr": 3.942650891985977e-06, + "credit_card_default_test_accuracy": 0.8144727945327759, + "credit_card_default_test_precision": 0.7970324158668518, + "credit_card_default_test_recall": 0.8144727945327759, + "credit_card_default_test_f1-score": 0.794328510761261, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7594409584999084, + "credit_card_default_test_score": 0.8144727945327759 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.4513669013977051, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8106188774108887, + "credit_card_default_val_precision": 0.7906190752983093, + "credit_card_default_val_recall": 0.8106188774108887, + "credit_card_default_val_f1-score": 0.7881050705909729, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7520156502723694, + "credit_card_default_val_score": 0.8106188774108887 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44810912013053894, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8126259446144104, + "credit_card_default_test_precision": 0.7946444153785706, + "credit_card_default_test_recall": 0.8126259446144104, + "credit_card_default_test_f1-score": 0.7928016185760498, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7593835592269897, + "credit_card_default_test_score": 0.8126259446144104 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.45220452547073364, + "lr": 9.86635996014229e-07, + "credit_card_default_val_accuracy": 0.8129861354827881, + "credit_card_default_val_precision": 0.7938436269760132, + "credit_card_default_val_recall": 0.8129861354827881, + "credit_card_default_val_f1-score": 0.7889845967292786, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7494129538536072, + "credit_card_default_val_score": 0.8129861354827881 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44776108860969543, + "lr": 9.866355412668781e-07, + "credit_card_default_test_accuracy": 0.8158159852027893, + "credit_card_default_test_precision": 0.7988069653511047, + "credit_card_default_test_recall": 0.8158159852027893, + "credit_card_default_test_f1-score": 0.7953466176986694, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7596734166145325, + "credit_card_default_test_score": 0.8158159852027893 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.44736173152923586, + "std": 0.0007735398310272306, + "n": 5, + "ci95": 0.0009603225807708227 + }, + "lr": { + "mean": 9.779485071703676e-06, + "std": 1.4573928903195385e-05, + "n": 5, + "ci95": 1.8093021787515935e-05 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8144728064537048, + "std": 0.001389594356623852, + "n": 5, + "ci95": 0.0017251326761098772 + }, + "credit_card_default_test_precision": { + "mean": 0.7970494866371155, + "std": 0.001823788990087108, + "n": 5, + "ci95": 0.002264170091171697 + }, + "credit_card_default_test_recall": { + "mean": 0.8144728064537048, + "std": 0.001389594356623852, + "n": 5, + "ci95": 0.0017251326761098772 + }, + "credit_card_default_test_f1-score": { + "mean": 0.794100308418274, + "std": 0.0009688002276090045, + "n": 5, + "ci95": 0.0012027315175139163 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7602586269378662, + "std": 0.0011310023747805697, + "n": 5, + "ci95": 0.0014040997965998358 + }, + "credit_card_default_test_score": { + "mean": 0.8144728064537048, + "std": 0.001389594356623852, + "n": 5, + "ci95": 0.0017251326761098772 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/downstream_results/fttransformer_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/downstream_results/fttransformer_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..d57b50eaf9c8db75783ee0312be6d80146cd4ae1 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/downstream_results/fttransformer_credit_card_default_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "credit_card_default", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpjq29br_h", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpjq29br_h/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.44726452231407166, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8143388628959656, + "credit_card_default_val_precision": 0.7957068681716919, + "credit_card_default_val_recall": 0.8143388628959656, + "credit_card_default_val_f1-score": 0.791758120059967, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7507679462432861, + "credit_card_default_val_score": 0.8143388628959656 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4400160014629364, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8190060257911682, + "credit_card_default_test_precision": 0.8030407428741455, + "credit_card_default_test_recall": 0.8190060257911682, + "credit_card_default_test_f1-score": 0.7994917035102844, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7647790312767029, + "credit_card_default_test_score": 0.8190060257911682 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.44776108860969543, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8150151968002319, + "credit_card_default_val_precision": 0.7966932058334351, + "credit_card_default_val_recall": 0.8150151968002319, + "credit_card_default_val_f1-score": 0.7914837002754211, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7525211572647095, + "credit_card_default_val_score": 0.8150151968002319 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44099584221839905, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8185023665428162, + "credit_card_default_test_precision": 0.8026161789894104, + "credit_card_default_test_recall": 0.8185023665428162, + "credit_card_default_test_f1-score": 0.7966155409812927, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7639510631561279, + "credit_card_default_test_score": 0.8185023665428162 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4494698643684387, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8160297870635986, + "credit_card_default_val_precision": 0.7982540726661682, + "credit_card_default_val_recall": 0.8160297870635986, + "credit_card_default_val_f1-score": 0.7916812300682068, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7508789300918579, + "credit_card_default_val_score": 0.8160297870635986 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4420795738697052, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8176628351211548, + "credit_card_default_test_precision": 0.8013173937797546, + "credit_card_default_test_recall": 0.8176628351211548, + "credit_card_default_test_f1-score": 0.7967845797538757, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.761913001537323, + "credit_card_default_test_score": 0.8176628351211548 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.44779202342033386, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8156915903091431, + "credit_card_default_val_precision": 0.7979090213775635, + "credit_card_default_val_recall": 0.8156915903091431, + "credit_card_default_val_f1-score": 0.7905467748641968, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7489837408065796, + "credit_card_default_val_score": 0.8156915903091431 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44313955307006836, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8181665539741516, + "credit_card_default_test_precision": 0.8021530508995056, + "credit_card_default_test_recall": 0.8181665539741516, + "credit_card_default_test_f1-score": 0.7962392568588257, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.759540319442749, + "credit_card_default_test_score": 0.8181665539741516 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.44690796732902527, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8133243322372437, + "credit_card_default_val_precision": 0.7943000197410583, + "credit_card_default_val_recall": 0.8133243322372437, + "credit_card_default_val_f1-score": 0.7903100252151489, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.752513587474823, + "credit_card_default_val_score": 0.8133243322372437 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.43984168767929077, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8179986476898193, + "credit_card_default_test_precision": 0.8018132448196411, + "credit_card_default_test_recall": 0.8179986476898193, + "credit_card_default_test_f1-score": 0.7967769503593445, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7656611800193787, + "credit_card_default_test_score": 0.8179986476898193 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.44121453166007996, + "std": 0.0013984818334808497, + "n": 5, + "ci95": 0.0017361661670427474 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_accuracy": { + "mean": 0.818267285823822, + "std": 0.0005120238491346953, + "n": 5, + "ci95": 0.0006356596577118361 + }, + "credit_card_default_test_precision": { + "mean": 0.8021881222724915, + "std": 0.0006727045520112721, + "n": 5, + "ci95": 0.0008351391170456787 + }, + "credit_card_default_test_recall": { + "mean": 0.818267285823822, + "std": 0.0005120238491346953, + "n": 5, + "ci95": 0.0006356596577118361 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7971816062927246, + "std": 0.0013101887310540189, + "n": 5, + "ci95": 0.0016265533758381898 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7631689190864563, + "std": 0.0024576228063520936, + "n": 5, + "ci95": 0.0030510525525532925 + }, + "credit_card_default_test_score": { + "mean": 0.818267285823822, + "std": 0.0005120238491346953, + "n": 5, + "ci95": 0.0006356596577118361 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/downstream_results/mlp_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/downstream_results/mlp_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..a51f4843c1b66427a072aff8f4fa4829b24c6e33 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/downstream_results/mlp_credit_card_default_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "credit_card_default", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp7lkoj9si", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp7lkoj9si/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.456789493560791, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8123098015785217, + "credit_card_default_val_precision": 0.7928820848464966, + "credit_card_default_val_recall": 0.8123098015785217, + "credit_card_default_val_f1-score": 0.7886457443237305, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.73887699842453, + "credit_card_default_val_score": 0.8123098015785217 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.45091870427131653, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8183344602584839, + "credit_card_default_test_precision": 0.8021538257598877, + "credit_card_default_test_recall": 0.8183344602584839, + "credit_card_default_test_f1-score": 0.7986555695533752, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7522221207618713, + "credit_card_default_test_score": 0.8183344602584839 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.4575861692428589, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8099424839019775, + "credit_card_default_val_precision": 0.7934061884880066, + "credit_card_default_val_recall": 0.8099424839019775, + "credit_card_default_val_f1-score": 0.773453950881958, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7348690629005432, + "credit_card_default_val_score": 0.8099424839019775 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.45597708225250244, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8107790350914001, + "credit_card_default_test_precision": 0.7945290803909302, + "credit_card_default_test_recall": 0.8107790350914001, + "credit_card_default_test_f1-score": 0.7787691950798035, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7376885414123535, + "credit_card_default_test_score": 0.8107790350914001 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4570556879043579, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8133243322372437, + "credit_card_default_val_precision": 0.7943274974822998, + "credit_card_default_val_recall": 0.8133243322372437, + "credit_card_default_val_f1-score": 0.7892600297927856, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7351929545402527, + "credit_card_default_val_score": 0.8133243322372437 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4498947262763977, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8176628351211548, + "credit_card_default_test_precision": 0.8012663125991821, + "credit_card_default_test_recall": 0.8176628351211548, + "credit_card_default_test_f1-score": 0.7977256178855896, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7465294599533081, + "credit_card_default_test_score": 0.8176628351211548 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.4659510552883148, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8089279532432556, + "credit_card_default_val_precision": 0.7939008474349976, + "credit_card_default_val_recall": 0.8089279532432556, + "credit_card_default_val_f1-score": 0.7692520022392273, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7287012338638306, + "credit_card_default_val_score": 0.8089279532432556 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.46061089634895325, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.809603750705719, + "credit_card_default_test_precision": 0.7940508723258972, + "credit_card_default_test_recall": 0.809603750705719, + "credit_card_default_test_f1-score": 0.7750416398048401, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7370550632476807, + "credit_card_default_test_score": 0.809603750705719 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.47154441475868225, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8116334080696106, + "credit_card_default_val_precision": 0.7974916100502014, + "credit_card_default_val_recall": 0.8116334080696106, + "credit_card_default_val_f1-score": 0.774224579334259, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7265719771385193, + "credit_card_default_val_score": 0.8116334080696106 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.462413489818573, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8101074695587158, + "credit_card_default_test_precision": 0.7948746085166931, + "credit_card_default_test_recall": 0.8101074695587158, + "credit_card_default_test_f1-score": 0.775701642036438, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.74432772397995, + "credit_card_default_test_score": 0.8101074695587158 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.4559629797935486, + "std": 0.005600911154912593, + "n": 5, + "ci95": 0.006953334837084047 + }, + "lr": { + "mean": 0.0009999994654208422, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8132975101470947, + "std": 0.004318271150767074, + "n": 5, + "ci95": 0.005360982239874915 + }, + "credit_card_default_test_precision": { + "mean": 0.797374939918518, + "std": 0.003980594495634056, + "n": 5, + "ci95": 0.004941768511096697 + }, + "credit_card_default_test_recall": { + "mean": 0.8132975101470947, + "std": 0.004318271150767074, + "n": 5, + "ci95": 0.005360982239874915 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7851787328720092, + "std": 0.011965654304592491, + "n": 5, + "ci95": 0.014854940316568206 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7435645818710327, + "std": 0.006348770584675134, + "n": 5, + "ci95": 0.007881776100011176 + }, + "credit_card_default_test_score": { + "mean": 0.8132975101470947, + "std": 0.004318271150767074, + "n": 5, + "ci95": 0.005360982239874915 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/downstream_results/resnet_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/downstream_results/resnet_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..68de63c3d7ba661463ba381cdc4b705e67031918 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/downstream_results/resnet_credit_card_default_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "credit_card_default", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmphon_xfhk", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmphon_xfhk/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.45115408301353455, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8068988919258118, + "credit_card_default_val_precision": 0.7871233224868774, + "credit_card_default_val_recall": 0.8068988919258118, + "credit_card_default_val_f1-score": 0.7888129949569702, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7570883631706238, + "credit_card_default_val_score": 0.8068988919258118 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.443339079618454, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8171591758728027, + "credit_card_default_test_precision": 0.8012056350708008, + "credit_card_default_test_recall": 0.8171591758728027, + "credit_card_default_test_f1-score": 0.8014456629753113, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7628201842308044, + "credit_card_default_test_score": 0.8171591758728027 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.4531085193157196, + "lr": 0.0005000000237487257, + "credit_card_default_val_accuracy": 0.8048697710037231, + "credit_card_default_val_precision": 0.7840221524238586, + "credit_card_default_val_recall": 0.8048697710037231, + "credit_card_default_val_f1-score": 0.7854407429695129, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7518139481544495, + "credit_card_default_val_score": 0.8048697710037231 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4463523030281067, + "lr": 0.0004999997327104211, + "credit_card_default_test_accuracy": 0.8166554570198059, + "credit_card_default_test_precision": 0.8007574677467346, + "credit_card_default_test_recall": 0.8166554570198059, + "credit_card_default_test_f1-score": 0.8014382123947144, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7634868025779724, + "credit_card_default_test_score": 0.8166554570198059 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4580335021018982, + "lr": 0.0005000000237487257, + "credit_card_default_val_accuracy": 0.8109570741653442, + "credit_card_default_val_precision": 0.791175901889801, + "credit_card_default_val_recall": 0.8109570741653442, + "credit_card_default_val_f1-score": 0.7891989350318909, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7497589588165283, + "credit_card_default_val_score": 0.8109570741653442 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44910746812820435, + "lr": 0.0004999997327104211, + "credit_card_default_test_accuracy": 0.8174949884414673, + "credit_card_default_test_precision": 0.8012334108352661, + "credit_card_default_test_recall": 0.8174949884414673, + "credit_card_default_test_f1-score": 0.8001068234443665, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7649736404418945, + "credit_card_default_test_score": 0.8174949884414673 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.45125502347946167, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8116334080696106, + "credit_card_default_val_precision": 0.791999340057373, + "credit_card_default_val_recall": 0.8116334080696106, + "credit_card_default_val_f1-score": 0.7891373634338379, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7552090287208557, + "credit_card_default_val_score": 0.8116334080696106 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4443984627723694, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8174949884414673, + "credit_card_default_test_precision": 0.8010596632957458, + "credit_card_default_test_recall": 0.8174949884414673, + "credit_card_default_test_f1-score": 0.7985050678253174, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.764964759349823, + "credit_card_default_test_score": 0.8174949884414673 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.45458972454071045, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8099424839019775, + "credit_card_default_val_precision": 0.789559543132782, + "credit_card_default_val_recall": 0.8099424839019775, + "credit_card_default_val_f1-score": 0.783901572227478, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7507177591323853, + "credit_card_default_val_score": 0.8099424839019775 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44696947932243347, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8168233633041382, + "credit_card_default_test_precision": 0.8004374504089355, + "credit_card_default_test_recall": 0.8168233633041382, + "credit_card_default_test_f1-score": 0.7939406037330627, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7622969746589661, + "credit_card_default_test_score": 0.8168233633041382 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.4460333585739136, + "std": 0.0022561150538143587, + "n": 5, + "ci95": 0.0028008877424162515 + }, + "lr": { + "mean": 0.0005999996792525053, + "std": 0.00041832978963653156, + "n": 5, + "ci95": 0.0005193417676547898 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8171255946159363, + "std": 0.00038288523823073767, + "n": 5, + "ci95": 0.00047533859973120053 + }, + "credit_card_default_test_precision": { + "mean": 0.8009387254714966, + "std": 0.00033794866706583895, + "n": 5, + "ci95": 0.0004195514220563793 + }, + "credit_card_default_test_recall": { + "mean": 0.8171255946159363, + "std": 0.00038288523823073767, + "n": 5, + "ci95": 0.00047533859973120053 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7990872740745545, + "std": 0.003120719862088489, + "n": 5, + "ci95": 0.003874264299801887 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7637084722518921, + "std": 0.0012257065644966931, + "n": 5, + "ci95": 0.001521671727908433 + }, + "credit_card_default_test_score": { + "mean": 0.8171255946159363, + "std": 0.00038288523823073767, + "n": 5, + "ci95": 0.00047533859973120053 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy b/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..1fc7c49f8c6d4ee1f2710b357e47a0f3cc159e94 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:db71cf3f6d0f18efc7a9e05d8937e97b529b985ee334e4f17fe3c202cc535494 +size 176640128 diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/vtjepa_best_model.pth b/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..a6133af641fe80080a1964323ddccd55ef7a7f14 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1e9ad677e26f2d8a4cb101425ea45cf47d8c6d7a9eb3039e68d486aa2aac1a03 +size 15531742 diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/vtjepa_config.yaml b/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..397113492a0b53de8865bde61f6801222b3a5d37 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: credit_card_default +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.001 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.3 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.1 +mask_min_trgt_share: 0.1 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 256 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.1 +recon_y_weight: 1.0 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.001 +sw_s_y_weight: 0.0001 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 15 +vtjepa_kl_x_end_weight: 0.0001 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 15 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 15 +wjepa_joint_end_weight: 0.00125 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/deterministic_shard00of01_protocol.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/deterministic_shard00of01_protocol.json new file mode 100644 index 0000000000000000000000000000000000000000..bad0a95c2f948808c5e60d0240fb4d5569d488ca --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/deterministic_shard00of01_protocol.json @@ -0,0 +1,50 @@ +{ + "created_at_utc": "2026-09-15T20:39:55.684902+00:00", + "stage": "deterministic", + "execute": true, + "batch_size": 512, + "shard_index": 0, + "num_shards": 1, + "total_specifications": 1, + "source_sha256": { + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5" + }, + "specifications": [ + { + "dataset": "bank_marketing", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.001, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0001, + "lambda_joint": 0.004, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..be4543de1375c9dffb723a9a0267e84796888482 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-15T19:55:40.223365+00:00", + "spec": { + "dataset": "electricity", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 5e-05, + "sw_z_weight": 0.0, + "sw_s_y_weight": 1e-05, + "lambda_joint": 0.00025, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..511e21e41413fa4a5be1400ebee31e468744be8c --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "electricity", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 5e-05, + "sw_z_weight": 0.0, + "sw_s_y_weight": 1e-05, + "lambda_joint": 0.00025, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "electricity", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.00025", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "5e-05", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "1e-05", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..e789b1f9c19db368a7c140a263524132f61c0764 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "electricity", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.00025, + "s_x": 5e-05, + "z": 0.0, + "s_y": 1e-05 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 40 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "electricity", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "64", + "--model_num_layers", + "6", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.15", + "--mask_max_ctx_share", + "0.6", + "--mask_min_trgt_share", + "0.15", + "--mask_max_trgt_share", + "0.9", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "128", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.0005", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "1e-06", + "--vtjepa_kl_x_annealing_epochs", + "40", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-05", + "--vtjepa_kl_z_annealing_epochs", + "40", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-06", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.001", + "--recon_y_weight", + "0.25", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.00025", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "40", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "5e-05", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "1e-05", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results/autoint_electricity_downstream_results.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results/autoint_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..32a4b54ff671fa3d5b52496d0a67359054d1e8ad --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results/autoint_electricity_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "electricity", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpp3gogwpv", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpp3gogwpv/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.38083615899086, + "lr": 0.0, + "electricity_val_accuracy": 0.8225879073143005, + "electricity_val_precision": 0.8220837712287903, + "electricity_val_recall": 0.8225879073143005, + "electricity_val_f1-score": 0.8221973180770874, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9075164198875427, + "electricity_val_score": 0.8225879073143005 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3831469416618347, + "lr": 0.0, + "electricity_test_accuracy": 0.826154351234436, + "electricity_test_precision": 0.8257611393928528, + "electricity_test_recall": 0.826154351234436, + "electricity_test_f1-score": 0.8258849382400513, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9055925607681274, + "electricity_test_score": 0.826154351234436 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.39049676060676575, + "lr": 0.0, + "electricity_val_accuracy": 0.8221370577812195, + "electricity_val_precision": 0.8215466737747192, + "electricity_val_recall": 0.8221370577812195, + "electricity_val_f1-score": 0.8215613961219788, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9021830558776855, + "electricity_val_score": 0.8221370577812195 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.39296814799308777, + "lr": 0.0, + "electricity_test_accuracy": 0.8236075639724731, + "electricity_test_precision": 0.8230963349342346, + "electricity_test_recall": 0.8236075639724731, + "electricity_test_f1-score": 0.8231960535049438, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9012018442153931, + "electricity_test_score": 0.8236075639724731 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.37689918279647827, + "lr": 0.0, + "electricity_val_accuracy": 0.8264201879501343, + "electricity_val_precision": 0.8259297013282776, + "electricity_val_recall": 0.8264201879501343, + "electricity_val_f1-score": 0.8260307908058167, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9090912342071533, + "electricity_val_score": 0.8264201879501343 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3787509799003601, + "lr": 0.0, + "electricity_test_accuracy": 0.8295869827270508, + "electricity_test_precision": 0.8290953040122986, + "electricity_test_recall": 0.8295869827270508, + "electricity_test_f1-score": 0.8291682600975037, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9082685112953186, + "electricity_test_score": 0.8295869827270508 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.39570626616477966, + "lr": 0.0, + "electricity_val_accuracy": 0.8155996203422546, + "electricity_val_precision": 0.8151055574417114, + "electricity_val_recall": 0.8155996203422546, + "electricity_val_f1-score": 0.8152470588684082, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8984375596046448, + "electricity_val_score": 0.8155996203422546 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3995775878429413, + "lr": 0.0, + "electricity_test_accuracy": 0.8150813579559326, + "electricity_test_precision": 0.8144586682319641, + "electricity_test_recall": 0.8150813579559326, + "electricity_test_f1-score": 0.8145215511322021, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8969169855117798, + "electricity_test_score": 0.8150813579559326 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.39763420820236206, + "lr": 0.0, + "electricity_val_accuracy": 0.8162758946418762, + "electricity_val_precision": 0.816234827041626, + "electricity_val_recall": 0.8162758946418762, + "electricity_val_f1-score": 0.8146142959594727, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8983796834945679, + "electricity_val_score": 0.8162758946418762 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.399333119392395, + "lr": 0.0, + "electricity_test_accuracy": 0.8195105791091919, + "electricity_test_precision": 0.8192992210388184, + "electricity_test_recall": 0.8195105791091919, + "electricity_test_f1-score": 0.8180992603302002, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8965386748313904, + "electricity_test_score": 0.8195105791091919 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.3907553553581238, + "std": 0.009464397262821271, + "n": 5, + "ci95": 0.011749717390510018 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.8227881669998169, + "std": 0.005664891680622579, + "n": 5, + "ci95": 0.007032764416666647 + }, + "electricity_test_precision": { + "mean": 0.8223421335220337, + "std": 0.005684893885238435, + "n": 5, + "ci95": 0.007057596452442099 + }, + "electricity_test_recall": { + "mean": 0.8227881669998169, + "std": 0.005664891680622579, + "n": 5, + "ci95": 0.007032764416666647 + }, + "electricity_test_f1-score": { + "mean": 0.8221740126609802, + "std": 0.005892766908466559, + "n": 5, + "ci95": 0.007315663522981918 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.9017037153244019, + "std": 0.005197642860674944, + "n": 5, + "ci95": 0.006452691387927628 + }, + "electricity_test_score": { + "mean": 0.8227881669998169, + "std": 0.005664891680622579, + "n": 5, + "ci95": 0.007032764416666647 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results/dcnv2_electricity_downstream_results.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results/dcnv2_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..9ed889c624e33967b7329cb7f5ba6f089908d91f --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results/dcnv2_electricity_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "electricity", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpj6fuqx4a", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpj6fuqx4a/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.36828869581222534, + "lr": 0.0, + "electricity_val_accuracy": 0.8331830501556396, + "electricity_val_precision": 0.832675039768219, + "electricity_val_recall": 0.8331830501556396, + "electricity_val_f1-score": 0.8325451612472534, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9150740504264832, + "electricity_val_score": 0.8331830501556396 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.37148943543434143, + "lr": 0.0, + "electricity_test_accuracy": 0.8351234793663025, + "electricity_test_precision": 0.8347146511077881, + "electricity_test_recall": 0.8351234793663025, + "electricity_test_f1-score": 0.8348125219345093, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.913567841053009, + "electricity_test_score": 0.8351234793663025 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.37160301208496094, + "lr": 0.0, + "electricity_val_accuracy": 0.8325067758560181, + "electricity_val_precision": 0.8320543169975281, + "electricity_val_recall": 0.8325067758560181, + "electricity_val_f1-score": 0.831684410572052, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9119824767112732, + "electricity_val_score": 0.8325067758560181 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.37350720167160034, + "lr": 0.0, + "electricity_test_accuracy": 0.834126889705658, + "electricity_test_precision": 0.8336272835731506, + "electricity_test_recall": 0.834126889705658, + "electricity_test_f1-score": 0.8335598111152649, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9122480750083923, + "electricity_test_score": 0.834126889705658 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.3726440668106079, + "lr": 0.0, + "electricity_val_accuracy": 0.8345355987548828, + "electricity_val_precision": 0.8340369462966919, + "electricity_val_recall": 0.8345355987548828, + "electricity_val_f1-score": 0.8340221047401428, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9129883646965027, + "electricity_val_score": 0.8345355987548828 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.37364310026168823, + "lr": 0.0, + "electricity_test_accuracy": 0.8334625363349915, + "electricity_test_precision": 0.8330397605895996, + "electricity_test_recall": 0.8334625363349915, + "electricity_test_f1-score": 0.8331384062767029, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.913081705570221, + "electricity_test_score": 0.8334625363349915 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.3699244558811188, + "lr": 0.0, + "electricity_val_accuracy": 0.8279982209205627, + "electricity_val_precision": 0.8274524807929993, + "electricity_val_recall": 0.8279982209205627, + "electricity_val_f1-score": 0.8272846937179565, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.912844717502594, + "electricity_val_score": 0.8279982209205627 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3732665479183197, + "lr": 0.0, + "electricity_test_accuracy": 0.8370058536529541, + "electricity_test_precision": 0.8366317749023438, + "electricity_test_recall": 0.8370058536529541, + "electricity_test_f1-score": 0.8367340564727783, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9129190444946289, + "electricity_test_score": 0.8370058536529541 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.3687976896762848, + "lr": 0.0, + "electricity_val_accuracy": 0.829801619052887, + "electricity_val_precision": 0.8292626142501831, + "electricity_val_recall": 0.829801619052887, + "electricity_val_f1-score": 0.8292050361633301, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9137072563171387, + "electricity_val_score": 0.829801619052887 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.37557747960090637, + "lr": 0.0, + "electricity_test_accuracy": 0.8304728269577026, + "electricity_test_precision": 0.8299964070320129, + "electricity_test_recall": 0.8304728269577026, + "electricity_test_f1-score": 0.8300773501396179, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9117783904075623, + "electricity_test_score": 0.8304728269577026 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.3734967529773712, + "std": 0.0014520044828179868, + "n": 5, + "ci95": 0.001802612659750014 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.8340383172035217, + "std": 0.0023997884744270666, + "n": 5, + "ci95": 0.0029792532570759745 + }, + "electricity_test_precision": { + "mean": 0.833601975440979, + "std": 0.0024359075058381623, + "n": 5, + "ci95": 0.003024093768279624 + }, + "electricity_test_recall": { + "mean": 0.8340383172035217, + "std": 0.0023997884744270666, + "n": 5, + "ci95": 0.0029792532570759745 + }, + "electricity_test_f1-score": { + "mean": 0.8336644291877746, + "std": 0.0024441516813475763, + "n": 5, + "ci95": 0.003034328623142902 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.9127190113067627, + "std": 0.0007068629669606207, + "n": 5, + "ci95": 0.0008775455916491105 + }, + "electricity_test_score": { + "mean": 0.8340383172035217, + "std": 0.0023997884744270666, + "n": 5, + "ci95": 0.0029792532570759745 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results/fttransformer_electricity_downstream_results.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results/fttransformer_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..db61496e1d0882340f034166275a48afd7a51d41 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results/fttransformer_electricity_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "electricity", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpv0altbwp", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpv0altbwp/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.4253382086753845, + "lr": 0.0, + "electricity_val_accuracy": 0.8032010793685913, + "electricity_val_precision": 0.8025455474853516, + "electricity_val_recall": 0.8032010793685913, + "electricity_val_f1-score": 0.8018348813056946, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8801766633987427, + "electricity_val_score": 0.8032010793685913 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4293154776096344, + "lr": 0.0, + "electricity_test_accuracy": 0.8034547567367554, + "electricity_test_precision": 0.802844762802124, + "electricity_test_recall": 0.8034547567367554, + "electricity_test_f1-score": 0.8020485639572144, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8774914741516113, + "electricity_test_score": 0.8034547567367554 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.4262220859527588, + "lr": 0.0, + "electricity_val_accuracy": 0.8011722564697266, + "electricity_val_precision": 0.8005000352859497, + "electricity_val_recall": 0.8011722564697266, + "electricity_val_f1-score": 0.7997601628303528, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8806599974632263, + "electricity_val_score": 0.8011722564697266 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4318920969963074, + "lr": 0.0, + "electricity_test_accuracy": 0.7970324158668518, + "electricity_test_precision": 0.7962751984596252, + "electricity_test_recall": 0.7970324158668518, + "electricity_test_f1-score": 0.7956543564796448, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8763056397438049, + "electricity_test_score": 0.7970324158668518 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.42772355675697327, + "lr": 0.0, + "electricity_val_accuracy": 0.798692524433136, + "electricity_val_precision": 0.7980766892433167, + "electricity_val_recall": 0.798692524433136, + "electricity_val_f1-score": 0.7970987558364868, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8773455023765564, + "electricity_val_score": 0.798692524433136 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.43564078211784363, + "lr": 0.0, + "electricity_test_accuracy": 0.7938212752342224, + "electricity_test_precision": 0.7933134436607361, + "electricity_test_recall": 0.7938212752342224, + "electricity_test_f1-score": 0.7919358015060425, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8736765384674072, + "electricity_test_score": 0.7938212752342224 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.43353602290153503, + "lr": 0.0, + "electricity_val_accuracy": 0.7984670996665955, + "electricity_val_precision": 0.7988982200622559, + "electricity_val_recall": 0.7984670996665955, + "electricity_val_f1-score": 0.7958927154541016, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8743888139724731, + "electricity_val_score": 0.7984670996665955 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4379521310329437, + "lr": 0.0, + "electricity_test_accuracy": 0.7916066646575928, + "electricity_test_precision": 0.7918879985809326, + "electricity_test_recall": 0.7916066646575928, + "electricity_test_f1-score": 0.7889108061790466, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8720296621322632, + "electricity_test_score": 0.7916066646575928 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.3902754783630371, + "lr": 0.0, + "electricity_val_accuracy": 0.8237150311470032, + "electricity_val_precision": 0.823143482208252, + "electricity_val_recall": 0.8237150311470032, + "electricity_val_f1-score": 0.822908878326416, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9009010195732117, + "electricity_val_score": 0.8237150311470032 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3970939815044403, + "lr": 0.0, + "electricity_test_accuracy": 0.8150813579559326, + "electricity_test_precision": 0.8144314885139465, + "electricity_test_recall": 0.8150813579559326, + "electricity_test_f1-score": 0.8142136931419373, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8975225687026978, + "electricity_test_score": 0.8150813579559326 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.4263788938522339, + "std": 0.01670589280295734, + "n": 5, + "ci95": 0.020739780224778043 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.800199294090271, + "std": 0.00944213144265759, + "n": 5, + "ci95": 0.011722075155391795 + }, + "electricity_test_precision": { + "mean": 0.7997505784034729, + "std": 0.009225484129311703, + "n": 5, + "ci95": 0.011453115111287665 + }, + "electricity_test_recall": { + "mean": 0.800199294090271, + "std": 0.00944213144265759, + "n": 5, + "ci95": 0.011722075155391795 + }, + "electricity_test_f1-score": { + "mean": 0.798552644252777, + "std": 0.010032916179749019, + "n": 5, + "ci95": 0.012455513694232444 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8794051766395569, + "std": 0.010352835250606068, + "n": 5, + "ci95": 0.012852682004693278 + }, + "electricity_test_score": { + "mean": 0.800199294090271, + "std": 0.00944213144265759, + "n": 5, + "ci95": 0.011722075155391795 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results/mlp_electricity_downstream_results.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results/mlp_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..bce3786216531bb38e822d2b74e3452f641fb2ba --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results/mlp_electricity_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "electricity", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpo8zc5dpm", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpo8zc5dpm/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.45377448201179504, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7842651009559631, + "electricity_val_precision": 0.7901862859725952, + "electricity_val_recall": 0.7842651009559631, + "electricity_val_f1-score": 0.7781946659088135, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8673897981643677, + "electricity_val_score": 0.7842651009559631 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4639207720756531, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7785406112670898, + "electricity_test_precision": 0.783967912197113, + "electricity_test_recall": 0.7785406112670898, + "electricity_test_f1-score": 0.7723544239997864, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8621642589569092, + "electricity_test_score": 0.7785406112670898 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.4568510353565216, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7824617028236389, + "electricity_val_precision": 0.7891604900360107, + "electricity_val_recall": 0.7824617028236389, + "electricity_val_f1-score": 0.7759427428245544, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8643500804901123, + "electricity_val_score": 0.7824617028236389 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.46997353434562683, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.775218665599823, + "electricity_test_precision": 0.781640887260437, + "electricity_test_recall": 0.775218665599823, + "electricity_test_f1-score": 0.7683533430099487, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8589786887168884, + "electricity_test_score": 0.775218665599823 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.446729838848114, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7880973815917969, + "electricity_val_precision": 0.7878503799438477, + "electricity_val_recall": 0.7880973815917969, + "electricity_val_f1-score": 0.7856864333152771, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8661814332008362, + "electricity_val_score": 0.7880973815917969 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.45572444796562195, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7847414612770081, + "electricity_test_precision": 0.7845763564109802, + "electricity_test_recall": 0.7847414612770081, + "electricity_test_f1-score": 0.7821466326713562, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8619311451911926, + "electricity_test_score": 0.7847414612770081 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.45924338698387146, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7840396761894226, + "electricity_val_precision": 0.7830287218093872, + "electricity_val_recall": 0.7840396761894226, + "electricity_val_f1-score": 0.782926619052887, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8588030934333801, + "electricity_val_score": 0.7840396761894226 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4670161306858063, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7766581773757935, + "electricity_test_precision": 0.7755798697471619, + "electricity_test_recall": 0.7766581773757935, + "electricity_test_f1-score": 0.7755626440048218, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8529821634292603, + "electricity_test_score": 0.7766581773757935 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.47163620591163635, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7790802717208862, + "electricity_val_precision": 0.787118673324585, + "electricity_val_recall": 0.7790802717208862, + "electricity_val_f1-score": 0.771767258644104, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8560341596603394, + "electricity_val_score": 0.7790802717208862 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.48404136300086975, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7685748934745789, + "electricity_test_precision": 0.7760655283927917, + "electricity_test_recall": 0.7685748934745789, + "electricity_test_f1-score": 0.7607215046882629, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8485464453697205, + "electricity_test_score": 0.7685748934745789 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.4681352496147156, + "std": 0.010361297933687471, + "n": 5, + "ci95": 0.012863188129046547 + }, + "lr": { + "mean": 0.0009999994654208422, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.7767467617988586, + "std": 0.005836425977502845, + "n": 5, + "ci95": 0.00724571823244109 + }, + "electricity_test_precision": { + "mean": 0.7803661108016968, + "std": 0.004293230394765992, + "n": 5, + "ci95": 0.005329895019200736 + }, + "electricity_test_recall": { + "mean": 0.7767467617988586, + "std": 0.005836425977502845, + "n": 5, + "ci95": 0.00724571823244109 + }, + "electricity_test_f1-score": { + "mean": 0.7718277096748352, + "std": 0.008001973930902734, + "n": 5, + "ci95": 0.009934170094874979 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8569205403327942, + "std": 0.00596814460095666, + "n": 5, + "ci95": 0.0074092422855499915 + }, + "electricity_test_score": { + "mean": 0.7767467617988586, + "std": 0.005836425977502845, + "n": 5, + "ci95": 0.00724571823244109 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results/resnet_electricity_downstream_results.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results/resnet_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..0596ed1445d71a4620217cb4221aa84b511afbf2 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results/resnet_electricity_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "electricity", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpcftj6il5", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpcftj6il5/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.4003864824771881, + "lr": 0.0, + "electricity_val_accuracy": 0.8151487708091736, + "electricity_val_precision": 0.8146806955337524, + "electricity_val_recall": 0.8151487708091736, + "electricity_val_f1-score": 0.8148254752159119, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8960868716239929, + "electricity_val_score": 0.8151487708091736 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.40950292348861694, + "lr": 0.0, + "electricity_test_accuracy": 0.8135311603546143, + "electricity_test_precision": 0.8133290410041809, + "electricity_test_recall": 0.8135311603546143, + "electricity_test_f1-score": 0.8134183883666992, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8915355801582336, + "electricity_test_score": 0.8135311603546143 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.40759947896003723, + "lr": 0.0, + "electricity_val_accuracy": 0.8095130920410156, + "electricity_val_precision": 0.8156034350395203, + "electricity_val_recall": 0.8095130920410156, + "electricity_val_f1-score": 0.8105146288871765, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9005982279777527, + "electricity_val_score": 0.8095130920410156 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.41420847177505493, + "lr": 0.0, + "electricity_test_accuracy": 0.8124238848686218, + "electricity_test_precision": 0.8188508152961731, + "electricity_test_recall": 0.8124238848686218, + "electricity_test_f1-score": 0.8134207129478455, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8979076147079468, + "electricity_test_score": 0.8124238848686218 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.39245733618736267, + "lr": 0.0, + "electricity_val_accuracy": 0.8196573257446289, + "electricity_val_precision": 0.8201331496238708, + "electricity_val_recall": 0.8196573257446289, + "electricity_val_f1-score": 0.8198521733283997, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9018487930297852, + "electricity_val_score": 0.8196573257446289 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3950406312942505, + "lr": 0.0, + "electricity_test_accuracy": 0.8216144442558289, + "electricity_test_precision": 0.8226320147514343, + "electricity_test_recall": 0.8216144442558289, + "electricity_test_f1-score": 0.8219646215438843, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9012686610221863, + "electricity_test_score": 0.8216144442558289 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.39851632714271545, + "lr": 0.0, + "electricity_val_accuracy": 0.808836817741394, + "electricity_val_precision": 0.8083773255348206, + "electricity_val_recall": 0.808836817741394, + "electricity_val_f1-score": 0.8085333108901978, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8958708643913269, + "electricity_val_score": 0.808836817741394 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4052407741546631, + "lr": 0.0, + "electricity_test_accuracy": 0.8145277500152588, + "electricity_test_precision": 0.8143652677536011, + "electricity_test_recall": 0.8145277500152588, + "electricity_test_f1-score": 0.8144391179084778, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8931828737258911, + "electricity_test_score": 0.8145277500152588 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.3846282958984375, + "lr": 0.0, + "electricity_val_accuracy": 0.8198827505111694, + "electricity_val_precision": 0.8193657398223877, + "electricity_val_recall": 0.8198827505111694, + "electricity_val_f1-score": 0.8194862008094788, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9040712714195251, + "electricity_val_score": 0.8198827505111694 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3941192328929901, + "lr": 0.0, + "electricity_test_accuracy": 0.8197320103645325, + "electricity_test_precision": 0.8194493651390076, + "electricity_test_recall": 0.8197320103645325, + "electricity_test_f1-score": 0.819563090801239, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8995412588119507, + "electricity_test_score": 0.8197320103645325 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.4036224067211151, + "std": 0.00884903308579223, + "n": 5, + "ci95": 0.010985764338714761 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.8163658499717712, + "std": 0.004056844318513874, + "n": 5, + "ci95": 0.005036429992967676 + }, + "electricity_test_precision": { + "mean": 0.8177253007888794, + "std": 0.003838357226489016, + "n": 5, + "ci95": 0.004765185928134204 + }, + "electricity_test_recall": { + "mean": 0.8163658499717712, + "std": 0.004056844318513874, + "n": 5, + "ci95": 0.005036429992967676 + }, + "electricity_test_f1-score": { + "mean": 0.8165611863136292, + "std": 0.003951312093876537, + "n": 5, + "ci95": 0.004905415435923301 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8966871976852417, + "std": 0.004166663487993788, + "n": 5, + "ci95": 0.005172766641738574 + }, + "electricity_test_score": { + "mean": 0.8163658499717712, + "std": 0.004056844318513874, + "n": 5, + "ci95": 0.005036429992967676 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..92db635ad3b60314c48a5902e435d5bc72f62432 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6e50a71be4f826e6aa5792366b091e7f5262a3a0c9b646df0688dae100b04585 +size 92799104 diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/vtjepa_best_model.pth b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..df078589942f675860e26cc4fdcdd27647c41a83 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a34c89c118819c2802944d33c90a279a78bb2c834bd007303d71f41086222b07 +size 8332910 diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/vtjepa_config.yaml b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3788bf33925cd8931e59a0b02f66ce094204f312 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: electricity +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.0005 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.6 +mask_max_trgt_share: 0.9 +mask_min_ctx_share: 0.15 +mask_min_trgt_share: 0.15 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 64 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 6 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 128 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.001 +recon_y_weight: 0.25 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 5.0e-05 +sw_s_y_weight: 1.0e-05 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 40 +vtjepa_kl_x_end_weight: 1.0e-06 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-06 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 40 +vtjepa_kl_z_end_weight: 1.0e-05 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 40 +wjepa_joint_end_weight: 0.00025 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu2-adult-bank.exit_code b/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu2-adult-bank.exit_code new file mode 100644 index 0000000000000000000000000000000000000000..573541ac9702dd3969c9bc859d2b91ec1f7e6e56 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu2-adult-bank.exit_code @@ -0,0 +1 @@ +0 diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu2-adult-bank.log b/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu2-adult-bank.log new file mode 100644 index 0000000000000000000000000000000000000000..d9a4d3cd12f48962db5dbb33147e0465569d0695 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu2-adult-bank.log @@ -0,0 +1,33517 @@ +[gpu2-adult-bank] Starting adult on physical GPU 2 at Tue Sep 15 18:28:13 UTC 2026 ++ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py --dataset adult --results-root /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42 --epochs 500 --objective w_jepa_joint --latent-architecture deterministic_no_z --sw-full-end-weight 0.09 --torch-seed 42 --np-seed 42 --mask-seed 0 --sw-seed 0 --sw-full-num-projections 2048 --sw-marginal-num-projections 1024 --sw-p 2 --sw-s-x-end-weight 0.0005 --sw-z-end-weight 0.0 --sw-s-y-end-weight 0.0001 --model-selection unfiltered --deterministic-mode warn --sw-include-marginals --batch-size 512 +{ + "dataset": "adult", + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "adult", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.09", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.0005", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "0.0001", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} +Not SLURM job. +╒══════════════════════════════════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ mock │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ activation_checkpointing │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 42 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 42 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_seed │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_seed │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ deterministic_mode │ warn │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pre_evaluation │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 512 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 500 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ improvement │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_val_cache_cadence │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ init_type │ trunc_normal │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_start_lr │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_lr │ 0.001 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_final_lr │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 10 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_final_weight_decay │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_ipe_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_gradient_clipping │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_feature_type_embedding │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_feature_index_embedding │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dropout_prob │ 0.001 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_layer_norm_eps │ 1e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dim_hidden │ 64 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dim_feedforward │ 256 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_num_heads │ 4 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_num_layers │ 8 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_act_func │ relu │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ probe_cadence │ 2 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ probe_model │ linear_probe │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ n_cls_tokens │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_type │ transformer │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_num_layers │ 16 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_embed_dim │ 16 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_num_heads │ 4 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_p_dropout │ 0.002 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_layer_norm_eps │ 1e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_activation │ relu │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_dim_feedforward │ 256 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_allow_overlap │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_min_ctx_share │ 0.1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_max_ctx_share │ 0.3 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_min_trgt_share │ 0.1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_max_trgt_share │ 0.6 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_num_preds │ 4 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_num_encs │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mu_init_std │ 0.03 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ logvar_init_std │ 0.005 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ logvar_init_bias │ -4.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_sigma_x_init │ -2.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_sigma_y_init │ -2.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ downstream_evaluation_epochs │ 50 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ downstream_probe_patience │ 20 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ probe_representation_type │ mean │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ probe_encoder_type │ target_posterior │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pooled_context_representation_size_for_target_posterior │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pooled_context_representation_size_for_auxiliary_encoder │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_end_weight │ 0.0001 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_end_weight │ 1e-06 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_end_weight │ 1e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_annealing_epochs │ 15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_annealing_epochs │ 15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_annealing_epochs │ 15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ recon_x_weight │ 0.1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ recon_y_weight │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ latent_architecture │ deterministic_no_z │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ jepa_objective │ w_jepa_joint │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_full_num_projections │ 2048 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_marginal_num_projections │ 1024 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_p │ 2 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_include_marginals │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_s_x_weight │ 0.0005 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_z_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_s_y_weight │ 0.0001 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_end_weight │ 0.09 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_annealing_epochs │ 15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_marginal_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_marginal_end_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_end_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_start_epoch │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_annealing_epochs │ 30 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_end_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_start_epoch │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_annealing_epochs │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_end_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_start_epoch │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_annealing_epochs │ 30 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ safe_exit │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_selection │ unfiltered │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sigma_aggregation │ mean │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ subset_size │ -1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ is_batchlearning │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ iteration │ 0 │ +╘══════════════════════════════════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ + +CUDA Device Information: +CUDA_VISIBLE_DEVICES: 2 +Number of CUDA devices: 1 +Current CUDA device: 0 +CUDA device name: NVIDIA A100-SXM4-80GB + +Using device: cuda:0 +Device name: NVIDIA A100-SXM4-80GB +Encoder Module parameters: ++-------------------+----------------------------------------------------------------------+ +| Parameter | Value | ++===================+======================================================================+ +| hidden_dim | 64 | ++-------------------+----------------------------------------------------------------------+ +| idx_num_features | [0, 2, 4, 10, 11, 12] | ++-------------------+----------------------------------------------------------------------+ +| idx_cat_features | [1, 3, 5, 6, 7, 8, 9, 13] | ++-------------------+----------------------------------------------------------------------+ +| cardinalities | [(1, 9), (3, 16), (5, 7), (6, 15), (7, 6), (8, 5), (9, 2), (13, 42)] | ++-------------------+----------------------------------------------------------------------+ +| n_input_features | 14 | ++-------------------+----------------------------------------------------------------------+ +| device | cuda:0 | ++-------------------+----------------------------------------------------------------------+ +| num_layers | 8 | ++-------------------+----------------------------------------------------------------------+ +| num_heads | 4 | ++-------------------+----------------------------------------------------------------------+ +| p_dropout | 0.001 | ++-------------------+----------------------------------------------------------------------+ +| layer_norm_eps | 1e-05 | ++-------------------+----------------------------------------------------------------------+ +| gradient_clipping | 1.0 | ++-------------------+----------------------------------------------------------------------+ +| dim_feedforward | 256 | ++-------------------+----------------------------------------------------------------------+ +Encoder Module parameters: ++-------------------+----------------------------------------------------------------------+ +| Parameter | Value | ++===================+======================================================================+ +| hidden_dim | 64 | ++-------------------+----------------------------------------------------------------------+ +| idx_num_features | [0, 2, 4, 10, 11, 12] | ++-------------------+----------------------------------------------------------------------+ +| idx_cat_features | [1, 3, 5, 6, 7, 8, 9, 13] | ++-------------------+----------------------------------------------------------------------+ +| cardinalities | [(1, 9), (3, 16), (5, 7), (6, 15), (7, 6), (8, 5), (9, 2), (13, 42)] | ++-------------------+----------------------------------------------------------------------+ +| n_input_features | 14 | ++-------------------+----------------------------------------------------------------------+ +| device | cuda:0 | ++-------------------+----------------------------------------------------------------------+ +| num_layers | 8 | ++-------------------+----------------------------------------------------------------------+ +| num_heads | 4 | ++-------------------+----------------------------------------------------------------------+ +| p_dropout | 0.001 | ++-------------------+----------------------------------------------------------------------+ +| layer_norm_eps | 1e-05 | ++-------------------+----------------------------------------------------------------------+ +| gradient_clipping | 1.0 | ++-------------------+----------------------------------------------------------------------+ +| dim_feedforward | 256 | ++-------------------+----------------------------------------------------------------------+ +DJEPA parameters: ++-------------------------+---------+ +| Parameter | Value | ++=========================+=========+ +| hidden_dim | 64 | ++-------------------------+---------+ +| num_layers | 8 | ++-------------------------+---------+ +| num_heads | 4 | ++-------------------------+---------+ +| p_dropout | 0.001 | ++-------------------------+---------+ +| layer_norm_eps | 1e-05 | ++-------------------------+---------+ +| feature_type_embedding | True | ++-------------------------+---------+ +| feature_index_embedding | True | ++-------------------------+---------+ +| dim_feedforward | 256 | ++-------------------------+---------+ +| use_reg_token | True | ++-------------------------+---------+ +| gradient_clipping | 1.0 | ++-------------------------+---------+ +| activation | gelu | ++-------------------------+---------+ +| num_features | 14 | ++-------------------------+---------+ +| n_cls_tokens | 1 | ++-------------------------+---------+ +MaskCollator params: +----------------- -------------------------------------------------------------------- +Allow overlap False +Min context share 0.1 +Max context share 0.3 +Min target share 0.1 +Max target share 0.6 +Min context 1 +Max context 4 +Min target 1 +Max target 8 +Num preds 4 +Num encs 1 +Num features 14 +Cardinalities [(1, 9), (3, 16), (5, 7), (6, 15), (7, 6), (8, 5), (9, 2), (13, 42)] +Base mask seed 0 +----------------- -------------------------------------------------------------------- +KL-x annealing: start=0.0 for 96 steps, then anneal to 0.0001 over 1440 steps. +KL-z annealing: start=0.0 for 96 steps, then anneal to 1e-06 over 1440 steps. +KL-y annealing: start=0.0 for 96 steps, then anneal to 1e-05 over 1440 steps. +W-JEPA marginal-scale annealing (s_x): start=0.0, end=1.0, start_epoch=0, anneal_epochs=30. +W-JEPA marginal-scale annealing (z): start=0.0, end=0.0, start_epoch=0, anneal_epochs=0. +W-JEPA marginal-scale annealing (s_y): start=0.0, end=1.0, start_epoch=0, anneal_epochs=30. +Configuration saved to: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/vtjepa_config.yaml +All results will be saved to: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952 +[VTJEPA] Running downstream probe before training begins +Using full dataset for embedding generation: adult +Generating embedding using: JEPAEmbeddingEncoder on full dataset + 0%| | 0/96 [00:00= 15) +[VTJEPA] Early stopping triggered at epoch 78 + Best downstream validation score: 0.858097 at epoch 47 + No improvement for 15 probe evaluations +Restored best downstream-selected checkpoint from epoch 47. +Training completed. ++ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -W ignore -m src.downstream_pipeline --exp_path /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952 --num_runs 5 --epochs 50 --models mlp,dcnv2,resnet,autoint,fttransformer --no-filter_uncertainty +[Config] Loaded sigma_aggregation=mean from /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/vtjepa_config.yaml +[Config] Loaded training data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets from /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/vtjepa_config.yaml +Split statistics: + val: 4792 samples + test: 9739 samples + train: 34311 samples +Running downstream for dataset 'adult', model 'mlp'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpc5lu6ha6/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpc5lu6ha6/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpc5lu6ha6/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpc5lu6ha6/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ adult │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─Linear: 2-2 [128, 14] [128, 128] 1,920 -- 245,760 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +===================================================================================================================================================================== +Total params: 31,375 +Trainable params: 31,375 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 4.02 +===================================================================================================================================================================== +Input size (MB): 0.46 +Forward/backward pass size (MB): 0.41 +Params size (MB): 0.13 +Estimated Total Size (MB): 0.99 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 31.2 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 31.4 K +Non-trainable params: 0 +Total params: 31.4 K +Total estimated model params size (MB): 0.126 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 21/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:01 • 0:00:00 170.57it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8583055138587952 │ +│ adult_val_f1-score │ 0.8558571934700012 │ +│ adult_val_loss │ 0.31649187207221985 │ +│ adult_val_precision │ 0.8544908165931702 │ +│ adult_val_recall │ 0.8583055138587952 │ +│ adult_val_roc_auc │ 0.9098939895629883 │ +│ adult_val_score │ 0.8583055138587952 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 0.0009999998146668077 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 396.37it/s +Validation metrics: [{'adult_val_loss': 0.31649187207221985, 'lr': 0.0009999998146668077, 'adult_val_accuracy': 0.8583055138587952, 'adult_val_precision': 0.8544908165931702, 'adult_val_recall': 0.8583055138587952, 'adult_val_f1-score': 0.8558571934700012, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9098939895629883, 'adult_val_score': 0.8583055138587952}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7545 samples (77.47%) + Class 1: 2194 samples (22.53%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8506006598472595 │ +│ adult_test_f1-score │ 0.8484002351760864 │ +│ adult_test_loss │ 0.32713568210601807 │ +│ adult_test_precision │ 0.847040593624115 │ +│ adult_test_recall │ 0.8506006598472595 │ +│ adult_test_roc_auc │ 0.9046164155006409 │ +│ adult_test_score │ 0.8506006598472595 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 0.0009999996982514858 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 428.60it/s +Test metrics: [{'adult_test_loss': 0.32713568210601807, 'lr': 0.0009999996982514858, 'adult_test_accuracy': 0.8506006598472595, 'adult_test_precision': 0.847040593624115, 'adult_test_recall': 0.8506006598472595, 'adult_test_f1-score': 0.8484002351760864, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.9046164155006409, 'adult_test_score': 0.8506006598472595}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpc5lu6ha6/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpc5lu6ha6/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpc5lu6ha6/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpc5lu6ha6/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ adult │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─Linear: 2-2 [128, 14] [128, 128] 1,920 -- 245,760 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +===================================================================================================================================================================== +Total params: 31,375 +Trainable params: 31,375 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 4.02 +===================================================================================================================================================================== +Input size (MB): 0.46 +Forward/backward pass size (MB): 0.41 +Params size (MB): 0.13 +Estimated Total Size (MB): 0.99 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 31.2 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 31.4 K +Non-trainable params: 0 +Total params: 31.4 K +Total estimated model params size (MB): 0.126 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:01 • 0:00:00 181.10it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8589315414428711 │ +│ adult_val_f1-score │ 0.8520825505256653 │ +│ adult_val_loss │ 0.3074006736278534 │ +│ adult_val_precision │ 0.8526650071144104 │ +│ adult_val_recall │ 0.8589315414428711 │ +│ adult_val_roc_auc │ 0.9134336709976196 │ +│ adult_val_score │ 0.8589315414428711 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 0.0009999998146668077 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 361.74it/s +Validation metrics: [{'adult_val_loss': 0.3074006736278534, 'lr': 0.0009999998146668077, 'adult_val_accuracy': 0.8589315414428711, 'adult_val_precision': 0.8526650071144104, 'adult_val_recall': 0.8589315414428711, 'adult_val_f1-score': 0.8520825505256653, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9134336709976196, 'adult_val_score': 0.8589315414428711}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7869 samples (80.80%) + Class 1: 1870 samples (19.20%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8499845862388611 │ +│ adult_test_f1-score │ 0.843402087688446 │ +│ adult_test_loss │ 0.322750985622406 │ +│ adult_test_precision │ 0.8432666659355164 │ +│ adult_test_recall │ 0.8499845862388611 │ +│ adult_test_roc_auc │ 0.9061656594276428 │ +│ adult_test_score │ 0.8499845862388611 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 0.0009999996982514858 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 309.67it/s +Test metrics: [{'adult_test_loss': 0.322750985622406, 'lr': 0.0009999996982514858, 'adult_test_accuracy': 0.8499845862388611, 'adult_test_precision': 0.8432666659355164, 'adult_test_recall': 0.8499845862388611, 'adult_test_f1-score': 0.843402087688446, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.9061656594276428, 'adult_test_score': 0.8499845862388611}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpc5lu6ha6/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpc5lu6ha6/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpc5lu6ha6/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpc5lu6ha6/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ adult │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─Linear: 2-2 [128, 14] [128, 128] 1,920 -- 245,760 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +===================================================================================================================================================================== +Total params: 31,375 +Trainable params: 31,375 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 4.02 +===================================================================================================================================================================== +Input size (MB): 0.46 +Forward/backward pass size (MB): 0.41 +Params size (MB): 0.13 +Estimated Total Size (MB): 0.99 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 31.2 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 31.4 K +Non-trainable params: 0 +Total params: 31.4 K +Total estimated model params size (MB): 0.126 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 31/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:01 • 0:00:00 229.87it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8585141897201538 │ +│ adult_val_f1-score │ 0.849731981754303 │ +│ adult_val_loss │ 0.3134402334690094 │ +│ adult_val_precision │ 0.8525142669677734 │ +│ adult_val_recall │ 0.8585141897201538 │ +│ adult_val_roc_auc │ 0.9106864333152771 │ +│ adult_val_score │ 0.8585141897201538 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 0.0009999998146668077 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 462.86it/s +Validation metrics: [{'adult_val_loss': 0.3134402334690094, 'lr': 0.0009999998146668077, 'adult_val_accuracy': 0.8585141897201538, 'adult_val_precision': 0.8525142669677734, 'adult_val_recall': 0.8585141897201538, 'adult_val_f1-score': 0.849731981754303, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9106864333152771, 'adult_val_score': 0.8585141897201538}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 8005 samples (82.20%) + Class 1: 1734 samples (17.80%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8481363654136658 │ +│ adult_test_f1-score │ 0.8393124938011169 │ +│ adult_test_loss │ 0.32736653089523315 │ +│ adult_test_precision │ 0.8411210179328918 │ +│ adult_test_recall │ 0.8481363654136658 │ +│ adult_test_roc_auc │ 0.904833972454071 │ +│ adult_test_score │ 0.8481363654136658 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 0.0009999996982514858 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 477.36it/s +Test metrics: [{'adult_test_loss': 0.32736653089523315, 'lr': 0.0009999996982514858, 'adult_test_accuracy': 0.8481363654136658, 'adult_test_precision': 0.8411210179328918, 'adult_test_recall': 0.8481363654136658, 'adult_test_f1-score': 0.8393124938011169, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.904833972454071, 'adult_test_score': 0.8481363654136658}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpc5lu6ha6/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpc5lu6ha6/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpc5lu6ha6/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpc5lu6ha6/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ adult │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─Linear: 2-2 [128, 14] [128, 128] 1,920 -- 245,760 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +===================================================================================================================================================================== +Total params: 31,375 +Trainable params: 31,375 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 4.02 +===================================================================================================================================================================== +Input size (MB): 0.46 +Forward/backward pass size (MB): 0.41 +Params size (MB): 0.13 +Estimated Total Size (MB): 0.99 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 31.2 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 31.4 K +Non-trainable params: 0 +Total params: 31.4 K +Total estimated model params size (MB): 0.126 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 48/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:01 • 0:00:00 231.89it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8470367193222046 │ +│ adult_val_f1-score │ 0.848872184753418 │ +│ adult_val_loss │ 0.31735482811927795 │ +│ adult_val_precision │ 0.8513215184211731 │ +│ adult_val_recall │ 0.8470367193222046 │ +│ adult_val_roc_auc │ 0.9130023717880249 │ +│ adult_val_score │ 0.8470367193222046 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 0.0009999998146668077 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 492.87it/s +Validation metrics: [{'adult_val_loss': 0.31735482811927795, 'lr': 0.0009999998146668077, 'adult_val_accuracy': 0.8470367193222046, 'adult_val_precision': 0.8513215184211731, 'adult_val_recall': 0.8470367193222046, 'adult_val_f1-score': 0.848872184753418, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9130023717880249, 'adult_val_score': 0.8470367193222046}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7205 samples (73.98%) + Class 1: 2534 samples (26.02%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8475202918052673 │ +│ adult_test_f1-score │ 0.8490062952041626 │ +│ adult_test_loss │ 0.3306499421596527 │ +│ adult_test_precision │ 0.8509243130683899 │ +│ adult_test_recall │ 0.8475202918052673 │ +│ adult_test_roc_auc │ 0.905820906162262 │ +│ adult_test_score │ 0.8475202918052673 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 0.0009999996982514858 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 467.35it/s +Test metrics: [{'adult_test_loss': 0.3306499421596527, 'lr': 0.0009999996982514858, 'adult_test_accuracy': 0.8475202918052673, 'adult_test_precision': 0.8509243130683899, 'adult_test_recall': 0.8475202918052673, 'adult_test_f1-score': 0.8490062952041626, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.905820906162262, 'adult_test_score': 0.8475202918052673}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpc5lu6ha6/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpc5lu6ha6/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpc5lu6ha6/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpc5lu6ha6/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ adult │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─Linear: 2-2 [128, 14] [128, 128] 1,920 -- 245,760 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +===================================================================================================================================================================== +Total params: 31,375 +Trainable params: 31,375 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 4.02 +===================================================================================================================================================================== +Input size (MB): 0.46 +Forward/backward pass size (MB): 0.41 +Params size (MB): 0.13 +Estimated Total Size (MB): 0.99 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 31.2 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 31.4 K +Non-trainable params: 0 +Total params: 31.4 K +Total estimated model params size (MB): 0.126 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:01 • 0:00:00 193.60it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8589315414428711 │ +│ adult_val_f1-score │ 0.8508283495903015 │ +│ adult_val_loss │ 0.3108050227165222 │ +│ adult_val_precision │ 0.8527976870536804 │ +│ adult_val_recall │ 0.8589315414428711 │ +│ adult_val_roc_auc │ 0.9120212197303772 │ +│ adult_val_score │ 0.8589315414428711 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 0.0009999998146668077 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 296.60it/s +Validation metrics: [{'adult_val_loss': 0.3108050227165222, 'lr': 0.0009999998146668077, 'adult_val_accuracy': 0.8589315414428711, 'adult_val_precision': 0.8527976870536804, 'adult_val_recall': 0.8589315414428711, 'adult_val_f1-score': 0.8508283495903015, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9120212197303772, 'adult_val_score': 0.8589315414428711}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7947 samples (81.60%) + Class 1: 1792 samples (18.40%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8518328666687012 │ +│ adult_test_f1-score │ 0.8441458344459534 │ +│ adult_test_loss │ 0.3234279751777649 │ +│ adult_test_precision │ 0.8452309370040894 │ +│ adult_test_recall │ 0.8518328666687012 │ +│ adult_test_roc_auc │ 0.9069936871528625 │ +│ adult_test_score │ 0.8518328666687012 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 0.0009999996982514858 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 299.25it/s +Test metrics: [{'adult_test_loss': 0.3234279751777649, 'lr': 0.0009999996982514858, 'adult_test_accuracy': 0.8518328666687012, 'adult_test_precision': 0.8452309370040894, 'adult_test_recall': 0.8518328666687012, 'adult_test_f1-score': 0.8441458344459534, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.9069936871528625, 'adult_test_score': 0.8518328666687012}] +Wrote results for mlp on adult to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/downstream_results/mlp_adult_downstream_results.json +Running downstream for dataset 'adult', model 'dcnv2'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpr9qbkxni/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpr9qbkxni/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpr9qbkxni/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpr9qbkxni/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 13450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ adult │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─EncoderDCNv2: 2-2 [128, 14] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 14] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 14] [128, 128] 1,920 -- 245,760 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 130,191 +Trainable params: 130,191 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 16.66 +========================================================================================================================================================================== +Input size (MB): 0.46 +Forward/backward pass size (MB): 1.06 +Params size (MB): 0.52 +Estimated Total Size (MB): 2.04 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 130 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 130 K +Non-trainable params: 0 +Total params: 130 K +Total estimated model params size (MB): 0.521 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 20/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:02 • 0:00:00 127.48it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8576794862747192 │ +│ adult_val_f1-score │ 0.8511360883712769 │ +│ adult_val_loss │ 0.3154257535934448 │ +│ adult_val_precision │ 0.8512991666793823 │ +│ adult_val_recall │ 0.8576794862747192 │ +│ adult_val_roc_auc │ 0.9099377989768982 │ +│ adult_val_score │ 0.8576794862747192 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 0.000624345033429563 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 224.78it/s +Validation metrics: [{'adult_val_loss': 0.3154257535934448, 'lr': 0.000624345033429563, 'adult_val_accuracy': 0.8576794862747192, 'adult_val_precision': 0.8512991666793823, 'adult_val_recall': 0.8576794862747192, 'adult_val_f1-score': 0.8511360883712769, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9099377989768982, 'adult_val_score': 0.8576794862747192}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7911 samples (81.23%) + Class 1: 1828 samples (18.77%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8479310274124146 │ +│ adult_test_f1-score │ 0.8406109809875488 │ +│ adult_test_loss │ 0.3306029140949249 │ +│ adult_test_precision │ 0.8408699631690979 │ +│ adult_test_recall │ 0.8479310274124146 │ +│ adult_test_roc_auc │ 0.9038339853286743 │ +│ adult_test_score │ 0.8479310274124146 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 0.0006243446841835976 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 284.48it/s +Test metrics: [{'adult_test_loss': 0.3306029140949249, 'lr': 0.0006243446841835976, 'adult_test_accuracy': 0.8479310274124146, 'adult_test_precision': 0.8408699631690979, 'adult_test_recall': 0.8479310274124146, 'adult_test_f1-score': 0.8406109809875488, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.9038339853286743, 'adult_test_score': 0.8479310274124146}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpr9qbkxni/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpr9qbkxni/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpr9qbkxni/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpr9qbkxni/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 13450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ adult │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─EncoderDCNv2: 2-2 [128, 14] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 14] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 14] [128, 128] 1,920 -- 245,760 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 130,191 +Trainable params: 130,191 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 16.66 +========================================================================================================================================================================== +Input size (MB): 0.46 +Forward/backward pass size (MB): 1.06 +Params size (MB): 0.52 +Estimated Total Size (MB): 2.04 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 130 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 130 K +Non-trainable params: 0 +Total params: 130 K +Total estimated model params size (MB): 0.521 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 34/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:01 • 0:00:00 156.30it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.855592668056488 │ +│ adult_val_f1-score │ 0.8506670594215393 │ +│ adult_val_loss │ 0.3332693874835968 │ +│ adult_val_precision │ 0.8495977520942688 │ +│ adult_val_recall │ 0.855592668056488 │ +│ adult_val_roc_auc │ 0.9069664478302002 │ +│ adult_val_score │ 0.855592668056488 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 0.00020610730280168355 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 346.26it/s +Validation metrics: [{'adult_val_loss': 0.3332693874835968, 'lr': 0.00020610730280168355, 'adult_val_accuracy': 0.855592668056488, 'adult_val_precision': 0.8495977520942688, 'adult_val_recall': 0.855592668056488, 'adult_val_f1-score': 0.8506670594215393, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9069664478302002, 'adult_val_score': 0.855592668056488}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7730 samples (79.37%) + Class 1: 2009 samples (20.63%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8478283286094666 │ +│ adult_test_f1-score │ 0.8431715369224548 │ +│ adult_test_loss │ 0.36531636118888855 │ +│ adult_test_precision │ 0.8418262600898743 │ +│ adult_test_recall │ 0.8478283286094666 │ +│ adult_test_roc_auc │ 0.8969647288322449 │ +│ adult_test_score │ 0.8478283286094666 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 0.00020610728824976832 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 388.55it/s +Test metrics: [{'adult_test_loss': 0.36531636118888855, 'lr': 0.00020610728824976832, 'adult_test_accuracy': 0.8478283286094666, 'adult_test_precision': 0.8418262600898743, 'adult_test_recall': 0.8478283286094666, 'adult_test_f1-score': 0.8431715369224548, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.8969647288322449, 'adult_test_score': 0.8478283286094666}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpr9qbkxni/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpr9qbkxni/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpr9qbkxni/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpr9qbkxni/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 13450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ adult │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─EncoderDCNv2: 2-2 [128, 14] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 14] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 14] [128, 128] 1,920 -- 245,760 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 130,191 +Trainable params: 130,191 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 16.66 +========================================================================================================================================================================== +Input size (MB): 0.46 +Forward/backward pass size (MB): 1.06 +Params size (MB): 0.52 +Estimated Total Size (MB): 2.04 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 130 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 130 K +Non-trainable params: 0 +Total params: 130 K +Total estimated model params size (MB): 0.521 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 33/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:02 • 0:00:00 132.84it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8572620749473572 │ +│ adult_val_f1-score │ 0.8533260822296143 │ +│ adult_val_loss │ 0.33715713024139404 │ +│ adult_val_precision │ 0.8519992232322693 │ +│ adult_val_recall │ 0.8572620749473572 │ +│ adult_val_roc_auc │ 0.9056263566017151 │ +│ adult_val_score │ 0.8572620749473572 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 0.0002320865314686671 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 196.22it/s +Validation metrics: [{'adult_val_loss': 0.33715713024139404, 'lr': 0.0002320865314686671, 'adult_val_accuracy': 0.8572620749473572, 'adult_val_precision': 0.8519992232322693, 'adult_val_recall': 0.8572620749473572, 'adult_val_f1-score': 0.8533260822296143, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9056263566017151, 'adult_val_score': 0.8572620749473572}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7711 samples (79.18%) + Class 1: 2028 samples (20.82%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8520382046699524 │ +│ adult_test_f1-score │ 0.8477650284767151 │ +│ adult_test_loss │ 0.36276647448539734 │ +│ adult_test_precision │ 0.8464993238449097 │ +│ adult_test_recall │ 0.8520382046699524 │ +│ adult_test_roc_auc │ 0.900492250919342 │ +│ adult_test_score │ 0.8520382046699524 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 0.00023208664788398892 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 171.09it/s +Test metrics: [{'adult_test_loss': 0.36276647448539734, 'lr': 0.00023208664788398892, 'adult_test_accuracy': 0.8520382046699524, 'adult_test_precision': 0.8464993238449097, 'adult_test_recall': 0.8520382046699524, 'adult_test_f1-score': 0.8477650284767151, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.900492250919342, 'adult_test_score': 0.8520382046699524}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpr9qbkxni/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpr9qbkxni/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpr9qbkxni/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpr9qbkxni/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 13450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ adult │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─EncoderDCNv2: 2-2 [128, 14] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 14] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 14] [128, 128] 1,920 -- 245,760 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 130,191 +Trainable params: 130,191 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 16.66 +========================================================================================================================================================================== +Input size (MB): 0.46 +Forward/backward pass size (MB): 1.06 +Params size (MB): 0.52 +Estimated Total Size (MB): 2.04 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 130 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 130 K +Non-trainable params: 0 +Total params: 130 K +Total estimated model params size (MB): 0.521 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 22/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:02 • 0:00:00 122.43it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8572620749473572 │ +│ adult_val_f1-score │ 0.8545712828636169 │ +│ adult_val_loss │ 0.3125121295452118 │ +│ adult_val_precision │ 0.8531542420387268 │ +│ adult_val_recall │ 0.8572620749473572 │ +│ adult_val_roc_auc │ 0.911125123500824 │ +│ adult_val_score │ 0.8572620749473572 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 0.0005626665661111474 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 216.02it/s +Validation metrics: [{'adult_val_loss': 0.3125121295452118, 'lr': 0.0005626665661111474, 'adult_val_accuracy': 0.8572620749473572, 'adult_val_precision': 0.8531542420387268, 'adult_val_recall': 0.8572620749473572, 'adult_val_f1-score': 0.8545712828636169, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.911125123500824, 'adult_val_score': 0.8572620749473572}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7612 samples (78.16%) + Class 1: 2127 samples (21.84%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8554266095161438 │ +│ adult_test_f1-score │ 0.8524986505508423 │ +│ adult_test_loss │ 0.32436859607696533 │ +│ adult_test_precision │ 0.8511304259300232 │ +│ adult_test_recall │ 0.8554266095161438 │ +│ adult_test_roc_auc │ 0.9062830209732056 │ +│ adult_test_score │ 0.8554266095161438 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 0.0005626667407341301 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 197.47it/s +Test metrics: [{'adult_test_loss': 0.32436859607696533, 'lr': 0.0005626667407341301, 'adult_test_accuracy': 0.8554266095161438, 'adult_test_precision': 0.8511304259300232, 'adult_test_recall': 0.8554266095161438, 'adult_test_f1-score': 0.8524986505508423, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.9062830209732056, 'adult_test_score': 0.8554266095161438}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpr9qbkxni/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpr9qbkxni/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpr9qbkxni/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpr9qbkxni/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 13450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ adult │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─EncoderDCNv2: 2-2 [128, 14] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 14] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 14] [128, 128] 1,920 -- 245,760 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 130,191 +Trainable params: 130,191 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 16.66 +========================================================================================================================================================================== +Input size (MB): 0.46 +Forward/backward pass size (MB): 1.06 +Params size (MB): 0.52 +Estimated Total Size (MB): 2.04 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 130 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 130 K +Non-trainable params: 0 +Total params: 130 K +Total estimated model params size (MB): 0.521 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 37/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:01 • 0:00:00 154.04it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8580968379974365 │ +│ adult_val_f1-score │ 0.8548644185066223 │ +│ adult_val_loss │ 0.3415168225765228 │ +│ adult_val_precision │ 0.8534587025642395 │ +│ adult_val_recall │ 0.8580968379974365 │ +│ adult_val_roc_auc │ 0.9058709740638733 │ +│ adult_val_score │ 0.8580968379974365 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 0.00013551567099057138 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 386.18it/s +Validation metrics: [{'adult_val_loss': 0.3415168225765228, 'lr': 0.00013551567099057138, 'adult_val_accuracy': 0.8580968379974365, 'adult_val_precision': 0.8534587025642395, 'adult_val_recall': 0.8580968379974365, 'adult_val_f1-score': 0.8548644185066223, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9058709740638733, 'adult_val_score': 0.8580968379974365}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7631 samples (78.36%) + Class 1: 2108 samples (21.64%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8477256298065186 │ +│ adult_test_f1-score │ 0.8443962335586548 │ +│ adult_test_loss │ 0.3774588108062744 │ +│ adult_test_precision │ 0.8428343534469604 │ +│ adult_test_recall │ 0.8477256298065186 │ +│ adult_test_roc_auc │ 0.8972837924957275 │ +│ adult_test_score │ 0.8477256298065186 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 0.00013551574375014752 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 354.91it/s +Test metrics: [{'adult_test_loss': 0.3774588108062744, 'lr': 0.00013551574375014752, 'adult_test_accuracy': 0.8477256298065186, 'adult_test_precision': 0.8428343534469604, 'adult_test_recall': 0.8477256298065186, 'adult_test_f1-score': 0.8443962335586548, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.8972837924957275, 'adult_test_score': 0.8477256298065186}] +Wrote results for dcnv2 on adult to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/downstream_results/dcnv2_adult_downstream_results.json +Running downstream for dataset 'adult', model 'resnet'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpq97zol4o/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpq97zol4o/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpq97zol4o/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpq97zol4o/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─ResNet: 2-2 [128, 14] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 14] [128, 128] 1,920 -- 245,760 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 229,903 +Trainable params: 229,903 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 29.43 +========================================================================================================================================================================== +Input size (MB): 0.46 +Forward/backward pass size (MB): 1.98 +Params size (MB): 0.92 +Estimated Total Size (MB): 3.36 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 229 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 229 K +Non-trainable params: 0 +Total params: 229 K +Total estimated model params size (MB): 0.920 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 42/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:01 • 0:00:00 159.31it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8583055138587952 │ +│ adult_val_f1-score │ 0.849209189414978 │ +│ adult_val_loss │ 0.3041922152042389 │ +│ adult_val_precision │ 0.8523894548416138 │ +│ adult_val_recall │ 0.8583055138587952 │ +│ adult_val_roc_auc │ 0.9153968691825867 │ +│ adult_val_score │ 0.8583055138587952 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 0.00020610730280168355 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 376.90it/s +Validation metrics: [{'adult_val_loss': 0.3041922152042389, 'lr': 0.00020610730280168355, 'adult_val_accuracy': 0.8583055138587952, 'adult_val_precision': 0.8523894548416138, 'adult_val_recall': 0.8583055138587952, 'adult_val_f1-score': 0.849209189414978, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9153968691825867, 'adult_val_score': 0.8583055138587952}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 8053 samples (82.69%) + Class 1: 1686 samples (17.31%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8508060574531555 │ +│ adult_test_f1-score │ 0.8413415551185608 │ +│ adult_test_loss │ 0.3245809078216553 │ +│ adult_test_precision │ 0.8444136381149292 │ +│ adult_test_recall │ 0.8508060574531555 │ +│ adult_test_roc_auc │ 0.9065249562263489 │ +│ adult_test_score │ 0.8508060574531555 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 0.00020610728824976832 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 391.58it/s +Test metrics: [{'adult_test_loss': 0.3245809078216553, 'lr': 0.00020610728824976832, 'adult_test_accuracy': 0.8508060574531555, 'adult_test_precision': 0.8444136381149292, 'adult_test_recall': 0.8508060574531555, 'adult_test_f1-score': 0.8413415551185608, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.9065249562263489, 'adult_test_score': 0.8508060574531555}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpq97zol4o/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpq97zol4o/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpq97zol4o/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpq97zol4o/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─ResNet: 2-2 [128, 14] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 14] [128, 128] 1,920 -- 245,760 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 229,903 +Trainable params: 229,903 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 29.43 +========================================================================================================================================================================== +Input size (MB): 0.46 +Forward/backward pass size (MB): 1.98 +Params size (MB): 0.92 +Estimated Total Size (MB): 3.36 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 229 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 229 K +Non-trainable params: 0 +Total params: 229 K +Total estimated model params size (MB): 0.920 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 47/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:01 • 0:00:00 164.89it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8522537350654602 │ +│ adult_val_f1-score │ 0.8487269282341003 │ +│ adult_val_loss │ 0.3081151247024536 │ +│ adult_val_precision │ 0.8471648693084717 │ +│ adult_val_recall │ 0.8522537350654602 │ +│ adult_val_roc_auc │ 0.9119558930397034 │ +│ adult_val_score │ 0.8522537350654602 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 9.54915230977349e-05 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 384.95it/s +Validation metrics: [{'adult_val_loss': 0.3081151247024536, 'lr': 9.54915230977349e-05, 'adult_val_accuracy': 0.8522537350654602, 'adult_val_precision': 0.8471648693084717, 'adult_val_recall': 0.8522537350654602, 'adult_val_f1-score': 0.8487269282341003, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9119558930397034, 'adult_val_score': 0.8522537350654602}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7675 samples (78.81%) + Class 1: 2064 samples (21.19%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8495738506317139 │ +│ adult_test_f1-score │ 0.8457114100456238 │ +│ adult_test_loss │ 0.3286660611629486 │ +│ adult_test_precision │ 0.8442495465278625 │ +│ adult_test_recall │ 0.8495738506317139 │ +│ adult_test_roc_auc │ 0.9046452641487122 │ +│ adult_test_score │ 0.8495738506317139 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 9.54914212343283e-05 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 396.31it/s +Test metrics: [{'adult_test_loss': 0.3286660611629486, 'lr': 9.54914212343283e-05, 'adult_test_accuracy': 0.8495738506317139, 'adult_test_precision': 0.8442495465278625, 'adult_test_recall': 0.8495738506317139, 'adult_test_f1-score': 0.8457114100456238, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.9046452641487122, 'adult_test_score': 0.8495738506317139}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpq97zol4o/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpq97zol4o/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpq97zol4o/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpq97zol4o/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─ResNet: 2-2 [128, 14] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 14] [128, 128] 1,920 -- 245,760 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 229,903 +Trainable params: 229,903 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 29.43 +========================================================================================================================================================================== +Input size (MB): 0.46 +Forward/backward pass size (MB): 1.98 +Params size (MB): 0.92 +Estimated Total Size (MB): 3.36 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 229 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 229 K +Non-trainable params: 0 +Total params: 229 K +Total estimated model params size (MB): 0.920 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 47/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:01 • 0:00:00 156.00it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8564273715019226 │ +│ adult_val_f1-score │ 0.8522515296936035 │ +│ adult_val_loss │ 0.31368014216423035 │ +│ adult_val_precision │ 0.8509513735771179 │ +│ adult_val_recall │ 0.8564273715019226 │ +│ adult_val_roc_auc │ 0.9105032086372375 │ +│ adult_val_score │ 0.8564273715019226 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 9.54915230977349e-05 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 236.68it/s +Validation metrics: [{'adult_val_loss': 0.31368014216423035, 'lr': 9.54915230977349e-05, 'adult_val_accuracy': 0.8564273715019226, 'adult_val_precision': 0.8509513735771179, 'adult_val_recall': 0.8564273715019226, 'adult_val_f1-score': 0.8522515296936035, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9105032086372375, 'adult_val_score': 0.8564273715019226}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7727 samples (79.34%) + Class 1: 2012 samples (20.66%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8540918231010437 │ +│ adult_test_f1-score │ 0.8496665954589844 │ +│ adult_test_loss │ 0.3249024748802185 │ +│ adult_test_precision │ 0.848552405834198 │ +│ adult_test_recall │ 0.8540918231010437 │ +│ adult_test_roc_auc │ 0.9066640138626099 │ +│ adult_test_score │ 0.8540918231010437 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 9.54914212343283e-05 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 240.41it/s +Test metrics: [{'adult_test_loss': 0.3249024748802185, 'lr': 9.54914212343283e-05, 'adult_test_accuracy': 0.8540918231010437, 'adult_test_precision': 0.848552405834198, 'adult_test_recall': 0.8540918231010437, 'adult_test_f1-score': 0.8496665954589844, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.9066640138626099, 'adult_test_score': 0.8540918231010437}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpq97zol4o/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpq97zol4o/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpq97zol4o/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpq97zol4o/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─ResNet: 2-2 [128, 14] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 14] [128, 128] 1,920 -- 245,760 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 229,903 +Trainable params: 229,903 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 29.43 +========================================================================================================================================================================== +Input size (MB): 0.46 +Forward/backward pass size (MB): 1.98 +Params size (MB): 0.92 +Estimated Total Size (MB): 3.36 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 229 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 229 K +Non-trainable params: 0 +Total params: 229 K +Total estimated model params size (MB): 0.920 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:01 • 0:00:00 136.42it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8572620749473572 │ +│ adult_val_f1-score │ 0.8518243432044983 │ +│ adult_val_loss │ 0.31039074063301086 │ +│ adult_val_precision │ 0.8511360883712769 │ +│ adult_val_recall │ 0.8572620749473572 │ +│ adult_val_roc_auc │ 0.9105858206748962 │ +│ adult_val_score │ 0.8572620749473572 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 266.57it/s +Validation metrics: [{'adult_val_loss': 0.31039074063301086, 'lr': 0.0, 'adult_val_accuracy': 0.8572620749473572, 'adult_val_precision': 0.8511360883712769, 'adult_val_recall': 0.8572620749473572, 'adult_val_f1-score': 0.8518243432044983, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9105858206748962, 'adult_val_score': 0.8572620749473572}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7843 samples (80.53%) + Class 1: 1896 samples (19.47%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8510113954544067 │ +│ adult_test_f1-score │ 0.844857931137085 │ +│ adult_test_loss │ 0.3238856792449951 │ +│ adult_test_precision │ 0.8444961905479431 │ +│ adult_test_recall │ 0.8510113954544067 │ +│ adult_test_roc_auc │ 0.9052449464797974 │ +│ adult_test_score │ 0.8510113954544067 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 269.60it/s +Test metrics: [{'adult_test_loss': 0.3238856792449951, 'lr': 0.0, 'adult_test_accuracy': 0.8510113954544067, 'adult_test_precision': 0.8444961905479431, 'adult_test_recall': 0.8510113954544067, 'adult_test_f1-score': 0.844857931137085, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.9052449464797974, 'adult_test_score': 0.8510113954544067}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpq97zol4o/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpq97zol4o/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpq97zol4o/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpq97zol4o/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─ResNet: 2-2 [128, 14] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 14] [128, 128] 1,920 -- 245,760 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 229,903 +Trainable params: 229,903 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 29.43 +========================================================================================================================================================================== +Input size (MB): 0.46 +Forward/backward pass size (MB): 1.98 +Params size (MB): 0.92 +Estimated Total Size (MB): 3.36 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 229 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 229 K +Non-trainable params: 0 +Total params: 229 K +Total estimated model params size (MB): 0.920 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 46/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:01 • 0:00:00 155.53it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8597663044929504 │ +│ adult_val_f1-score │ 0.8529577851295471 │ +│ adult_val_loss │ 0.3099907338619232 │ +│ adult_val_precision │ 0.8535987138748169 │ +│ adult_val_recall │ 0.8597663044929504 │ +│ adult_val_roc_auc │ 0.9129458665847778 │ +│ adult_val_score │ 0.8597663044929504 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 0.0007938927737995982 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 252.75it/s +Validation metrics: [{'adult_val_loss': 0.3099907338619232, 'lr': 0.0007938927737995982, 'adult_val_accuracy': 0.8597663044929504, 'adult_val_precision': 0.8535987138748169, 'adult_val_recall': 0.8597663044929504, 'adult_val_f1-score': 0.8529577851295471, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9129458665847778, 'adult_val_score': 0.8597663044929504}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7955 samples (81.68%) + Class 1: 1784 samples (18.32%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8553239703178406 │ +│ adult_test_f1-score │ 0.8476958870887756 │ +│ adult_test_loss │ 0.3275333046913147 │ +│ adult_test_precision │ 0.8491895198822021 │ +│ adult_test_recall │ 0.8553239703178406 │ +│ adult_test_roc_auc │ 0.9070183038711548 │ +│ adult_test_score │ 0.8553239703178406 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 0.0007938928320072591 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 254.63it/s +Test metrics: [{'adult_test_loss': 0.3275333046913147, 'lr': 0.0007938928320072591, 'adult_test_accuracy': 0.8553239703178406, 'adult_test_precision': 0.8491895198822021, 'adult_test_recall': 0.8553239703178406, 'adult_test_f1-score': 0.8476958870887756, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.9070183038711548, 'adult_test_score': 0.8553239703178406}] +Wrote results for resnet on adult to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/downstream_results/resnet_adult_downstream_results.json +Running downstream for dataset 'adult', model 'autoint'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp23hefb1m/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=0 --np_seed=0 --output_file=/tmp/tmp23hefb1m/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp23hefb1m/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp23hefb1m/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 13450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ adult │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─EncoderAutoIntBase: 2-2 [128, 14] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 14] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 14] [128, 14, 128] 1,792 -- 229,376 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 14, 14] [256, 14, 14] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 14, 128] [128, 14, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 14, 128] [128, 14, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 14, 14] [256, 14, 14] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 14, 128] [128, 14, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 14, 128] [128, 14, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 14, 14] [256, 14, 14] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 14, 128] [128, 14, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 14, 128] [128, 14, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 1792] [128, 128] 229,504 -- 29,376,512 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +==================================================================================================================================================================================== +Total params: 442,511 +Trainable params: 442,511 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 56.64 +==================================================================================================================================================================================== +Input size (MB): 0.46 +Forward/backward pass size (MB): 29.51 +Params size (MB): 1.77 +Estimated Total Size (MB): 31.74 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 442 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 442 K +Non-trainable params: 0 +Total params: 442 K +Total estimated model params size (MB): 1.770 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 41/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:02 • 0:00:00 99.82it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8568447232246399 │ +│ adult_val_f1-score │ 0.8530578017234802 │ +│ adult_val_loss │ 0.3129625618457794 │ +│ adult_val_precision │ 0.8516823649406433 │ +│ adult_val_recall │ 0.8568447232246399 │ +│ adult_val_roc_auc │ 0.911578357219696 │ +│ adult_val_score │ 0.8568447232246399 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 6.184666563058272e-05 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 196.16it/s +Validation metrics: [{'adult_val_loss': 0.3129625618457794, 'lr': 6.184666563058272e-05, 'adult_val_accuracy': 0.8568447232246399, 'adult_val_precision': 0.8516823649406433, 'adult_val_recall': 0.8568447232246399, 'adult_val_f1-score': 0.8530578017234802, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.911578357219696, 'adult_val_score': 0.8568447232246399}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7679 samples (78.85%) + Class 1: 2060 samples (21.15%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.855734646320343 │ +│ adult_test_f1-score │ 0.8519795536994934 │ +│ adult_test_loss │ 0.32497116923332214 │ +│ adult_test_precision │ 0.8507261872291565 │ +│ adult_test_recall │ 0.855734646320343 │ +│ adult_test_roc_auc │ 0.9085766673088074 │ +│ adult_test_score │ 0.855734646320343 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 6.18466583546251e-05 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 204.91it/s +Test metrics: [{'adult_test_loss': 0.32497116923332214, 'lr': 6.18466583546251e-05, 'adult_test_accuracy': 0.855734646320343, 'adult_test_precision': 0.8507261872291565, 'adult_test_recall': 0.855734646320343, 'adult_test_f1-score': 0.8519795536994934, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.9085766673088074, 'adult_test_score': 0.855734646320343}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp23hefb1m/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=1 --np_seed=1 --output_file=/tmp/tmp23hefb1m/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp23hefb1m/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp23hefb1m/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 13450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ adult │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─EncoderAutoIntBase: 2-2 [128, 14] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 14] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 14] [128, 14, 128] 1,792 -- 229,376 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 14, 14] [256, 14, 14] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 14, 128] [128, 14, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 14, 128] [128, 14, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 14, 14] [256, 14, 14] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 14, 128] [128, 14, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 14, 128] [128, 14, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 14, 14] [256, 14, 14] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 14, 128] [128, 14, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 14, 128] [128, 14, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 1792] [128, 128] 229,504 -- 29,376,512 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +==================================================================================================================================================================================== +Total params: 442,511 +Trainable params: 442,511 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 56.64 +==================================================================================================================================================================================== +Input size (MB): 0.46 +Forward/backward pass size (MB): 29.51 +Params size (MB): 1.77 +Estimated Total Size (MB): 31.74 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 442 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 442 K +Non-trainable params: 0 +Total params: 442 K +Total estimated model params size (MB): 1.770 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 35/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:02 • 0:00:00 103.60it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8626878261566162 │ +│ adult_val_f1-score │ 0.8564330339431763 │ +│ adult_val_loss │ 0.3086968958377838 │ +│ adult_val_precision │ 0.8568668961524963 │ +│ adult_val_recall │ 0.8626878261566162 │ +│ adult_val_roc_auc │ 0.9119505882263184 │ +│ adult_val_score │ 0.8626878261566162 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 0.0001812879927456379 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 198.93it/s +Validation metrics: [{'adult_val_loss': 0.3086968958377838, 'lr': 0.0001812879927456379, 'adult_val_accuracy': 0.8626878261566162, 'adult_val_precision': 0.8568668961524963, 'adult_val_recall': 0.8626878261566162, 'adult_val_f1-score': 0.8564330339431763, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9119505882263184, 'adult_val_score': 0.8626878261566162}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7870 samples (80.81%) + Class 1: 1869 samples (19.19%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8570694923400879 │ +│ adult_test_f1-score │ 0.8507835865020752 │ +│ adult_test_loss │ 0.3204444944858551 │ +│ adult_test_precision │ 0.8510997891426086 │ +│ adult_test_recall │ 0.8570694923400879 │ +│ adult_test_roc_auc │ 0.9077580571174622 │ +│ adult_test_score │ 0.8570694923400879 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 0.0001812881964724511 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 205.75it/s +Test metrics: [{'adult_test_loss': 0.3204444944858551, 'lr': 0.0001812881964724511, 'adult_test_accuracy': 0.8570694923400879, 'adult_test_precision': 0.8510997891426086, 'adult_test_recall': 0.8570694923400879, 'adult_test_f1-score': 0.8507835865020752, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.9077580571174622, 'adult_test_score': 0.8570694923400879}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp23hefb1m/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=2 --np_seed=2 --output_file=/tmp/tmp23hefb1m/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp23hefb1m/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp23hefb1m/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 13450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ adult │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─EncoderAutoIntBase: 2-2 [128, 14] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 14] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 14] [128, 14, 128] 1,792 -- 229,376 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 14, 14] [256, 14, 14] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 14, 128] [128, 14, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 14, 128] [128, 14, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 14, 14] [256, 14, 14] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 14, 128] [128, 14, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 14, 128] [128, 14, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 14, 14] [256, 14, 14] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 14, 128] [128, 14, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 14, 128] [128, 14, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 1792] [128, 128] 229,504 -- 29,376,512 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +==================================================================================================================================================================================== +Total params: 442,511 +Trainable params: 442,511 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 56.64 +==================================================================================================================================================================================== +Input size (MB): 0.46 +Forward/backward pass size (MB): 29.51 +Params size (MB): 1.77 +Estimated Total Size (MB): 31.74 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 442 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 442 K +Non-trainable params: 0 +Total params: 442 K +Total estimated model params size (MB): 1.770 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 45/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:02 • 0:00:00 105.22it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8558013439178467 │ +│ adult_val_f1-score │ 0.8512482643127441 │ +│ adult_val_loss │ 0.3128451406955719 │ +│ adult_val_precision │ 0.8500397205352783 │ +│ adult_val_recall │ 0.8558013439178467 │ +│ adult_val_roc_auc │ 0.9117394089698792 │ +│ adult_val_score │ 0.8558013439178467 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 1.570841959619429e-05 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 301.68it/s +Validation metrics: [{'adult_val_loss': 0.3128451406955719, 'lr': 1.570841959619429e-05, 'adult_val_accuracy': 0.8558013439178467, 'adult_val_precision': 0.8500397205352783, 'adult_val_recall': 0.8558013439178467, 'adult_val_f1-score': 0.8512482643127441, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9117394089698792, 'adult_val_score': 0.8558013439178467}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7748 samples (79.56%) + Class 1: 1991 samples (20.44%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8523462414741516 │ +│ adult_test_f1-score │ 0.8475839495658875 │ +│ adult_test_loss │ 0.33015069365501404 │ +│ adult_test_precision │ 0.846514880657196 │ +│ adult_test_recall │ 0.8523462414741516 │ +│ adult_test_roc_auc │ 0.9068588614463806 │ +│ adult_test_score │ 0.8523462414741516 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 1.5708421415183693e-05 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 297.83it/s +Test metrics: [{'adult_test_loss': 0.33015069365501404, 'lr': 1.5708421415183693e-05, 'adult_test_accuracy': 0.8523462414741516, 'adult_test_precision': 0.846514880657196, 'adult_test_recall': 0.8523462414741516, 'adult_test_f1-score': 0.8475839495658875, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.9068588614463806, 'adult_test_score': 0.8523462414741516}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp23hefb1m/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=3 --np_seed=3 --output_file=/tmp/tmp23hefb1m/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp23hefb1m/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp23hefb1m/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 13450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ adult │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─EncoderAutoIntBase: 2-2 [128, 14] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 14] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 14] [128, 14, 128] 1,792 -- 229,376 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 14, 14] [256, 14, 14] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 14, 128] [128, 14, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 14, 128] [128, 14, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 14, 14] [256, 14, 14] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 14, 128] [128, 14, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 14, 128] [128, 14, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 14, 14] [256, 14, 14] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 14, 128] [128, 14, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 14, 128] [128, 14, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 1792] [128, 128] 229,504 -- 29,376,512 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +==================================================================================================================================================================================== +Total params: 442,511 +Trainable params: 442,511 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 56.64 +==================================================================================================================================================================================== +Input size (MB): 0.46 +Forward/backward pass size (MB): 29.51 +Params size (MB): 1.77 +Estimated Total Size (MB): 31.74 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 442 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 442 K +Non-trainable params: 0 +Total params: 442 K +Total estimated model params size (MB): 1.770 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 42/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:02 • 0:00:00 108.52it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8583055138587952 │ +│ adult_val_f1-score │ 0.852821409702301 │ +│ adult_val_loss │ 0.3171083331108093 │ +│ adult_val_precision │ 0.8522410988807678 │ +│ adult_val_recall │ 0.8583055138587952 │ +│ adult_val_roc_auc │ 0.9097170233726501 │ +│ adult_val_score │ 0.8583055138587952 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 4.758648356073536e-05 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 306.06it/s +Validation metrics: [{'adult_val_loss': 0.3171083331108093, 'lr': 4.758648356073536e-05, 'adult_val_accuracy': 0.8583055138587952, 'adult_val_precision': 0.8522410988807678, 'adult_val_recall': 0.8583055138587952, 'adult_val_f1-score': 0.852821409702301, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9097170233726501, 'adult_val_score': 0.8583055138587952}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7812 samples (80.21%) + Class 1: 1927 samples (19.79%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8550159335136414 │ +│ adult_test_f1-score │ 0.849465012550354 │ +│ adult_test_loss │ 0.3377856910228729 │ +│ adult_test_precision │ 0.8490124344825745 │ +│ adult_test_recall │ 0.8550159335136414 │ +│ adult_test_roc_auc │ 0.9036782383918762 │ +│ adult_test_score │ 0.8550159335136414 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 4.758644354296848e-05 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 310.69it/s +Test metrics: [{'adult_test_loss': 0.3377856910228729, 'lr': 4.758644354296848e-05, 'adult_test_accuracy': 0.8550159335136414, 'adult_test_precision': 0.8490124344825745, 'adult_test_recall': 0.8550159335136414, 'adult_test_f1-score': 0.849465012550354, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.9036782383918762, 'adult_test_score': 0.8550159335136414}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp23hefb1m/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=4 --np_seed=4 --output_file=/tmp/tmp23hefb1m/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp23hefb1m/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp23hefb1m/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 13450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ adult │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─EncoderAutoIntBase: 2-2 [128, 14] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 14] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 14] [128, 14, 128] 1,792 -- 229,376 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 14, 14] [256, 14, 14] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 14, 128] [128, 14, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 14, 128] [128, 14, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 14, 14] [256, 14, 14] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 14, 128] [128, 14, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 14, 128] [128, 14, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 14, 128] [128, 14, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 14, 14] [256, 14, 14] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 14, 128] [128, 14, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 14, 128] [128, 14, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 14, 128] [128, 14, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 1792] [128, 128] 229,504 -- 29,376,512 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +==================================================================================================================================================================================== +Total params: 442,511 +Trainable params: 442,511 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 56.64 +==================================================================================================================================================================================== +Input size (MB): 0.46 +Forward/backward pass size (MB): 29.51 +Params size (MB): 1.77 +Estimated Total Size (MB): 31.74 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 442 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 442 K +Non-trainable params: 0 +Total params: 442 K +Total estimated model params size (MB): 1.770 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 44/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:02 • 0:00:00 101.35it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8628965020179749 │ +│ adult_val_f1-score │ 0.858567476272583 │ +│ adult_val_loss │ 0.314855694770813 │ +│ adult_val_precision │ 0.8576472997665405 │ +│ adult_val_recall │ 0.8628965020179749 │ +│ adult_val_roc_auc │ 0.9121567606925964 │ +│ adult_val_score │ 0.8628965020179749 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 2.4471735741826706e-05 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 203.62it/s +Validation metrics: [{'adult_val_loss': 0.314855694770813, 'lr': 2.4471735741826706e-05, 'adult_val_accuracy': 0.8628965020179749, 'adult_val_precision': 0.8576472997665405, 'adult_val_recall': 0.8628965020179749, 'adult_val_f1-score': 0.858567476272583, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9121567606925964, 'adult_val_score': 0.8628965020179749}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7781 samples (79.90%) + Class 1: 1958 samples (20.10%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8518328666687012 │ +│ adult_test_f1-score │ 0.8465977907180786 │ +│ adult_test_loss │ 0.33142223954200745 │ +│ adult_test_precision │ 0.845725417137146 │ +│ adult_test_recall │ 0.8518328666687012 │ +│ adult_test_roc_auc │ 0.9062855839729309 │ +│ adult_test_score │ 0.8518328666687012 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 2.4471757569699548e-05 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 207.76it/s +Test metrics: [{'adult_test_loss': 0.33142223954200745, 'lr': 2.4471757569699548e-05, 'adult_test_accuracy': 0.8518328666687012, 'adult_test_precision': 0.845725417137146, 'adult_test_recall': 0.8518328666687012, 'adult_test_f1-score': 0.8465977907180786, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.9062855839729309, 'adult_test_score': 0.8518328666687012}] +Wrote results for autoint on adult to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/downstream_results/autoint_adult_downstream_results.json +Running downstream for dataset 'adult', model 'fttransformer'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpe2iufojv/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpe2iufojv/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpe2iufojv/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpe2iufojv/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 13450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ adult │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─EncoderFTTransformer: 2-2 [128, 14] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 14] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 14] [128, 14, 128] 3,584 -- 458,752 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 15, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 15, 15] [512, 15, 15] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 15, 128] [128, 15, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 15, 512] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 15, 256] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 15, 256] [128, 15, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 15, 15] [512, 15, 15] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 15, 128] [128, 15, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 15, 512] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 15, 256] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 15, 256] [128, 15, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 15] [512, 1, 15] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 15, 128] [128, 15, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 15, 512] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 15, 256] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 15, 256] [128, 15, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================================= +Total params: 529,423 +Trainable params: 529,423 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 67.77 +========================================================================================================================================================================================= +Input size (MB): 0.46 +Forward/backward pass size (MB): 61.49 +Params size (MB): 2.12 +Estimated Total Size (MB): 64.06 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 529 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 529 K +Non-trainable params: 0 +Total params: 529 K +Total estimated model params size (MB): 2.118 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:03 • 0:00:00 85.65it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8572620749473572 │ +│ adult_val_f1-score │ 0.8521676659584045 │ +│ adult_val_loss │ 0.3083151876926422 │ +│ adult_val_precision │ 0.8512853384017944 │ +│ adult_val_recall │ 0.8572620749473572 │ +│ adult_val_roc_auc │ 0.913361132144928 │ +│ adult_val_score │ 0.8572620749473572 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 242.35it/s +Validation metrics: [{'adult_val_loss': 0.3083151876926422, 'lr': 0.0, 'adult_val_accuracy': 0.8572620749473572, 'adult_val_precision': 0.8512853384017944, 'adult_val_recall': 0.8572620749473572, 'adult_val_f1-score': 0.8521676659584045, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.913361132144928, 'adult_val_score': 0.8572620749473572}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7778 samples (79.86%) + Class 1: 1961 samples (20.14%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8548105359077454 │ +│ adult_test_f1-score │ 0.8497217297554016 │ +│ adult_test_loss │ 0.3239251375198364 │ +│ adult_test_precision │ 0.8489660024642944 │ +│ adult_test_recall │ 0.8548105359077454 │ +│ adult_test_roc_auc │ 0.9074366688728333 │ +│ adult_test_score │ 0.8548105359077454 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 231.97it/s +Test metrics: [{'adult_test_loss': 0.3239251375198364, 'lr': 0.0, 'adult_test_accuracy': 0.8548105359077454, 'adult_test_precision': 0.8489660024642944, 'adult_test_recall': 0.8548105359077454, 'adult_test_f1-score': 0.8497217297554016, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.9074366688728333, 'adult_test_score': 0.8548105359077454}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpe2iufojv/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpe2iufojv/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpe2iufojv/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpe2iufojv/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 13450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ adult │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─EncoderFTTransformer: 2-2 [128, 14] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 14] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 14] [128, 14, 128] 3,584 -- 458,752 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 15, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 15, 15] [512, 15, 15] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 15, 128] [128, 15, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 15, 512] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 15, 256] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 15, 256] [128, 15, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 15, 15] [512, 15, 15] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 15, 128] [128, 15, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 15, 512] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 15, 256] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 15, 256] [128, 15, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 15] [512, 1, 15] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 15, 128] [128, 15, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 15, 512] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 15, 256] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 15, 256] [128, 15, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================================= +Total params: 529,423 +Trainable params: 529,423 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 67.77 +========================================================================================================================================================================================= +Input size (MB): 0.46 +Forward/backward pass size (MB): 61.49 +Params size (MB): 2.12 +Estimated Total Size (MB): 64.06 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 529 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 529 K +Non-trainable params: 0 +Total params: 529 K +Total estimated model params size (MB): 2.118 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:03 • 0:00:00 69.81it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8651919960975647 │ +│ adult_val_f1-score │ 0.8615251779556274 │ +│ adult_val_loss │ 0.3046202063560486 │ +│ adult_val_precision │ 0.8604550361633301 │ +│ adult_val_recall │ 0.8651919960975647 │ +│ adult_val_roc_auc │ 0.9156761169433594 │ +│ adult_val_score │ 0.8651919960975647 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 154.59it/s +Validation metrics: [{'adult_val_loss': 0.3046202063560486, 'lr': 0.0, 'adult_val_accuracy': 0.8651919960975647, 'adult_val_precision': 0.8604550361633301, 'adult_val_recall': 0.8651919960975647, 'adult_val_f1-score': 0.8615251779556274, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9156761169433594, 'adult_val_score': 0.8651919960975647}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7748 samples (79.56%) + Class 1: 1991 samples (20.44%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8570694923400879 │ +│ adult_test_f1-score │ 0.8524595499038696 │ +│ adult_test_loss │ 0.3174950182437897 │ +│ adult_test_precision │ 0.8515889048576355 │ +│ adult_test_recall │ 0.8570694923400879 │ +│ adult_test_roc_auc │ 0.9097620248794556 │ +│ adult_test_score │ 0.8570694923400879 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 143.52it/s +Test metrics: [{'adult_test_loss': 0.3174950182437897, 'lr': 0.0, 'adult_test_accuracy': 0.8570694923400879, 'adult_test_precision': 0.8515889048576355, 'adult_test_recall': 0.8570694923400879, 'adult_test_f1-score': 0.8524595499038696, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.9097620248794556, 'adult_test_score': 0.8570694923400879}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpe2iufojv/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpe2iufojv/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpe2iufojv/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpe2iufojv/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 13450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ adult │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─EncoderFTTransformer: 2-2 [128, 14] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 14] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 14] [128, 14, 128] 3,584 -- 458,752 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 15, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 15, 15] [512, 15, 15] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 15, 128] [128, 15, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 15, 512] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 15, 256] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 15, 256] [128, 15, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 15, 15] [512, 15, 15] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 15, 128] [128, 15, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 15, 512] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 15, 256] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 15, 256] [128, 15, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 15] [512, 1, 15] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 15, 128] [128, 15, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 15, 512] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 15, 256] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 15, 256] [128, 15, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================================= +Total params: 529,423 +Trainable params: 529,423 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 67.77 +========================================================================================================================================================================================= +Input size (MB): 0.46 +Forward/backward pass size (MB): 61.49 +Params size (MB): 2.12 +Estimated Total Size (MB): 64.06 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 529 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 529 K +Non-trainable params: 0 +Total params: 529 K +Total estimated model params size (MB): 2.118 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 35/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:03 • 0:00:00 87.41it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8618530631065369 │ +│ adult_val_f1-score │ 0.8542932271957397 │ +│ adult_val_loss │ 0.30751922726631165 │ +│ adult_val_precision │ 0.8560426235198975 │ +│ adult_val_recall │ 0.8618530631065369 │ +│ adult_val_roc_auc │ 0.9131970405578613 │ +│ adult_val_score │ 0.8618530631065369 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 0.0001812879927456379 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 252.96it/s +Validation metrics: [{'adult_val_loss': 0.30751922726631165, 'lr': 0.0001812879927456379, 'adult_val_accuracy': 0.8618530631065369, 'adult_val_precision': 0.8560426235198975, 'adult_val_recall': 0.8618530631065369, 'adult_val_f1-score': 0.8542932271957397, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9131970405578613, 'adult_val_score': 0.8618530631065369}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 8001 samples (82.15%) + Class 1: 1738 samples (17.85%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.852859616279602 │ +│ adult_test_f1-score │ 0.8443744778633118 │ +│ adult_test_loss │ 0.32495948672294617 │ +│ adult_test_precision │ 0.8465269207954407 │ +│ adult_test_recall │ 0.852859616279602 │ +│ adult_test_roc_auc │ 0.9071648120880127 │ +│ adult_test_score │ 0.852859616279602 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 0.0001812881964724511 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 256.53it/s +Test metrics: [{'adult_test_loss': 0.32495948672294617, 'lr': 0.0001812881964724511, 'adult_test_accuracy': 0.852859616279602, 'adult_test_precision': 0.8465269207954407, 'adult_test_recall': 0.852859616279602, 'adult_test_f1-score': 0.8443744778633118, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.9071648120880127, 'adult_test_score': 0.852859616279602}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpe2iufojv/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpe2iufojv/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpe2iufojv/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpe2iufojv/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 13450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ adult │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─EncoderFTTransformer: 2-2 [128, 14] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 14] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 14] [128, 14, 128] 3,584 -- 458,752 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 15, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 15, 15] [512, 15, 15] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 15, 128] [128, 15, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 15, 512] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 15, 256] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 15, 256] [128, 15, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 15, 15] [512, 15, 15] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 15, 128] [128, 15, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 15, 512] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 15, 256] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 15, 256] [128, 15, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 15] [512, 1, 15] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 15, 128] [128, 15, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 15, 512] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 15, 256] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 15, 256] [128, 15, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================================= +Total params: 529,423 +Trainable params: 529,423 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 67.77 +========================================================================================================================================================================================= +Input size (MB): 0.46 +Forward/backward pass size (MB): 61.49 +Params size (MB): 2.12 +Estimated Total Size (MB): 64.06 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 529 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 529 K +Non-trainable params: 0 +Total params: 529 K +Total estimated model params size (MB): 2.118 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 45/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:03 • 0:00:00 86.34it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8603923320770264 │ +│ adult_val_f1-score │ 0.8552144169807434 │ +│ adult_val_loss │ 0.3071005642414093 │ +│ adult_val_precision │ 0.854596734046936 │ +│ adult_val_recall │ 0.8603923320770264 │ +│ adult_val_roc_auc │ 0.9135773777961731 │ +│ adult_val_score │ 0.8603923320770264 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 1.570841959619429e-05 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 237.46it/s +Validation metrics: [{'adult_val_loss': 0.3071005642414093, 'lr': 1.570841959619429e-05, 'adult_val_accuracy': 0.8603923320770264, 'adult_val_precision': 0.854596734046936, 'adult_val_recall': 0.8603923320770264, 'adult_val_f1-score': 0.8552144169807434, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9135773777961731, 'adult_val_score': 0.8603923320770264}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7802 samples (80.11%) + Class 1: 1937 samples (19.89%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.8556320071220398 │ +│ adult_test_f1-score │ 0.8502432107925415 │ +│ adult_test_loss │ 0.3209441900253296 │ +│ adult_test_precision │ 0.8497291803359985 │ +│ adult_test_recall │ 0.8556320071220398 │ +│ adult_test_roc_auc │ 0.9095597267150879 │ +│ adult_test_score │ 0.8556320071220398 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 1.5708421415183693e-05 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 236.53it/s +Test metrics: [{'adult_test_loss': 0.3209441900253296, 'lr': 1.5708421415183693e-05, 'adult_test_accuracy': 0.8556320071220398, 'adult_test_precision': 0.8497291803359985, 'adult_test_recall': 0.8556320071220398, 'adult_test_f1-score': 0.8502432107925415, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.9095597267150879, 'adult_test_score': 0.8556320071220398}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=adult --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpe2iufojv/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpe2iufojv/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ adult │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpe2iufojv/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpe2iufojv/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 14 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 269 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 14, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 14 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 13450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ adult │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 14, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 14, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 14, 64] [128, 14] -- -- -- +│ │ └─Linear: 3-1 [128, 896] [128, 14] 12,558 -- 1,607,424 +│ └─EncoderFTTransformer: 2-2 [128, 14] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 14] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 14] [128, 14, 128] 3,584 -- 458,752 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 15, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 15, 15] [512, 15, 15] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 15, 128] [128, 15, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 15, 512] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 15, 256] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 15, 256] [128, 15, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 15, 15] [512, 15, 15] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 15, 128] [128, 15, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 15, 512] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 15, 256] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 15, 256] [128, 15, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 15, 128] [128, 15, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 15] [512, 1, 15] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 15, 128] [128, 15, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 15, 128] [128, 15, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 15, 512] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 15, 256] [128, 15, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 15, 256] [128, 15, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 15, 128] [128, 15, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================================= +Total params: 529,423 +Trainable params: 529,423 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 67.77 +========================================================================================================================================================================================= +Input size (MB): 0.46 +Forward/backward pass size (MB): 61.49 +Params size (MB): 2.12 +Estimated Total Size (MB): 64.06 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 529 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 529 K +Non-trainable params: 0 +Total params: 529 K +Total estimated model params size (MB): 2.118 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 48/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 269/269 0:00:03 • 0:00:00 88.95it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_val_accuracy │ 0.8589315414428711 │ +│ adult_val_f1-score │ 0.8547211289405823 │ +│ adult_val_loss │ 0.3103259801864624 │ +│ adult_val_precision │ 0.8535484671592712 │ +│ adult_val_recall │ 0.8589315414428711 │ +│ adult_val_roc_auc │ 0.9123598337173462 │ +│ adult_val_score │ 0.8589315414428711 │ +│ adult_val_support │ 4792.0 │ +│ lr │ 9.866357686405536e-07 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38/38 0:00:00 • 0:00:00 228.74it/s +Validation metrics: [{'adult_val_loss': 0.3103259801864624, 'lr': 9.866357686405536e-07, 'adult_val_accuracy': 0.8589315414428711, 'adult_val_precision': 0.8535484671592712, 'adult_val_recall': 0.8589315414428711, 'adult_val_f1-score': 0.8547211289405823, 'adult_val_support': 4792.0, 'adult_val_roc_auc': 0.9123598337173462, 'adult_val_score': 0.8589315414428711}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7350 samples (75.47%) + Class 1.0: 2389 samples (24.53%) +Predicted class frequencies: + Class 0: 7770 samples (79.78%) + Class 1: 1969 samples (20.22%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ adult_test_accuracy │ 0.856042742729187 │ +│ adult_test_f1-score │ 0.8511052131652832 │ +│ adult_test_loss │ 0.32588112354278564 │ +│ adult_test_precision │ 0.8503450155258179 │ +│ adult_test_recall │ 0.856042742729187 │ +│ adult_test_roc_auc │ 0.9073761701583862 │ +│ adult_test_score │ 0.856042742729187 │ +│ adult_test_support │ 9739.0 │ +│ lr │ 9.866350865195272e-07 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77/77 0:00:00 • 0:00:00 231.91it/s +Test metrics: [{'adult_test_loss': 0.32588112354278564, 'lr': 9.866350865195272e-07, 'adult_test_accuracy': 0.856042742729187, 'adult_test_precision': 0.8503450155258179, 'adult_test_recall': 0.856042742729187, 'adult_test_f1-score': 0.8511052131652832, 'adult_test_support': 9739.0, 'adult_test_roc_auc': 0.9073761701583862, 'adult_test_score': 0.856042742729187}] +Wrote results for fttransformer on adult to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/downstream_results/fttransformer_adult_downstream_results.json + +Downstream pipeline complete. Results under /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_182952/downstream_results +[gpu2-adult-bank] Starting bank_marketing on physical GPU 2 at Tue Sep 15 20:39:54 UTC 2026 ++ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py --dataset bank_marketing --results-root /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42 --epochs 500 --objective w_jepa_joint --latent-architecture deterministic_no_z --sw-full-end-weight 0.004 --torch-seed 42 --np-seed 42 --mask-seed 0 --sw-seed 0 --sw-full-num-projections 2048 --sw-marginal-num-projections 1024 --sw-p 2 --sw-s-x-end-weight 0.001 --sw-z-end-weight 0.0 --sw-s-y-end-weight 0.0001 --model-selection unfiltered --deterministic-mode warn --sw-include-marginals --batch-size 512 +{ + "dataset": "bank_marketing", + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "bank_marketing", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.004", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.001", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "0.0001", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} +Not SLURM job. +╒══════════════════════════════════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ mock │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ activation_checkpointing │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 42 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 42 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_seed │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_seed │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ deterministic_mode │ warn │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pre_evaluation │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 512 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 500 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ improvement │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_val_cache_cadence │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ init_type │ trunc_normal │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_start_lr │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_lr │ 0.001 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_final_lr │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 10 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_final_weight_decay │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_ipe_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_gradient_clipping │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_feature_type_embedding │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_feature_index_embedding │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dropout_prob │ 0.001 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_layer_norm_eps │ 1e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dim_hidden │ 64 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dim_feedforward │ 256 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_num_heads │ 4 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_num_layers │ 8 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_act_func │ relu │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ probe_cadence │ 2 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ probe_model │ linear_probe │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ n_cls_tokens │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_type │ transformer │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_num_layers │ 16 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_embed_dim │ 16 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_num_heads │ 4 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_p_dropout │ 0.002 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_layer_norm_eps │ 1e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_activation │ relu │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_dim_feedforward │ 256 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_allow_overlap │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_min_ctx_share │ 0.1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_max_ctx_share │ 0.3 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_min_trgt_share │ 0.1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_max_trgt_share │ 0.6 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_num_preds │ 4 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_num_encs │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mu_init_std │ 0.03 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ logvar_init_std │ 0.005 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ logvar_init_bias │ -4.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_sigma_x_init │ -2.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_sigma_y_init │ -2.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ downstream_evaluation_epochs │ 50 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ downstream_probe_patience │ 20 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ probe_representation_type │ mean │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ probe_encoder_type │ target_posterior │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pooled_context_representation_size_for_target_posterior │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pooled_context_representation_size_for_auxiliary_encoder │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_end_weight │ 0.0001 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_end_weight │ 1e-06 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_end_weight │ 1e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_annealing_epochs │ 15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_annealing_epochs │ 15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_annealing_epochs │ 15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ recon_x_weight │ 0.1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ recon_y_weight │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ latent_architecture │ deterministic_no_z │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ jepa_objective │ w_jepa_joint │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_full_num_projections │ 2048 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_marginal_num_projections │ 1024 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_p │ 2 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_include_marginals │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_s_x_weight │ 0.001 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_z_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_s_y_weight │ 0.0001 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_end_weight │ 0.004 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_annealing_epochs │ 15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_marginal_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_marginal_end_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_end_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_start_epoch │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_annealing_epochs │ 30 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_end_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_start_epoch │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_annealing_epochs │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_end_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_start_epoch │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_annealing_epochs │ 30 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ safe_exit │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_selection │ unfiltered │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sigma_aggregation │ mean │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ subset_size │ -1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ is_batchlearning │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ iteration │ 0 │ +╘══════════════════════════════════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ + +CUDA Device Information: +CUDA_VISIBLE_DEVICES: 2 +Number of CUDA devices: 1 +Current CUDA device: 0 +CUDA device name: NVIDIA A100-SXM4-80GB + +Using device: cuda:0 +Device name: NVIDIA A100-SXM4-80GB +Encoder Module parameters: ++-------------------+--------------------------------------------------------+ +| Parameter | Value | ++===================+========================================================+ +| hidden_dim | 64 | ++-------------------+--------------------------------------------------------+ +| idx_num_features | [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15] | ++-------------------+--------------------------------------------------------+ +| idx_cat_features | [] | ++-------------------+--------------------------------------------------------+ +| cardinalities | [] | ++-------------------+--------------------------------------------------------+ +| n_input_features | 16 | ++-------------------+--------------------------------------------------------+ +| device | cuda:0 | ++-------------------+--------------------------------------------------------+ +| num_layers | 8 | ++-------------------+--------------------------------------------------------+ +| num_heads | 4 | ++-------------------+--------------------------------------------------------+ +| p_dropout | 0.001 | ++-------------------+--------------------------------------------------------+ +| layer_norm_eps | 1e-05 | ++-------------------+--------------------------------------------------------+ +| gradient_clipping | 1.0 | ++-------------------+--------------------------------------------------------+ +| dim_feedforward | 256 | ++-------------------+--------------------------------------------------------+ +Encoder Module parameters: ++-------------------+--------------------------------------------------------+ +| Parameter | Value | ++===================+========================================================+ +| hidden_dim | 64 | ++-------------------+--------------------------------------------------------+ +| idx_num_features | [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15] | ++-------------------+--------------------------------------------------------+ +| idx_cat_features | [] | ++-------------------+--------------------------------------------------------+ +| cardinalities | [] | ++-------------------+--------------------------------------------------------+ +| n_input_features | 16 | ++-------------------+--------------------------------------------------------+ +| device | cuda:0 | ++-------------------+--------------------------------------------------------+ +| num_layers | 8 | ++-------------------+--------------------------------------------------------+ +| num_heads | 4 | ++-------------------+--------------------------------------------------------+ +| p_dropout | 0.001 | ++-------------------+--------------------------------------------------------+ +| layer_norm_eps | 1e-05 | ++-------------------+--------------------------------------------------------+ +| gradient_clipping | 1.0 | ++-------------------+--------------------------------------------------------+ +| dim_feedforward | 256 | ++-------------------+--------------------------------------------------------+ +DJEPA parameters: ++-------------------------+---------+ +| Parameter | Value | ++=========================+=========+ +| hidden_dim | 64 | ++-------------------------+---------+ +| num_layers | 8 | ++-------------------------+---------+ +| num_heads | 4 | ++-------------------------+---------+ +| p_dropout | 0.001 | ++-------------------------+---------+ +| layer_norm_eps | 1e-05 | ++-------------------------+---------+ +| feature_type_embedding | True | ++-------------------------+---------+ +| feature_index_embedding | True | ++-------------------------+---------+ +| dim_feedforward | 256 | ++-------------------------+---------+ +| use_reg_token | True | ++-------------------------+---------+ +| gradient_clipping | 1.0 | ++-------------------------+---------+ +| activation | gelu | ++-------------------------+---------+ +| num_features | 16 | ++-------------------------+---------+ +| n_cls_tokens | 1 | ++-------------------------+---------+ +MaskCollator params: +----------------- ----- +Allow overlap False +Min context share 0.1 +Max context share 0.3 +Min target share 0.1 +Max target share 0.6 +Min context 1 +Max context 4 +Min target 1 +Max target 9 +Num preds 4 +Num encs 1 +Num features 16 +Cardinalities [] +Base mask seed 0 +----------------- ----- +KL-x annealing: start=0.0 for 89 steps, then anneal to 0.0001 over 1335 steps. +KL-z annealing: start=0.0 for 89 steps, then anneal to 1e-06 over 1335 steps. +KL-y annealing: start=0.0 for 89 steps, then anneal to 1e-05 over 1335 steps. +W-JEPA marginal-scale annealing (s_x): start=0.0, end=1.0, start_epoch=0, anneal_epochs=30. +W-JEPA marginal-scale annealing (z): start=0.0, end=0.0, start_epoch=0, anneal_epochs=0. +W-JEPA marginal-scale annealing (s_y): start=0.0, end=1.0, start_epoch=0, anneal_epochs=30. +Configuration saved to: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/vtjepa_config.yaml +All results will be saved to: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009 +[VTJEPA] Running downstream probe before training begins +Using full dataset for embedding generation: bank_marketing +Generating embedding using: JEPAEmbeddingEncoder on full dataset + 0%| | 0/89 [00:00= 15) +[VTJEPA] Early stopping triggered at epoch 76 + Best downstream validation score: 0.897627 at epoch 45 + No improvement for 15 probe evaluations +Restored best downstream-selected checkpoint from epoch 45. +Training completed. ++ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -W ignore -m src.downstream_pipeline --exp_path /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009 --num_runs 5 --epochs 50 --models mlp,dcnv2,resnet,autoint,fttransformer --no-filter_uncertainty +[Config] Loaded sigma_aggregation=mean from /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/vtjepa_config.yaml +[Config] Loaded training data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets from /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/vtjepa_config.yaml +Split statistics: + val: 4425 samples + test: 9008 samples + train: 31778 samples +Running downstream for dataset 'bank_marketing', model 'mlp'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpfsnlr28c/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpfsnlr28c/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpfsnlr28c/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpfsnlr28c/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ bank_marketing │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─Linear: 2-2 [128, 16] [128, 128] 2,176 -- 278,528 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +===================================================================================================================================================================== +Total params: 35,473 +Trainable params: 35,473 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 4.54 +===================================================================================================================================================================== +Input size (MB): 0.52 +Forward/backward pass size (MB): 0.41 +Params size (MB): 0.14 +Estimated Total Size (MB): 1.08 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 35.3 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 35.5 K +Non-trainable params: 0 +Total params: 35.5 K +Total estimated model params size (MB): 0.142 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 176.94it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.8978531360626221 │ +│ bank_marketing_val_f1-score │ 0.88892662525177 │ +│ bank_marketing_val_loss │ 0.22977767884731293 │ +│ bank_marketing_val_precision │ 0.8853709697723389 │ +│ bank_marketing_val_recall │ 0.8978531360626221 │ +│ bank_marketing_val_roc_auc │ 0.9060643911361694 │ +│ bank_marketing_val_score │ 0.8978531360626221 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 0.0009999999310821295 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 232.95it/s +Validation metrics: [{'bank_marketing_val_loss': 0.22977767884731293, 'lr': 0.0009999999310821295, 'bank_marketing_val_accuracy': 0.8978531360626221, 'bank_marketing_val_precision': 0.8853709697723389, 'bank_marketing_val_recall': 0.8978531360626221, 'bank_marketing_val_f1-score': 0.88892662525177, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9060643911361694, 'bank_marketing_val_score': 0.8978531360626221}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8309 samples (92.24%) + Class 1: 699 samples (7.76%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.8974245190620422 │ +│ bank_marketing_test_f1-score │ 0.8872795104980469 │ +│ bank_marketing_test_loss │ 0.22559453547000885 │ +│ bank_marketing_test_precision │ 0.884318470954895 │ +│ bank_marketing_test_recall │ 0.8974245190620422 │ +│ bank_marketing_test_roc_auc │ 0.9136142730712891 │ +│ bank_marketing_test_score │ 0.8974245190620422 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 0.0009999994654208422 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 344.35it/s +Test metrics: [{'bank_marketing_test_loss': 0.22559453547000885, 'lr': 0.0009999994654208422, 'bank_marketing_test_accuracy': 0.8974245190620422, 'bank_marketing_test_precision': 0.884318470954895, 'bank_marketing_test_recall': 0.8974245190620422, 'bank_marketing_test_f1-score': 0.8872795104980469, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9136142730712891, 'bank_marketing_test_score': 0.8974245190620422}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpfsnlr28c/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpfsnlr28c/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpfsnlr28c/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpfsnlr28c/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ bank_marketing │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─Linear: 2-2 [128, 16] [128, 128] 2,176 -- 278,528 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +===================================================================================================================================================================== +Total params: 35,473 +Trainable params: 35,473 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 4.54 +===================================================================================================================================================================== +Input size (MB): 0.52 +Forward/backward pass size (MB): 0.41 +Params size (MB): 0.14 +Estimated Total Size (MB): 1.08 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 35.3 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 35.5 K +Non-trainable params: 0 +Total params: 35.5 K +Total estimated model params size (MB): 0.142 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 190.61it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.8967231512069702 │ +│ bank_marketing_val_f1-score │ 0.8859037756919861 │ +│ bank_marketing_val_loss │ 0.22912684082984924 │ +│ bank_marketing_val_precision │ 0.8824706077575684 │ +│ bank_marketing_val_recall │ 0.8967231512069702 │ +│ bank_marketing_val_roc_auc │ 0.9075892567634583 │ +│ bank_marketing_val_score │ 0.8967231512069702 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 0.0009999999310821295 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 159.19it/s +Validation metrics: [{'bank_marketing_val_loss': 0.22912684082984924, 'lr': 0.0009999999310821295, 'bank_marketing_val_accuracy': 0.8967231512069702, 'bank_marketing_val_precision': 0.8824706077575684, 'bank_marketing_val_recall': 0.8967231512069702, 'bank_marketing_val_f1-score': 0.8859037756919861, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9075892567634583, 'bank_marketing_val_score': 0.8967231512069702}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8389 samples (93.13%) + Class 1: 619 samples (6.87%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.8996447324752808 │ +│ bank_marketing_test_f1-score │ 0.8871031999588013 │ +│ bank_marketing_test_loss │ 0.22638066112995148 │ +│ bank_marketing_test_precision │ 0.8857904672622681 │ +│ bank_marketing_test_recall │ 0.8996447324752808 │ +│ bank_marketing_test_roc_auc │ 0.9144390821456909 │ +│ bank_marketing_test_score │ 0.8996447324752808 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 0.0009999994654208422 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 197.73it/s +Test metrics: [{'bank_marketing_test_loss': 0.22638066112995148, 'lr': 0.0009999994654208422, 'bank_marketing_test_accuracy': 0.8996447324752808, 'bank_marketing_test_precision': 0.8857904672622681, 'bank_marketing_test_recall': 0.8996447324752808, 'bank_marketing_test_f1-score': 0.8871031999588013, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9144390821456909, 'bank_marketing_test_score': 0.8996447324752808}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpfsnlr28c/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpfsnlr28c/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpfsnlr28c/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpfsnlr28c/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ bank_marketing │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─Linear: 2-2 [128, 16] [128, 128] 2,176 -- 278,528 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +===================================================================================================================================================================== +Total params: 35,473 +Trainable params: 35,473 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 4.54 +===================================================================================================================================================================== +Input size (MB): 0.52 +Forward/backward pass size (MB): 0.41 +Params size (MB): 0.14 +Estimated Total Size (MB): 1.08 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 35.3 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 35.5 K +Non-trainable params: 0 +Total params: 35.5 K +Total estimated model params size (MB): 0.142 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 190.23it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.8974011540412903 │ +│ bank_marketing_val_f1-score │ 0.8868569731712341 │ +│ bank_marketing_val_loss │ 0.2247917205095291 │ +│ bank_marketing_val_precision │ 0.8834943771362305 │ +│ bank_marketing_val_recall │ 0.8974011540412903 │ +│ bank_marketing_val_roc_auc │ 0.909329354763031 │ +│ bank_marketing_val_score │ 0.8974011540412903 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 0.0009999999310821295 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 187.02it/s +Validation metrics: [{'bank_marketing_val_loss': 0.2247917205095291, 'lr': 0.0009999999310821295, 'bank_marketing_val_accuracy': 0.8974011540412903, 'bank_marketing_val_precision': 0.8834943771362305, 'bank_marketing_val_recall': 0.8974011540412903, 'bank_marketing_val_f1-score': 0.8868569731712341, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.909329354763031, 'bank_marketing_val_score': 0.8974011540412903}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8366 samples (92.87%) + Class 1: 642 samples (7.13%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.8984236121177673 │ +│ bank_marketing_test_f1-score │ 0.8865182399749756 │ +│ bank_marketing_test_loss │ 0.223946213722229 │ +│ bank_marketing_test_precision │ 0.8844873905181885 │ +│ bank_marketing_test_recall │ 0.8984236121177673 │ +│ bank_marketing_test_roc_auc │ 0.914363443851471 │ +│ bank_marketing_test_score │ 0.8984236121177673 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 0.0009999994654208422 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 497.02it/s +Test metrics: [{'bank_marketing_test_loss': 0.223946213722229, 'lr': 0.0009999994654208422, 'bank_marketing_test_accuracy': 0.8984236121177673, 'bank_marketing_test_precision': 0.8844873905181885, 'bank_marketing_test_recall': 0.8984236121177673, 'bank_marketing_test_f1-score': 0.8865182399749756, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.914363443851471, 'bank_marketing_test_score': 0.8984236121177673}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpfsnlr28c/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpfsnlr28c/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpfsnlr28c/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpfsnlr28c/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ bank_marketing │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─Linear: 2-2 [128, 16] [128, 128] 2,176 -- 278,528 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +===================================================================================================================================================================== +Total params: 35,473 +Trainable params: 35,473 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 4.54 +===================================================================================================================================================================== +Input size (MB): 0.52 +Forward/backward pass size (MB): 0.41 +Params size (MB): 0.14 +Estimated Total Size (MB): 1.08 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 35.3 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 35.5 K +Non-trainable params: 0 +Total params: 35.5 K +Total estimated model params size (MB): 0.142 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 238.01it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.8985310792922974 │ +│ bank_marketing_val_f1-score │ 0.8931238651275635 │ +│ bank_marketing_val_loss │ 0.22808919847011566 │ +│ bank_marketing_val_precision │ 0.8898804187774658 │ +│ bank_marketing_val_recall │ 0.8985310792922974 │ +│ bank_marketing_val_roc_auc │ 0.907413125038147 │ +│ bank_marketing_val_score │ 0.8985310792922974 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 0.0009999999310821295 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 448.11it/s +Validation metrics: [{'bank_marketing_val_loss': 0.22808919847011566, 'lr': 0.0009999999310821295, 'bank_marketing_val_accuracy': 0.8985310792922974, 'bank_marketing_val_precision': 0.8898804187774658, 'bank_marketing_val_recall': 0.8985310792922974, 'bank_marketing_val_f1-score': 0.8931238651275635, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.907413125038147, 'bank_marketing_val_score': 0.8985310792922974}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8199 samples (91.02%) + Class 1: 809 samples (8.98%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.8992007374763489 │ +│ bank_marketing_test_f1-score │ 0.8924477696418762 │ +│ bank_marketing_test_loss │ 0.2233138233423233 │ +│ bank_marketing_test_precision │ 0.8892562389373779 │ +│ bank_marketing_test_recall │ 0.8992007374763489 │ +│ bank_marketing_test_roc_auc │ 0.9150081872940063 │ +│ bank_marketing_test_score │ 0.8992007374763489 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 0.0009999994654208422 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 454.43it/s +Test metrics: [{'bank_marketing_test_loss': 0.2233138233423233, 'lr': 0.0009999994654208422, 'bank_marketing_test_accuracy': 0.8992007374763489, 'bank_marketing_test_precision': 0.8892562389373779, 'bank_marketing_test_recall': 0.8992007374763489, 'bank_marketing_test_f1-score': 0.8924477696418762, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9150081872940063, 'bank_marketing_test_score': 0.8992007374763489}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpfsnlr28c/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpfsnlr28c/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpfsnlr28c/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpfsnlr28c/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ bank_marketing │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─Linear: 2-2 [128, 16] [128, 128] 2,176 -- 278,528 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +===================================================================================================================================================================== +Total params: 35,473 +Trainable params: 35,473 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 4.54 +===================================================================================================================================================================== +Input size (MB): 0.52 +Forward/backward pass size (MB): 0.41 +Params size (MB): 0.14 +Estimated Total Size (MB): 1.08 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 35.3 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 35.5 K +Non-trainable params: 0 +Total params: 35.5 K +Total estimated model params size (MB): 0.142 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 189.85it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.8983050584793091 │ +│ bank_marketing_val_f1-score │ 0.8877190947532654 │ +│ bank_marketing_val_loss │ 0.22907376289367676 │ +│ bank_marketing_val_precision │ 0.8845475912094116 │ +│ bank_marketing_val_recall │ 0.8983050584793091 │ +│ bank_marketing_val_roc_auc │ 0.9051838517189026 │ +│ bank_marketing_val_score │ 0.8983050584793091 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 0.0009999999310821295 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 189.65it/s +Validation metrics: [{'bank_marketing_val_loss': 0.22907376289367676, 'lr': 0.0009999999310821295, 'bank_marketing_val_accuracy': 0.8983050584793091, 'bank_marketing_val_precision': 0.8845475912094116, 'bank_marketing_val_recall': 0.8983050584793091, 'bank_marketing_val_f1-score': 0.8877190947532654, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9051838517189026, 'bank_marketing_val_score': 0.8983050584793091}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8375 samples (92.97%) + Class 1: 633 samples (7.03%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.8949822187423706 │ +│ bank_marketing_test_f1-score │ 0.8823571801185608 │ +│ bank_marketing_test_loss │ 0.22502531111240387 │ +│ bank_marketing_test_precision │ 0.8797292113304138 │ +│ bank_marketing_test_recall │ 0.8949822187423706 │ +│ bank_marketing_test_roc_auc │ 0.9136731624603271 │ +│ bank_marketing_test_score │ 0.8949822187423706 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 0.0009999994654208422 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 185.16it/s +Test metrics: [{'bank_marketing_test_loss': 0.22502531111240387, 'lr': 0.0009999994654208422, 'bank_marketing_test_accuracy': 0.8949822187423706, 'bank_marketing_test_precision': 0.8797292113304138, 'bank_marketing_test_recall': 0.8949822187423706, 'bank_marketing_test_f1-score': 0.8823571801185608, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9136731624603271, 'bank_marketing_test_score': 0.8949822187423706}] +Wrote results for mlp on bank_marketing to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/downstream_results/mlp_bank_marketing_downstream_results.json +Running downstream for dataset 'bank_marketing', model 'dcnv2'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpe6ljcoqg/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpe6ljcoqg/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpe6ljcoqg/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpe6ljcoqg/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ bank_marketing │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─EncoderDCNv2: 2-2 [128, 16] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 16] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 16] [128, 128] 2,176 -- 278,528 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 134,289 +Trainable params: 134,289 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 17.19 +========================================================================================================================================================================== +Input size (MB): 0.52 +Forward/backward pass size (MB): 1.07 +Params size (MB): 0.54 +Estimated Total Size (MB): 2.13 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 134 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 134 K +Non-trainable params: 0 +Total params: 134 K +Total estimated model params size (MB): 0.537 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 39/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 128.92it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.8996610045433044 │ +│ bank_marketing_val_f1-score │ 0.8952116966247559 │ +│ bank_marketing_val_loss │ 0.24878370761871338 │ +│ bank_marketing_val_precision │ 0.8923066258430481 │ +│ bank_marketing_val_recall │ 0.8996610045433044 │ +│ bank_marketing_val_roc_auc │ 0.9083406925201416 │ +│ bank_marketing_val_score │ 0.8996610045433044 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 9.54915230977349e-05 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 189.73it/s +Validation metrics: [{'bank_marketing_val_loss': 0.24878370761871338, 'lr': 9.54915230977349e-05, 'bank_marketing_val_accuracy': 0.8996610045433044, 'bank_marketing_val_precision': 0.8923066258430481, 'bank_marketing_val_recall': 0.8996610045433044, 'bank_marketing_val_f1-score': 0.8952116966247559, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9083406925201416, 'bank_marketing_val_score': 0.8996610045433044}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8105 samples (89.98%) + Class 1: 903 samples (10.02%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.8958703279495239 │ +│ bank_marketing_test_f1-score │ 0.8914102911949158 │ +│ bank_marketing_test_loss │ 0.24591408669948578 │ +│ bank_marketing_test_precision │ 0.88845294713974 │ +│ bank_marketing_test_recall │ 0.8958703279495239 │ +│ bank_marketing_test_roc_auc │ 0.9115240573883057 │ +│ bank_marketing_test_score │ 0.8958703279495239 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 9.549142851028591e-05 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 192.91it/s +Test metrics: [{'bank_marketing_test_loss': 0.24591408669948578, 'lr': 9.549142851028591e-05, 'bank_marketing_test_accuracy': 0.8958703279495239, 'bank_marketing_test_precision': 0.88845294713974, 'bank_marketing_test_recall': 0.8958703279495239, 'bank_marketing_test_f1-score': 0.8914102911949158, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9115240573883057, 'bank_marketing_test_score': 0.8958703279495239}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpe6ljcoqg/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpe6ljcoqg/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpe6ljcoqg/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpe6ljcoqg/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ bank_marketing │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─EncoderDCNv2: 2-2 [128, 16] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 16] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 16] [128, 128] 2,176 -- 278,528 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 134,289 +Trainable params: 134,289 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 17.19 +========================================================================================================================================================================== +Input size (MB): 0.52 +Forward/backward pass size (MB): 1.07 +Params size (MB): 0.54 +Estimated Total Size (MB): 2.13 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 134 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 134 K +Non-trainable params: 0 +Total params: 134 K +Total estimated model params size (MB): 0.537 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 30/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 146.13it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.8955932259559631 │ +│ bank_marketing_val_f1-score │ 0.8953327536582947 │ +│ bank_marketing_val_loss │ 0.24530549347400665 │ +│ bank_marketing_val_precision │ 0.8950775861740112 │ +│ bank_marketing_val_recall │ 0.8955932259559631 │ +│ bank_marketing_val_roc_auc │ 0.9046961069107056 │ +│ bank_marketing_val_score │ 0.8955932259559631 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 0.000315937795676291 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 173.86it/s +Validation metrics: [{'bank_marketing_val_loss': 0.24530549347400665, 'lr': 0.000315937795676291, 'bank_marketing_val_accuracy': 0.8955932259559631, 'bank_marketing_val_precision': 0.8950775861740112, 'bank_marketing_val_recall': 0.8955932259559631, 'bank_marketing_val_f1-score': 0.8953327536582947, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9046961069107056, 'bank_marketing_val_score': 0.8955932259559631}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 7926 samples (87.99%) + Class 1: 1082 samples (12.01%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.8948712348937988 │ +│ bank_marketing_test_f1-score │ 0.8945179581642151 │ +│ bank_marketing_test_loss │ 0.2277309000492096 │ +│ bank_marketing_test_precision │ 0.8941749334335327 │ +│ bank_marketing_test_recall │ 0.8948712348937988 │ +│ bank_marketing_test_roc_auc │ 0.9153793454170227 │ +│ bank_marketing_test_score │ 0.8948712348937988 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 0.0003159379994031042 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 190.96it/s +Test metrics: [{'bank_marketing_test_loss': 0.2277309000492096, 'lr': 0.0003159379994031042, 'bank_marketing_test_accuracy': 0.8948712348937988, 'bank_marketing_test_precision': 0.8941749334335327, 'bank_marketing_test_recall': 0.8948712348937988, 'bank_marketing_test_f1-score': 0.8945179581642151, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9153793454170227, 'bank_marketing_test_score': 0.8948712348937988}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpe6ljcoqg/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpe6ljcoqg/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpe6ljcoqg/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpe6ljcoqg/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ bank_marketing │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─EncoderDCNv2: 2-2 [128, 16] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 16] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 16] [128, 128] 2,176 -- 278,528 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 134,289 +Trainable params: 134,289 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 17.19 +========================================================================================================================================================================== +Input size (MB): 0.52 +Forward/backward pass size (MB): 1.07 +Params size (MB): 0.54 +Estimated Total Size (MB): 2.13 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 134 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 134 K +Non-trainable params: 0 +Total params: 134 K +Total estimated model params size (MB): 0.537 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 36/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:02 • 0:00:00 115.99it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.8964971899986267 │ +│ bank_marketing_val_f1-score │ 0.8905891180038452 │ +│ bank_marketing_val_loss │ 0.2553221881389618 │ +│ bank_marketing_val_precision │ 0.8871020674705505 │ +│ bank_marketing_val_recall │ 0.8964971899986267 │ +│ bank_marketing_val_roc_auc │ 0.9006679654121399 │ +│ bank_marketing_val_score │ 0.8964971899986267 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 0.00015772644837852567 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 108.19it/s +Validation metrics: [{'bank_marketing_val_loss': 0.2553221881389618, 'lr': 0.00015772644837852567, 'bank_marketing_val_accuracy': 0.8964971899986267, 'bank_marketing_val_precision': 0.8871020674705505, 'bank_marketing_val_recall': 0.8964971899986267, 'bank_marketing_val_f1-score': 0.8905891180038452, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9006679654121399, 'bank_marketing_val_score': 0.8964971899986267}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8148 samples (90.45%) + Class 1: 860 samples (9.55%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.8957593441009521 │ +│ bank_marketing_test_f1-score │ 0.8901763558387756 │ +│ bank_marketing_test_loss │ 0.24071477353572845 │ +│ bank_marketing_test_precision │ 0.8868793845176697 │ +│ bank_marketing_test_recall │ 0.8957593441009521 │ +│ bank_marketing_test_roc_auc │ 0.9093999266624451 │ +│ bank_marketing_test_score │ 0.8957593441009521 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 0.00015772631741128862 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 122.93it/s +Test metrics: [{'bank_marketing_test_loss': 0.24071477353572845, 'lr': 0.00015772631741128862, 'bank_marketing_test_accuracy': 0.8957593441009521, 'bank_marketing_test_precision': 0.8868793845176697, 'bank_marketing_test_recall': 0.8957593441009521, 'bank_marketing_test_f1-score': 0.8901763558387756, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9093999266624451, 'bank_marketing_test_score': 0.8957593441009521}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpe6ljcoqg/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpe6ljcoqg/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpe6ljcoqg/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpe6ljcoqg/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ bank_marketing │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─EncoderDCNv2: 2-2 [128, 16] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 16] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 16] [128, 128] 2,176 -- 278,528 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 134,289 +Trainable params: 134,289 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 17.19 +========================================================================================================================================================================== +Input size (MB): 0.52 +Forward/backward pass size (MB): 1.07 +Params size (MB): 0.54 +Estimated Total Size (MB): 2.13 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 134 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 134 K +Non-trainable params: 0 +Total params: 134 K +Total estimated model params size (MB): 0.537 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 41/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 146.05it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.894463300704956 │ +│ bank_marketing_val_f1-score │ 0.8910015821456909 │ +│ bank_marketing_val_loss │ 0.2640564441680908 │ +│ bank_marketing_val_precision │ 0.8883780241012573 │ +│ bank_marketing_val_recall │ 0.894463300704956 │ +│ bank_marketing_val_roc_auc │ 0.8993784189224243 │ +│ bank_marketing_val_score │ 0.894463300704956 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 6.184667290654033e-05 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 215.87it/s +Validation metrics: [{'bank_marketing_val_loss': 0.2640564441680908, 'lr': 6.184667290654033e-05, 'bank_marketing_val_accuracy': 0.894463300704956, 'bank_marketing_val_precision': 0.8883780241012573, 'bank_marketing_val_recall': 0.894463300704956, 'bank_marketing_val_f1-score': 0.8910015821456909, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.8993784189224243, 'bank_marketing_val_score': 0.894463300704956}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8098 samples (89.90%) + Class 1: 910 samples (10.10%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.8948712348937988 │ +│ bank_marketing_test_f1-score │ 0.8905467987060547 │ +│ bank_marketing_test_loss │ 0.2508843243122101 │ +│ bank_marketing_test_precision │ 0.8876044750213623 │ +│ bank_marketing_test_recall │ 0.8948712348937988 │ +│ bank_marketing_test_roc_auc │ 0.908989429473877 │ +│ bank_marketing_test_score │ 0.8948712348937988 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 6.184665107866749e-05 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 196.26it/s +Test metrics: [{'bank_marketing_test_loss': 0.2508843243122101, 'lr': 6.184665107866749e-05, 'bank_marketing_test_accuracy': 0.8948712348937988, 'bank_marketing_test_precision': 0.8876044750213623, 'bank_marketing_test_recall': 0.8948712348937988, 'bank_marketing_test_f1-score': 0.8905467987060547, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.908989429473877, 'bank_marketing_test_score': 0.8948712348937988}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpe6ljcoqg/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpe6ljcoqg/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpe6ljcoqg/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpe6ljcoqg/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ bank_marketing │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─EncoderDCNv2: 2-2 [128, 16] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 16] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 16] [128, 128] 2,176 -- 278,528 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 134,289 +Trainable params: 134,289 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 17.19 +========================================================================================================================================================================== +Input size (MB): 0.52 +Forward/backward pass size (MB): 1.07 +Params size (MB): 0.54 +Estimated Total Size (MB): 2.13 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 134 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 134 K +Non-trainable params: 0 +Total params: 134 K +Total estimated model params size (MB): 0.537 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 34/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 136.31it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.8978531360626221 │ +│ bank_marketing_val_f1-score │ 0.8914575576782227 │ +│ bank_marketing_val_loss │ 0.2521127462387085 │ +│ bank_marketing_val_precision │ 0.8879690170288086 │ +│ bank_marketing_val_recall │ 0.8978531360626221 │ +│ bank_marketing_val_roc_auc │ 0.9041525721549988 │ +│ bank_marketing_val_score │ 0.8978531360626221 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 0.00020610730280168355 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 173.29it/s +Validation metrics: [{'bank_marketing_val_loss': 0.2521127462387085, 'lr': 0.00020610730280168355, 'bank_marketing_val_accuracy': 0.8978531360626221, 'bank_marketing_val_precision': 0.8879690170288086, 'bank_marketing_val_recall': 0.8978531360626221, 'bank_marketing_val_f1-score': 0.8914575576782227, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9041525721549988, 'bank_marketing_val_score': 0.8978531360626221}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8156 samples (90.54%) + Class 1: 852 samples (9.46%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.8948712348937988 │ +│ bank_marketing_test_f1-score │ 0.8890246748924255 │ +│ bank_marketing_test_loss │ 0.23682673275470734 │ +│ bank_marketing_test_precision │ 0.8856142163276672 │ +│ bank_marketing_test_recall │ 0.8948712348937988 │ +│ bank_marketing_test_roc_auc │ 0.9119967222213745 │ +│ bank_marketing_test_score │ 0.8948712348937988 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 0.00020610728824976832 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 177.22it/s +Test metrics: [{'bank_marketing_test_loss': 0.23682673275470734, 'lr': 0.00020610728824976832, 'bank_marketing_test_accuracy': 0.8948712348937988, 'bank_marketing_test_precision': 0.8856142163276672, 'bank_marketing_test_recall': 0.8948712348937988, 'bank_marketing_test_f1-score': 0.8890246748924255, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9119967222213745, 'bank_marketing_test_score': 0.8948712348937988}] +Wrote results for dcnv2 on bank_marketing to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/downstream_results/dcnv2_bank_marketing_downstream_results.json +Running downstream for dataset 'bank_marketing', model 'resnet'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp567h46t2/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=0 --np_seed=0 --output_file=/tmp/tmp567h46t2/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp567h46t2/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp567h46t2/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─ResNet: 2-2 [128, 16] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 16] [128, 128] 2,176 -- 278,528 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 234,001 +Trainable params: 234,001 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 29.95 +========================================================================================================================================================================== +Input size (MB): 0.52 +Forward/backward pass size (MB): 1.98 +Params size (MB): 0.94 +Estimated Total Size (MB): 3.44 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 233 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 234 K +Non-trainable params: 0 +Total params: 234 K +Total estimated model params size (MB): 0.936 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 44/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 163.85it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.9010169506072998 │ +│ bank_marketing_val_f1-score │ 0.8970322012901306 │ +│ bank_marketing_val_loss │ 0.21889719367027283 │ +│ bank_marketing_val_precision │ 0.8943459391593933 │ +│ bank_marketing_val_recall │ 0.9010169506072998 │ +│ bank_marketing_val_roc_auc │ 0.9175759553909302 │ +│ bank_marketing_val_score │ 0.9010169506072998 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 0.0004999999655410647 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 409.74it/s +Validation metrics: [{'bank_marketing_val_loss': 0.21889719367027283, 'lr': 0.0004999999655410647, 'bank_marketing_val_accuracy': 0.9010169506072998, 'bank_marketing_val_precision': 0.8943459391593933, 'bank_marketing_val_recall': 0.9010169506072998, 'bank_marketing_val_f1-score': 0.8970322012901306, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9175759553909302, 'bank_marketing_val_score': 0.9010169506072998}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8102 samples (89.94%) + Class 1: 906 samples (10.06%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.8986456394195557 │ +│ bank_marketing_test_f1-score │ 0.8943783640861511 │ +│ bank_marketing_test_loss │ 0.21612519025802612 │ +│ bank_marketing_test_precision │ 0.891594648361206 │ +│ bank_marketing_test_recall │ 0.8986456394195557 │ +│ bank_marketing_test_roc_auc │ 0.9227721095085144 │ +│ bank_marketing_test_score │ 0.8986456394195557 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 0.0004999997327104211 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 409.78it/s +Test metrics: [{'bank_marketing_test_loss': 0.21612519025802612, 'lr': 0.0004999997327104211, 'bank_marketing_test_accuracy': 0.8986456394195557, 'bank_marketing_test_precision': 0.891594648361206, 'bank_marketing_test_recall': 0.8986456394195557, 'bank_marketing_test_f1-score': 0.8943783640861511, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9227721095085144, 'bank_marketing_test_score': 0.8986456394195557}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp567h46t2/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=1 --np_seed=1 --output_file=/tmp/tmp567h46t2/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp567h46t2/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp567h46t2/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─ResNet: 2-2 [128, 16] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 16] [128, 128] 2,176 -- 278,528 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 234,001 +Trainable params: 234,001 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 29.95 +========================================================================================================================================================================== +Input size (MB): 0.52 +Forward/backward pass size (MB): 1.98 +Params size (MB): 0.94 +Estimated Total Size (MB): 3.44 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 233 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 234 K +Non-trainable params: 0 +Total params: 234 K +Total estimated model params size (MB): 0.936 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 168.24it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.9010169506072998 │ +│ bank_marketing_val_f1-score │ 0.8955821394920349 │ +│ bank_marketing_val_loss │ 0.21721623837947845 │ +│ bank_marketing_val_precision │ 0.8924801349639893 │ +│ bank_marketing_val_recall │ 0.9010169506072998 │ +│ bank_marketing_val_roc_auc │ 0.9208208322525024 │ +│ bank_marketing_val_score │ 0.9010169506072998 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 0.0 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 230.43it/s +Validation metrics: [{'bank_marketing_val_loss': 0.21721623837947845, 'lr': 0.0, 'bank_marketing_val_accuracy': 0.9010169506072998, 'bank_marketing_val_precision': 0.8924801349639893, 'bank_marketing_val_recall': 0.9010169506072998, 'bank_marketing_val_f1-score': 0.8955821394920349, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9208208322525024, 'bank_marketing_val_score': 0.9010169506072998}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8126 samples (90.21%) + Class 1: 882 samples (9.79%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.8997557759284973 │ +│ bank_marketing_test_f1-score │ 0.8949436545372009 │ +│ bank_marketing_test_loss │ 0.21192345023155212 │ +│ bank_marketing_test_precision │ 0.8920477032661438 │ +│ bank_marketing_test_recall │ 0.8997557759284973 │ +│ bank_marketing_test_roc_auc │ 0.9275190234184265 │ +│ bank_marketing_test_score │ 0.8997557759284973 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 0.0 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 278.31it/s +Test metrics: [{'bank_marketing_test_loss': 0.21192345023155212, 'lr': 0.0, 'bank_marketing_test_accuracy': 0.8997557759284973, 'bank_marketing_test_precision': 0.8920477032661438, 'bank_marketing_test_recall': 0.8997557759284973, 'bank_marketing_test_f1-score': 0.8949436545372009, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9275190234184265, 'bank_marketing_test_score': 0.8997557759284973}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp567h46t2/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=2 --np_seed=2 --output_file=/tmp/tmp567h46t2/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp567h46t2/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp567h46t2/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─ResNet: 2-2 [128, 16] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 16] [128, 128] 2,176 -- 278,528 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 234,001 +Trainable params: 234,001 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 29.95 +========================================================================================================================================================================== +Input size (MB): 0.52 +Forward/backward pass size (MB): 1.98 +Params size (MB): 0.94 +Estimated Total Size (MB): 3.44 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 233 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 234 K +Non-trainable params: 0 +Total params: 234 K +Total estimated model params size (MB): 0.936 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 155.69it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.9007909893989563 │ +│ bank_marketing_val_f1-score │ 0.8929097652435303 │ +│ bank_marketing_val_loss │ 0.2241469770669937 │ +│ bank_marketing_val_precision │ 0.8896549940109253 │ +│ bank_marketing_val_recall │ 0.9007909893989563 │ +│ bank_marketing_val_roc_auc │ 0.9155095219612122 │ +│ bank_marketing_val_score │ 0.9007909893989563 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 0.0 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 257.77it/s +Validation metrics: [{'bank_marketing_val_loss': 0.2241469770669937, 'lr': 0.0, 'bank_marketing_val_accuracy': 0.9007909893989563, 'bank_marketing_val_precision': 0.8896549940109253, 'bank_marketing_val_recall': 0.9007909893989563, 'bank_marketing_val_f1-score': 0.8929097652435303, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9155095219612122, 'bank_marketing_val_score': 0.9007909893989563}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8254 samples (91.63%) + Class 1: 754 samples (8.37%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.8957593441009521 │ +│ bank_marketing_test_f1-score │ 0.8871678113937378 │ +│ bank_marketing_test_loss │ 0.21976067125797272 │ +│ bank_marketing_test_precision │ 0.8836522102355957 │ +│ bank_marketing_test_recall │ 0.8957593441009521 │ +│ bank_marketing_test_roc_auc │ 0.921647310256958 │ +│ bank_marketing_test_score │ 0.8957593441009521 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 0.0 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 250.03it/s +Test metrics: [{'bank_marketing_test_loss': 0.21976067125797272, 'lr': 0.0, 'bank_marketing_test_accuracy': 0.8957593441009521, 'bank_marketing_test_precision': 0.8836522102355957, 'bank_marketing_test_recall': 0.8957593441009521, 'bank_marketing_test_f1-score': 0.8871678113937378, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.921647310256958, 'bank_marketing_test_score': 0.8957593441009521}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp567h46t2/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=3 --np_seed=3 --output_file=/tmp/tmp567h46t2/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp567h46t2/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp567h46t2/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─ResNet: 2-2 [128, 16] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 16] [128, 128] 2,176 -- 278,528 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 234,001 +Trainable params: 234,001 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 29.95 +========================================================================================================================================================================== +Input size (MB): 0.52 +Forward/backward pass size (MB): 1.98 +Params size (MB): 0.94 +Estimated Total Size (MB): 3.44 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 233 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 234 K +Non-trainable params: 0 +Total params: 234 K +Total estimated model params size (MB): 0.936 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 174.58it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.9030508399009705 │ +│ bank_marketing_val_f1-score │ 0.8967077136039734 │ +│ bank_marketing_val_loss │ 0.21638795733451843 │ +│ bank_marketing_val_precision │ 0.8936535716056824 │ +│ bank_marketing_val_recall │ 0.9030508399009705 │ +│ bank_marketing_val_roc_auc │ 0.9203115701675415 │ +│ bank_marketing_val_score │ 0.9030508399009705 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 0.0 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 300.45it/s +Validation metrics: [{'bank_marketing_val_loss': 0.21638795733451843, 'lr': 0.0, 'bank_marketing_val_accuracy': 0.9030508399009705, 'bank_marketing_val_precision': 0.8936535716056824, 'bank_marketing_val_recall': 0.9030508399009705, 'bank_marketing_val_f1-score': 0.8967077136039734, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9203115701675415, 'bank_marketing_val_score': 0.9030508399009705}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8176 samples (90.76%) + Class 1: 832 samples (9.24%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.9017540216445923 │ +│ bank_marketing_test_f1-score │ 0.89577716588974 │ +│ bank_marketing_test_loss │ 0.2147349715232849 │ +│ bank_marketing_test_precision │ 0.8928338885307312 │ +│ bank_marketing_test_recall │ 0.9017540216445923 │ +│ bank_marketing_test_roc_auc │ 0.924019992351532 │ +│ bank_marketing_test_score │ 0.9017540216445923 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 0.0 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 284.34it/s +Test metrics: [{'bank_marketing_test_loss': 0.2147349715232849, 'lr': 0.0, 'bank_marketing_test_accuracy': 0.9017540216445923, 'bank_marketing_test_precision': 0.8928338885307312, 'bank_marketing_test_recall': 0.9017540216445923, 'bank_marketing_test_f1-score': 0.89577716588974, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.924019992351532, 'bank_marketing_test_score': 0.9017540216445923}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp567h46t2/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=4 --np_seed=4 --output_file=/tmp/tmp567h46t2/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp567h46t2/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp567h46t2/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─ResNet: 2-2 [128, 16] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 16] [128, 128] 2,176 -- 278,528 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 234,001 +Trainable params: 234,001 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 29.95 +========================================================================================================================================================================== +Input size (MB): 0.52 +Forward/backward pass size (MB): 1.98 +Params size (MB): 0.94 +Estimated Total Size (MB): 3.44 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 233 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 234 K +Non-trainable params: 0 +Total params: 234 K +Total estimated model params size (MB): 0.936 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 178.78it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.9016948938369751 │ +│ bank_marketing_val_f1-score │ 0.8949263095855713 │ +│ bank_marketing_val_loss │ 0.21550880372524261 │ +│ bank_marketing_val_precision │ 0.8917353749275208 │ +│ bank_marketing_val_recall │ 0.9016948938369751 │ +│ bank_marketing_val_roc_auc │ 0.9207606911659241 │ +│ bank_marketing_val_score │ 0.9016948938369751 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 0.0 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 264.32it/s +Validation metrics: [{'bank_marketing_val_loss': 0.21550880372524261, 'lr': 0.0, 'bank_marketing_val_accuracy': 0.9016948938369751, 'bank_marketing_val_precision': 0.8917353749275208, 'bank_marketing_val_recall': 0.9016948938369751, 'bank_marketing_val_f1-score': 0.8949263095855713, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9207606911659241, 'bank_marketing_val_score': 0.9016948938369751}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8199 samples (91.02%) + Class 1: 809 samples (8.98%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.8985346555709839 │ +│ bank_marketing_test_f1-score │ 0.8917370438575745 │ +│ bank_marketing_test_loss │ 0.21002539992332458 │ +│ bank_marketing_test_precision │ 0.8884825706481934 │ +│ bank_marketing_test_recall │ 0.8985346555709839 │ +│ bank_marketing_test_roc_auc │ 0.9276644587516785 │ +│ bank_marketing_test_score │ 0.8985346555709839 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 0.0 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 256.67it/s +Test metrics: [{'bank_marketing_test_loss': 0.21002539992332458, 'lr': 0.0, 'bank_marketing_test_accuracy': 0.8985346555709839, 'bank_marketing_test_precision': 0.8884825706481934, 'bank_marketing_test_recall': 0.8985346555709839, 'bank_marketing_test_f1-score': 0.8917370438575745, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9276644587516785, 'bank_marketing_test_score': 0.8985346555709839}] +Wrote results for resnet on bank_marketing to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/downstream_results/resnet_bank_marketing_downstream_results.json +Running downstream for dataset 'bank_marketing', model 'autoint'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpb4aej848/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpb4aej848/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpb4aej848/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpb4aej848/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ bank_marketing │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─EncoderAutoIntBase: 2-2 [128, 16] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 16] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 16] [128, 16, 128] 2,048 -- 262,144 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 16, 16] [256, 16, 16] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 16, 128] [128, 16, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 16, 128] [128, 16, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 16, 16] [256, 16, 16] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 16, 128] [128, 16, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 16, 128] [128, 16, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 16, 16] [256, 16, 16] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 16, 128] [128, 16, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 16, 128] [128, 16, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 2048] [128, 128] 262,272 -- 33,570,816 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +==================================================================================================================================================================================== +Total params: 479,377 +Trainable params: 479,377 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 61.36 +==================================================================================================================================================================================== +Input size (MB): 0.52 +Forward/backward pass size (MB): 33.70 +Params size (MB): 1.92 +Estimated Total Size (MB): 36.14 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 479 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 479 K +Non-trainable params: 0 +Total params: 479 K +Total estimated model params size (MB): 1.918 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 47/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:02 • 0:00:00 110.83it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.9023728966712952 │ +│ bank_marketing_val_f1-score │ 0.8980438113212585 │ +│ bank_marketing_val_loss │ 0.21948149800300598 │ +│ bank_marketing_val_precision │ 0.8952900767326355 │ +│ bank_marketing_val_recall │ 0.9023728966712952 │ +│ bank_marketing_val_roc_auc │ 0.9229846000671387 │ +│ bank_marketing_val_score │ 0.9023728966712952 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 3.942650891985977e-06 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 290.28it/s +Validation metrics: [{'bank_marketing_val_loss': 0.21948149800300598, 'lr': 3.942650891985977e-06, 'bank_marketing_val_accuracy': 0.9023728966712952, 'bank_marketing_val_precision': 0.8952900767326355, 'bank_marketing_val_recall': 0.9023728966712952, 'bank_marketing_val_f1-score': 0.8980438113212585, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9229846000671387, 'bank_marketing_val_score': 0.9023728966712952}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8164 samples (90.63%) + Class 1: 844 samples (9.37%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.8984236121177673 │ +│ bank_marketing_test_f1-score │ 0.8925639390945435 │ +│ bank_marketing_test_loss │ 0.22231349349021912 │ +│ bank_marketing_test_precision │ 0.8893837332725525 │ +│ bank_marketing_test_recall │ 0.8984236121177673 │ +│ bank_marketing_test_roc_auc │ 0.9235153198242188 │ +│ bank_marketing_test_score │ 0.8984236121177673 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 3.942649982491275e-06 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 174.36it/s +Test metrics: [{'bank_marketing_test_loss': 0.22231349349021912, 'lr': 3.942649982491275e-06, 'bank_marketing_test_accuracy': 0.8984236121177673, 'bank_marketing_test_precision': 0.8893837332725525, 'bank_marketing_test_recall': 0.8984236121177673, 'bank_marketing_test_f1-score': 0.8925639390945435, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9235153198242188, 'bank_marketing_test_score': 0.8984236121177673}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpb4aej848/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpb4aej848/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpb4aej848/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpb4aej848/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ bank_marketing │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─EncoderAutoIntBase: 2-2 [128, 16] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 16] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 16] [128, 16, 128] 2,048 -- 262,144 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 16, 16] [256, 16, 16] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 16, 128] [128, 16, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 16, 128] [128, 16, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 16, 16] [256, 16, 16] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 16, 128] [128, 16, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 16, 128] [128, 16, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 16, 16] [256, 16, 16] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 16, 128] [128, 16, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 16, 128] [128, 16, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 2048] [128, 128] 262,272 -- 33,570,816 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +==================================================================================================================================================================================== +Total params: 479,377 +Trainable params: 479,377 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 61.36 +==================================================================================================================================================================================== +Input size (MB): 0.52 +Forward/backward pass size (MB): 33.70 +Params size (MB): 1.92 +Estimated Total Size (MB): 36.14 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 479 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 479 K +Non-trainable params: 0 +Total params: 479 K +Total estimated model params size (MB): 1.918 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:02 • 0:00:00 114.60it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.9007909893989563 │ +│ bank_marketing_val_f1-score │ 0.8960306644439697 │ +│ bank_marketing_val_loss │ 0.21828456223011017 │ +│ bank_marketing_val_precision │ 0.8930777907371521 │ +│ bank_marketing_val_recall │ 0.9007909893989563 │ +│ bank_marketing_val_roc_auc │ 0.9228183031082153 │ +│ bank_marketing_val_score │ 0.9007909893989563 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 0.0 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 251.88it/s +Validation metrics: [{'bank_marketing_val_loss': 0.21828456223011017, 'lr': 0.0, 'bank_marketing_val_accuracy': 0.9007909893989563, 'bank_marketing_val_precision': 0.8930777907371521, 'bank_marketing_val_recall': 0.9007909893989563, 'bank_marketing_val_f1-score': 0.8960306644439697, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9228183031082153, 'bank_marketing_val_score': 0.9007909893989563}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8177 samples (90.77%) + Class 1: 831 samples (9.23%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.8996447324752808 │ +│ bank_marketing_test_f1-score │ 0.8935130834579468 │ +│ bank_marketing_test_loss │ 0.21871507167816162 │ +│ bank_marketing_test_precision │ 0.8903918862342834 │ +│ bank_marketing_test_recall │ 0.8996447324752808 │ +│ bank_marketing_test_roc_auc │ 0.9251676797866821 │ +│ bank_marketing_test_score │ 0.8996447324752808 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 0.0 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 251.63it/s +Test metrics: [{'bank_marketing_test_loss': 0.21871507167816162, 'lr': 0.0, 'bank_marketing_test_accuracy': 0.8996447324752808, 'bank_marketing_test_precision': 0.8903918862342834, 'bank_marketing_test_recall': 0.8996447324752808, 'bank_marketing_test_f1-score': 0.8935130834579468, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9251676797866821, 'bank_marketing_test_score': 0.8996447324752808}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpb4aej848/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpb4aej848/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpb4aej848/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpb4aej848/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ bank_marketing │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─EncoderAutoIntBase: 2-2 [128, 16] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 16] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 16] [128, 16, 128] 2,048 -- 262,144 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 16, 16] [256, 16, 16] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 16, 128] [128, 16, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 16, 128] [128, 16, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 16, 16] [256, 16, 16] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 16, 128] [128, 16, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 16, 128] [128, 16, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 16, 16] [256, 16, 16] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 16, 128] [128, 16, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 16, 128] [128, 16, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 2048] [128, 128] 262,272 -- 33,570,816 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +==================================================================================================================================================================================== +Total params: 479,377 +Trainable params: 479,377 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 61.36 +==================================================================================================================================================================================== +Input size (MB): 0.52 +Forward/backward pass size (MB): 33.70 +Params size (MB): 1.92 +Estimated Total Size (MB): 36.14 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 479 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 479 K +Non-trainable params: 0 +Total params: 479 K +Total estimated model params size (MB): 1.918 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 48/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:02 • 0:00:00 116.71it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.8960452079772949 │ +│ bank_marketing_val_f1-score │ 0.8910025954246521 │ +│ bank_marketing_val_loss │ 0.22860261797904968 │ +│ bank_marketing_val_precision │ 0.887739360332489 │ +│ bank_marketing_val_recall │ 0.8960452079772949 │ +│ bank_marketing_val_roc_auc │ 0.916431188583374 │ +│ bank_marketing_val_score │ 0.8960452079772949 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 9.866358823273913e-07 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 247.96it/s +Validation metrics: [{'bank_marketing_val_loss': 0.22860261797904968, 'lr': 9.866358823273913e-07, 'bank_marketing_val_accuracy': 0.8960452079772949, 'bank_marketing_val_precision': 0.887739360332489, 'bank_marketing_val_recall': 0.8960452079772949, 'bank_marketing_val_f1-score': 0.8910025954246521, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.916431188583374, 'bank_marketing_val_score': 0.8960452079772949}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8157 samples (90.55%) + Class 1: 851 samples (9.45%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.9023090600967407 │ +│ bank_marketing_test_f1-score │ 0.8968508839607239 │ +│ bank_marketing_test_loss │ 0.21838077902793884 │ +│ bank_marketing_test_precision │ 0.8939895629882812 │ +│ bank_marketing_test_recall │ 0.9023090600967407 │ +│ bank_marketing_test_roc_auc │ 0.9256624579429626 │ +│ bank_marketing_test_score │ 0.9023090600967407 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 9.86635200206365e-07 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 254.00it/s +Test metrics: [{'bank_marketing_test_loss': 0.21838077902793884, 'lr': 9.86635200206365e-07, 'bank_marketing_test_accuracy': 0.9023090600967407, 'bank_marketing_test_precision': 0.8939895629882812, 'bank_marketing_test_recall': 0.9023090600967407, 'bank_marketing_test_f1-score': 0.8968508839607239, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9256624579429626, 'bank_marketing_test_score': 0.9023090600967407}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpb4aej848/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpb4aej848/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpb4aej848/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpb4aej848/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ bank_marketing │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─EncoderAutoIntBase: 2-2 [128, 16] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 16] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 16] [128, 16, 128] 2,048 -- 262,144 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 16, 16] [256, 16, 16] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 16, 128] [128, 16, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 16, 128] [128, 16, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 16, 16] [256, 16, 16] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 16, 128] [128, 16, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 16, 128] [128, 16, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 16, 16] [256, 16, 16] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 16, 128] [128, 16, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 16, 128] [128, 16, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 2048] [128, 128] 262,272 -- 33,570,816 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +==================================================================================================================================================================================== +Total params: 479,377 +Trainable params: 479,377 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 61.36 +==================================================================================================================================================================================== +Input size (MB): 0.52 +Forward/backward pass size (MB): 33.70 +Params size (MB): 1.92 +Estimated Total Size (MB): 36.14 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 479 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 479 K +Non-trainable params: 0 +Total params: 479 K +Total estimated model params size (MB): 1.918 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 47/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:02 • 0:00:00 116.75it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.9032768607139587 │ +│ bank_marketing_val_f1-score │ 0.8984830975532532 │ +│ bank_marketing_val_loss │ 0.2225060611963272 │ +│ bank_marketing_val_precision │ 0.8956463932991028 │ +│ bank_marketing_val_recall │ 0.9032768607139587 │ +│ bank_marketing_val_roc_auc │ 0.9205772280693054 │ +│ bank_marketing_val_score │ 0.9032768607139587 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 3.942650891985977e-06 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 299.53it/s +Validation metrics: [{'bank_marketing_val_loss': 0.2225060611963272, 'lr': 3.942650891985977e-06, 'bank_marketing_val_accuracy': 0.9032768607139587, 'bank_marketing_val_precision': 0.8956463932991028, 'bank_marketing_val_recall': 0.9032768607139587, 'bank_marketing_val_f1-score': 0.8984830975532532, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9205772280693054, 'bank_marketing_val_score': 0.9032768607139587}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8182 samples (90.83%) + Class 1: 826 samples (9.17%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.9030861258506775 │ +│ bank_marketing_test_f1-score │ 0.8970361948013306 │ +│ bank_marketing_test_loss │ 0.2180434763431549 │ +│ bank_marketing_test_precision │ 0.894201397895813 │ +│ bank_marketing_test_recall │ 0.9030861258506775 │ +│ bank_marketing_test_roc_auc │ 0.9257867336273193 │ +│ bank_marketing_test_score │ 0.9030861258506775 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 3.942649982491275e-06 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 297.03it/s +Test metrics: [{'bank_marketing_test_loss': 0.2180434763431549, 'lr': 3.942649982491275e-06, 'bank_marketing_test_accuracy': 0.9030861258506775, 'bank_marketing_test_precision': 0.894201397895813, 'bank_marketing_test_recall': 0.9030861258506775, 'bank_marketing_test_f1-score': 0.8970361948013306, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9257867336273193, 'bank_marketing_test_score': 0.9030861258506775}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpb4aej848/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpb4aej848/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpb4aej848/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpb4aej848/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ bank_marketing │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─EncoderAutoIntBase: 2-2 [128, 16] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 16] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 16] [128, 16, 128] 2,048 -- 262,144 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 16, 16] [256, 16, 16] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 16, 128] [128, 16, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 16, 128] [128, 16, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 16, 16] [256, 16, 16] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 16, 128] [128, 16, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 16, 128] [128, 16, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 16, 128] [128, 16, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 16, 16] [256, 16, 16] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 16, 128] [128, 16, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 16, 128] [128, 16, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 16, 128] [128, 16, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 2048] [128, 128] 262,272 -- 33,570,816 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +==================================================================================================================================================================================== +Total params: 479,377 +Trainable params: 479,377 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 61.36 +==================================================================================================================================================================================== +Input size (MB): 0.52 +Forward/backward pass size (MB): 33.70 +Params size (MB): 1.92 +Estimated Total Size (MB): 36.14 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 479 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 479 K +Non-trainable params: 0 +Total params: 479 K +Total estimated model params size (MB): 1.918 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 39/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:02 • 0:00:00 119.20it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.9003390073776245 │ +│ bank_marketing_val_f1-score │ 0.8949207663536072 │ +│ bank_marketing_val_loss │ 0.21735332906246185 │ +│ bank_marketing_val_precision │ 0.89178067445755 │ +│ bank_marketing_val_recall │ 0.9003390073776245 │ +│ bank_marketing_val_roc_auc │ 0.9221661686897278 │ +│ bank_marketing_val_score │ 0.9003390073776245 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 9.54915230977349e-05 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 303.63it/s +Validation metrics: [{'bank_marketing_val_loss': 0.21735332906246185, 'lr': 9.54915230977349e-05, 'bank_marketing_val_accuracy': 0.9003390073776245, 'bank_marketing_val_precision': 0.89178067445755, 'bank_marketing_val_recall': 0.9003390073776245, 'bank_marketing_val_f1-score': 0.8949207663536072, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9221661686897278, 'bank_marketing_val_score': 0.9003390073776245}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8143 samples (90.40%) + Class 1: 865 samples (9.60%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.9020870327949524 │ +│ bank_marketing_test_f1-score │ 0.8969678282737732 │ +│ bank_marketing_test_loss │ 0.21332308650016785 │ +│ bank_marketing_test_precision │ 0.8941370248794556 │ +│ bank_marketing_test_recall │ 0.9020870327949524 │ +│ bank_marketing_test_roc_auc │ 0.9272551536560059 │ +│ bank_marketing_test_score │ 0.9020870327949524 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 9.549142851028591e-05 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 293.88it/s +Test metrics: [{'bank_marketing_test_loss': 0.21332308650016785, 'lr': 9.549142851028591e-05, 'bank_marketing_test_accuracy': 0.9020870327949524, 'bank_marketing_test_precision': 0.8941370248794556, 'bank_marketing_test_recall': 0.9020870327949524, 'bank_marketing_test_f1-score': 0.8969678282737732, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9272551536560059, 'bank_marketing_test_score': 0.9020870327949524}] +Wrote results for autoint on bank_marketing to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/downstream_results/autoint_bank_marketing_downstream_results.json +Running downstream for dataset 'bank_marketing', model 'fttransformer'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp4t6fk5zi/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=0 --np_seed=0 --output_file=/tmp/tmp4t6fk5zi/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp4t6fk5zi/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp4t6fk5zi/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ bank_marketing │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─EncoderFTTransformer: 2-2 [128, 16] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 16] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 16] [128, 16, 128] 4,096 -- 524,288 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 17, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 17, 17] [512, 17, 17] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 17, 128] [128, 17, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 17, 512] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 17, 256] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 17, 256] [128, 17, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 17, 17] [512, 17, 17] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 17, 128] [128, 17, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 17, 512] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 17, 256] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 17, 256] [128, 17, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 17] [512, 1, 17] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 17, 128] [128, 17, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 17, 512] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 17, 256] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 17, 256] [128, 17, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================================= +Total params: 533,777 +Trainable params: 533,777 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 68.32 +========================================================================================================================================================================================= +Input size (MB): 0.52 +Forward/backward pass size (MB): 69.62 +Params size (MB): 2.14 +Estimated Total Size (MB): 72.28 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 533 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 533 K +Non-trainable params: 0 +Total params: 533 K +Total estimated model params size (MB): 2.135 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:02 • 0:00:00 90.32it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.897175133228302 │ +│ bank_marketing_val_f1-score │ 0.8902133107185364 │ +│ bank_marketing_val_loss │ 0.22677680850028992 │ +│ bank_marketing_val_precision │ 0.8865931034088135 │ +│ bank_marketing_val_recall │ 0.897175133228302 │ +│ bank_marketing_val_roc_auc │ 0.9102813601493835 │ +│ bank_marketing_val_score │ 0.897175133228302 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 0.0 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 242.26it/s +Validation metrics: [{'bank_marketing_val_loss': 0.22677680850028992, 'lr': 0.0, 'bank_marketing_val_accuracy': 0.897175133228302, 'bank_marketing_val_precision': 0.8865931034088135, 'bank_marketing_val_recall': 0.897175133228302, 'bank_marketing_val_f1-score': 0.8902133107185364, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9102813601493835, 'bank_marketing_val_score': 0.897175133228302}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8204 samples (91.07%) + Class 1: 804 samples (8.93%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.9004218578338623 │ +│ bank_marketing_test_f1-score │ 0.893615186214447 │ +│ bank_marketing_test_loss │ 0.22284826636314392 │ +│ bank_marketing_test_precision │ 0.890540599822998 │ +│ bank_marketing_test_recall │ 0.9004218578338623 │ +│ bank_marketing_test_roc_auc │ 0.9157853722572327 │ +│ bank_marketing_test_score │ 0.9004218578338623 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 0.0 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 251.77it/s +Test metrics: [{'bank_marketing_test_loss': 0.22284826636314392, 'lr': 0.0, 'bank_marketing_test_accuracy': 0.9004218578338623, 'bank_marketing_test_precision': 0.890540599822998, 'bank_marketing_test_recall': 0.9004218578338623, 'bank_marketing_test_f1-score': 0.893615186214447, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9157853722572327, 'bank_marketing_test_score': 0.9004218578338623}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp4t6fk5zi/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=1 --np_seed=1 --output_file=/tmp/tmp4t6fk5zi/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp4t6fk5zi/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp4t6fk5zi/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ bank_marketing │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─EncoderFTTransformer: 2-2 [128, 16] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 16] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 16] [128, 16, 128] 4,096 -- 524,288 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 17, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 17, 17] [512, 17, 17] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 17, 128] [128, 17, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 17, 512] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 17, 256] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 17, 256] [128, 17, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 17, 17] [512, 17, 17] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 17, 128] [128, 17, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 17, 512] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 17, 256] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 17, 256] [128, 17, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 17] [512, 1, 17] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 17, 128] [128, 17, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 17, 512] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 17, 256] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 17, 256] [128, 17, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================================= +Total params: 533,777 +Trainable params: 533,777 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 68.32 +========================================================================================================================================================================================= +Input size (MB): 0.52 +Forward/backward pass size (MB): 69.62 +Params size (MB): 2.14 +Estimated Total Size (MB): 72.28 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 533 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 533 K +Non-trainable params: 0 +Total params: 533 K +Total estimated model params size (MB): 2.135 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 44/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:03 • 0:00:00 81.72it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.9010169506072998 │ +│ bank_marketing_val_f1-score │ 0.8949300050735474 │ +│ bank_marketing_val_loss │ 0.23019227385520935 │ +│ bank_marketing_val_precision │ 0.8917264938354492 │ +│ bank_marketing_val_recall │ 0.9010169506072998 │ +│ bank_marketing_val_roc_auc │ 0.9059411287307739 │ +│ bank_marketing_val_score │ 0.9010169506072998 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 2.447173756081611e-05 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 164.88it/s +Validation metrics: [{'bank_marketing_val_loss': 0.23019227385520935, 'lr': 2.447173756081611e-05, 'bank_marketing_val_accuracy': 0.9010169506072998, 'bank_marketing_val_precision': 0.8917264938354492, 'bank_marketing_val_recall': 0.9010169506072998, 'bank_marketing_val_f1-score': 0.8949300050735474, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9059411287307739, 'bank_marketing_val_score': 0.9010169506072998}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8243 samples (91.51%) + Class 1: 765 samples (8.49%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.9003108143806458 │ +│ bank_marketing_test_f1-score │ 0.892410159111023 │ +│ bank_marketing_test_loss │ 0.22492770850658417 │ +│ bank_marketing_test_precision │ 0.8894136548042297 │ +│ bank_marketing_test_recall │ 0.9003108143806458 │ +│ bank_marketing_test_roc_auc │ 0.913726806640625 │ +│ bank_marketing_test_score │ 0.9003108143806458 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 2.447175393172074e-05 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 153.83it/s +Test metrics: [{'bank_marketing_test_loss': 0.22492770850658417, 'lr': 2.447175393172074e-05, 'bank_marketing_test_accuracy': 0.9003108143806458, 'bank_marketing_test_precision': 0.8894136548042297, 'bank_marketing_test_recall': 0.9003108143806458, 'bank_marketing_test_f1-score': 0.892410159111023, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.913726806640625, 'bank_marketing_test_score': 0.9003108143806458}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp4t6fk5zi/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=2 --np_seed=2 --output_file=/tmp/tmp4t6fk5zi/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp4t6fk5zi/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp4t6fk5zi/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ bank_marketing │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─EncoderFTTransformer: 2-2 [128, 16] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 16] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 16] [128, 16, 128] 4,096 -- 524,288 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 17, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 17, 17] [512, 17, 17] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 17, 128] [128, 17, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 17, 512] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 17, 256] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 17, 256] [128, 17, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 17, 17] [512, 17, 17] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 17, 128] [128, 17, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 17, 512] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 17, 256] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 17, 256] [128, 17, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 17] [512, 1, 17] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 17, 128] [128, 17, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 17, 512] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 17, 256] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 17, 256] [128, 17, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================================= +Total params: 533,777 +Trainable params: 533,777 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 68.32 +========================================================================================================================================================================================= +Input size (MB): 0.52 +Forward/backward pass size (MB): 69.62 +Params size (MB): 2.14 +Estimated Total Size (MB): 72.28 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 533 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 533 K +Non-trainable params: 0 +Total params: 533 K +Total estimated model params size (MB): 2.135 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:02 • 0:00:00 90.34it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.8969491720199585 │ +│ bank_marketing_val_f1-score │ 0.8899129629135132 │ +│ bank_marketing_val_loss │ 0.22250057756900787 │ +│ bank_marketing_val_precision │ 0.8862643837928772 │ +│ bank_marketing_val_recall │ 0.8969491720199585 │ +│ bank_marketing_val_roc_auc │ 0.9142277836799622 │ +│ bank_marketing_val_score │ 0.8969491720199585 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 0.0 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 246.43it/s +Validation metrics: [{'bank_marketing_val_loss': 0.22250057756900787, 'lr': 0.0, 'bank_marketing_val_accuracy': 0.8969491720199585, 'bank_marketing_val_precision': 0.8862643837928772, 'bank_marketing_val_recall': 0.8969491720199585, 'bank_marketing_val_f1-score': 0.8899129629135132, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9142277836799622, 'bank_marketing_val_score': 0.8969491720199585}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8238 samples (91.45%) + Class 1: 770 samples (8.55%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.8990896940231323 │ +│ bank_marketing_test_f1-score │ 0.8912361264228821 │ +│ bank_marketing_test_loss │ 0.2215438336133957 │ +│ bank_marketing_test_precision │ 0.8880785703659058 │ +│ bank_marketing_test_recall │ 0.8990896940231323 │ +│ bank_marketing_test_roc_auc │ 0.9173547625541687 │ +│ bank_marketing_test_score │ 0.8990896940231323 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 0.0 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 248.76it/s +Test metrics: [{'bank_marketing_test_loss': 0.2215438336133957, 'lr': 0.0, 'bank_marketing_test_accuracy': 0.8990896940231323, 'bank_marketing_test_precision': 0.8880785703659058, 'bank_marketing_test_recall': 0.8990896940231323, 'bank_marketing_test_f1-score': 0.8912361264228821, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9173547625541687, 'bank_marketing_test_score': 0.8990896940231323}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp4t6fk5zi/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=3 --np_seed=3 --output_file=/tmp/tmp4t6fk5zi/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp4t6fk5zi/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp4t6fk5zi/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ bank_marketing │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─EncoderFTTransformer: 2-2 [128, 16] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 16] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 16] [128, 16, 128] 4,096 -- 524,288 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 17, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 17, 17] [512, 17, 17] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 17, 128] [128, 17, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 17, 512] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 17, 256] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 17, 256] [128, 17, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 17, 17] [512, 17, 17] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 17, 128] [128, 17, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 17, 512] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 17, 256] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 17, 256] [128, 17, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 17] [512, 1, 17] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 17, 128] [128, 17, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 17, 512] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 17, 256] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 17, 256] [128, 17, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================================= +Total params: 533,777 +Trainable params: 533,777 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 68.32 +========================================================================================================================================================================================= +Input size (MB): 0.52 +Forward/backward pass size (MB): 69.62 +Params size (MB): 2.14 +Estimated Total Size (MB): 72.28 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 533 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 533 K +Non-trainable params: 0 +Total params: 533 K +Total estimated model params size (MB): 2.135 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:02 • 0:00:00 83.30it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.8994350433349609 │ +│ bank_marketing_val_f1-score │ 0.8926262259483337 │ +│ bank_marketing_val_loss │ 0.2246096283197403 │ +│ bank_marketing_val_precision │ 0.8892220258712769 │ +│ bank_marketing_val_recall │ 0.8994350433349609 │ +│ bank_marketing_val_roc_auc │ 0.9116213321685791 │ +│ bank_marketing_val_score │ 0.8994350433349609 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 0.0 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 237.07it/s +Validation metrics: [{'bank_marketing_val_loss': 0.2246096283197403, 'lr': 0.0, 'bank_marketing_val_accuracy': 0.8994350433349609, 'bank_marketing_val_precision': 0.8892220258712769, 'bank_marketing_val_recall': 0.8994350433349609, 'bank_marketing_val_f1-score': 0.8926262259483337, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9116213321685791, 'bank_marketing_val_score': 0.8994350433349609}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8204 samples (91.07%) + Class 1: 804 samples (8.93%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.8993117213249207 │ +│ bank_marketing_test_f1-score │ 0.8924291729927063 │ +│ bank_marketing_test_loss │ 0.22113539278507233 │ +│ bank_marketing_test_precision │ 0.8892467617988586 │ +│ bank_marketing_test_recall │ 0.8993117213249207 │ +│ bank_marketing_test_roc_auc │ 0.9176443219184875 │ +│ bank_marketing_test_score │ 0.8993117213249207 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 0.0 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 239.56it/s +Test metrics: [{'bank_marketing_test_loss': 0.22113539278507233, 'lr': 0.0, 'bank_marketing_test_accuracy': 0.8993117213249207, 'bank_marketing_test_precision': 0.8892467617988586, 'bank_marketing_test_recall': 0.8993117213249207, 'bank_marketing_test_f1-score': 0.8924291729927063, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9176443219184875, 'bank_marketing_test_score': 0.8993117213249207}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp4t6fk5zi/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=4 --np_seed=4 --output_file=/tmp/tmp4t6fk5zi/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp4t6fk5zi/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp4t6fk5zi/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 16 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 16, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 16 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ bank_marketing │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 16, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 16, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 16, 64] [128, 16] -- -- -- +│ │ └─Linear: 3-1 [128, 1024] [128, 16] 16,400 -- 2,099,200 +│ └─EncoderFTTransformer: 2-2 [128, 16] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 16] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 16] [128, 16, 128] 4,096 -- 524,288 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 17, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 17, 17] [512, 17, 17] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 17, 128] [128, 17, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 17, 512] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 17, 256] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 17, 256] [128, 17, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 17, 17] [512, 17, 17] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 17, 128] [128, 17, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 17, 512] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 17, 256] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 17, 256] [128, 17, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 17, 128] [128, 17, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 17] [512, 1, 17] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 17, 128] [128, 17, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 17, 128] [128, 17, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 17, 512] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 17, 256] [128, 17, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 17, 256] [128, 17, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 17, 128] [128, 17, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================================= +Total params: 533,777 +Trainable params: 533,777 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 68.32 +========================================================================================================================================================================================= +Input size (MB): 0.52 +Forward/backward pass size (MB): 69.62 +Params size (MB): 2.14 +Estimated Total Size (MB): 72.28 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 533 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 533 K +Non-trainable params: 0 +Total params: 533 K +Total estimated model params size (MB): 2.135 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:03 • 0:00:00 81.01it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.897175133228302 │ +│ bank_marketing_val_f1-score │ 0.8912495374679565 │ +│ bank_marketing_val_loss │ 0.22259756922721863 │ +│ bank_marketing_val_precision │ 0.8878017663955688 │ +│ bank_marketing_val_recall │ 0.897175133228302 │ +│ bank_marketing_val_roc_auc │ 0.9147111177444458 │ +│ bank_marketing_val_score │ 0.897175133228302 │ +│ bank_marketing_val_support │ 4425.0 │ +│ lr │ 0.0 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 241.36it/s +Validation metrics: [{'bank_marketing_val_loss': 0.22259756922721863, 'lr': 0.0, 'bank_marketing_val_accuracy': 0.897175133228302, 'bank_marketing_val_precision': 0.8878017663955688, 'bank_marketing_val_recall': 0.897175133228302, 'bank_marketing_val_f1-score': 0.8912495374679565, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9147111177444458, 'bank_marketing_val_score': 0.897175133228302}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8213 samples (91.17%) + Class 1: 795 samples (8.83%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.8998667597770691 │ +│ bank_marketing_test_f1-score │ 0.8927748203277588 │ +│ bank_marketing_test_loss │ 0.22052712738513947 │ +│ bank_marketing_test_precision │ 0.8896520137786865 │ +│ bank_marketing_test_recall │ 0.8998667597770691 │ +│ bank_marketing_test_roc_auc │ 0.9190995693206787 │ +│ bank_marketing_test_score │ 0.8998667597770691 │ +│ bank_marketing_test_support │ 9008.0 │ +│ lr │ 0.0 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 237.57it/s +Test metrics: [{'bank_marketing_test_loss': 0.22052712738513947, 'lr': 0.0, 'bank_marketing_test_accuracy': 0.8998667597770691, 'bank_marketing_test_precision': 0.8896520137786865, 'bank_marketing_test_recall': 0.8998667597770691, 'bank_marketing_test_f1-score': 0.8927748203277588, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9190995693206787, 'bank_marketing_test_score': 0.8998667597770691}] +Wrote results for fttransformer on bank_marketing to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/downstream_results/fttransformer_bank_marketing_downstream_results.json + +Downstream pipeline complete. Results under /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260915_204009/downstream_results diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu2-covertype.exit_code b/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu2-covertype.exit_code new file mode 100644 index 0000000000000000000000000000000000000000..573541ac9702dd3969c9bc859d2b91ec1f7e6e56 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu2-covertype.exit_code @@ -0,0 +1 @@ +0 diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu2-covertype.log b/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu2-covertype.log new file mode 100644 index 0000000000000000000000000000000000000000..f4d3d35ad1e2a1e0996bd5bcd1803d85e080cce7 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu2-covertype.log @@ -0,0 +1,13760 @@ +[gpu2-covertype] Starting covertype on physical GPU 2 at Tue Sep 15 18:28:13 UTC 2026 ++ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py --dataset covertype --results-root /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42 --epochs 500 --objective w_jepa_joint --latent-architecture deterministic_no_z --sw-full-end-weight 0.001875 --torch-seed 42 --np-seed 42 --mask-seed 0 --sw-seed 0 --sw-full-num-projections 2048 --sw-marginal-num-projections 1024 --sw-p 2 --sw-s-x-end-weight 0.00075 --sw-z-end-weight 0.0 --sw-s-y-end-weight 7.5e-05 --model-selection unfiltered --deterministic-mode warn --sw-include-marginals --batch-size 512 +{ + "dataset": "covertype", + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "covertype", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "64", + "--model_num_layers", + "8", + "--model_num_heads", + "8", + "--model_dropout_prob", + "0.0015", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.15", + "--mask_max_ctx_share", + "0.35", + "--mask_min_trgt_share", + "0.15", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.0005", + "--exp_warmup", + "12", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "1e-06", + "--vtjepa_kl_x_annealing_epochs", + "60", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-05", + "--vtjepa_kl_z_annealing_epochs", + "60", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-06", + "--vtjepa_kl_y_annealing_epochs", + "20", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.001", + "--recon_y_weight", + "0.5", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.001875", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "60", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.00075", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "7.5e-05", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} +Not SLURM job. +╒══════════════════════════════════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ mock │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ activation_checkpointing │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 42 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 42 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mask_seed │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ sw_seed │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ deterministic_mode │ warn │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pre_evaluation │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 512 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 500 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ improvement │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_val_cache_cadence │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ init_type │ trunc_normal │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_start_lr │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_lr │ 0.0005 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_final_lr │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 12 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_final_weight_decay │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_ipe_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_gradient_clipping │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_feature_type_embedding │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_feature_index_embedding │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dropout_prob │ 0.0015 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_layer_norm_eps │ 1e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dim_hidden │ 64 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dim_feedforward │ 64 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_num_heads │ 8 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_num_layers │ 8 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_act_func │ relu │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ probe_cadence │ 2 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ probe_model │ linear_probe │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ n_cls_tokens │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pred_type │ transformer │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pred_num_layers │ 16 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pred_embed_dim │ 16 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pred_num_heads │ 4 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pred_p_dropout │ 0.002 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pred_layer_norm_eps │ 1e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pred_activation │ relu │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pred_dim_feedforward │ 256 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mask_allow_overlap │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mask_min_ctx_share │ 0.15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mask_max_ctx_share │ 0.35 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mask_min_trgt_share │ 0.15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mask_max_trgt_share │ 0.6 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mask_num_preds │ 4 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mask_num_encs │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mu_init_std │ 0.03 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ logvar_init_std │ 0.005 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ logvar_init_bias │ -4.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_sigma_x_init │ -2.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_sigma_y_init │ -2.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ downstream_evaluation_epochs │ 50 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ downstream_probe_patience │ 20 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ probe_representation_type │ mean │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ probe_encoder_type │ target_posterior │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pooled_context_representation_size_for_target_posterior │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pooled_context_representation_size_for_auxiliary_encoder │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_end_weight │ 1e-06 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_end_weight │ 1e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_end_weight │ 1e-06 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_annealing_epochs │ 60 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_annealing_epochs │ 60 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_annealing_epochs │ 20 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ recon_x_weight │ 0.001 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ recon_y_weight │ 0.5 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ latent_architecture │ deterministic_no_z │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ jepa_objective │ w_jepa_joint │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ sw_full_num_projections │ 2048 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ sw_marginal_num_projections │ 1024 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ sw_p │ 2 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ sw_include_marginals │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ sw_s_x_weight │ 0.00075 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ sw_z_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ sw_s_y_weight │ 7.5e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_end_weight │ 0.001875 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_annealing_epochs │ 60 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_marginal_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_marginal_end_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_end_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_start_epoch │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_annealing_epochs │ 30 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_end_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_start_epoch │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_annealing_epochs │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_end_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_start_epoch │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_annealing_epochs │ 30 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ safe_exit │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_selection │ unfiltered │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ sigma_aggregation │ mean │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ subset_size │ -1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ is_batchlearning │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ iteration │ 0 │ +╘══════════════════════════════════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ + +CUDA Device Information: +CUDA_VISIBLE_DEVICES: 2 +Number of CUDA devices: 1 +Current CUDA device: 0 +CUDA device name: NVIDIA A100-SXM4-80GB + +Using device: cuda:0 +Device name: NVIDIA A100-SXM4-80GB +Encoder Module parameters: ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| Parameter | Value | ++===================+================================================================================================================================================================================================================+ +| hidden_dim | 64 | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| idx_num_features | [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53] | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| idx_cat_features | [] | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| cardinalities | [] | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| n_input_features | 54 | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| device | cuda:0 | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| num_layers | 8 | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| num_heads | 8 | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| p_dropout | 0.0015 | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| layer_norm_eps | 1e-05 | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| gradient_clipping | 1.0 | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| dim_feedforward | 64 | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +Encoder Module parameters: ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| Parameter | Value | ++===================+================================================================================================================================================================================================================+ +| hidden_dim | 64 | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| idx_num_features | [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53] | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| idx_cat_features | [] | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| cardinalities | [] | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| n_input_features | 54 | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| device | cuda:0 | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| num_layers | 8 | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| num_heads | 8 | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| p_dropout | 0.0015 | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| layer_norm_eps | 1e-05 | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| gradient_clipping | 1.0 | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| dim_feedforward | 64 | ++-------------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +DJEPA parameters: ++-------------------------+---------+ +| Parameter | Value | ++=========================+=========+ +| hidden_dim | 64 | ++-------------------------+---------+ +| num_layers | 8 | ++-------------------------+---------+ +| num_heads | 8 | ++-------------------------+---------+ +| p_dropout | 0.0015 | ++-------------------------+---------+ +| layer_norm_eps | 1e-05 | ++-------------------------+---------+ +| feature_type_embedding | True | ++-------------------------+---------+ +| feature_index_embedding | True | ++-------------------------+---------+ +| dim_feedforward | 64 | ++-------------------------+---------+ +| use_reg_token | True | ++-------------------------+---------+ +| gradient_clipping | 1.0 | ++-------------------------+---------+ +| activation | gelu | ++-------------------------+---------+ +| num_features | 54 | ++-------------------------+---------+ +| n_cls_tokens | 1 | ++-------------------------+---------+ +MaskCollator params: +----------------- ----- +Allow overlap False +Min context share 0.15 +Max context share 0.35 +Min target share 0.15 +Max target share 0.6 +Min context 8 +Max context 18 +Min target 8 +Max target 32 +Num preds 4 +Num encs 1 +Num features 54 +Cardinalities [] +Base mask seed 0 +----------------- ----- +KL-x annealing: start=0.0 for 20 steps, then anneal to 1e-06 over 1200 steps. +KL-z annealing: start=0.0 for 20 steps, then anneal to 1e-05 over 1200 steps. +KL-y annealing: start=0.0 for 20 steps, then anneal to 1e-06 over 400 steps. +W-JEPA marginal-scale annealing (s_x): start=0.0, end=1.0, start_epoch=0, anneal_epochs=30. +W-JEPA marginal-scale annealing (z): start=0.0, end=0.0, start_epoch=0, anneal_epochs=0. +W-JEPA marginal-scale annealing (s_y): start=0.0, end=1.0, start_epoch=0, anneal_epochs=30. +Configuration saved to: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/vtjepa_config.yaml +All results will be saved to: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951 +[VTJEPA] Running downstream probe before training begins +Using full dataset for embedding generation: covertype +Generating embedding using: JEPAEmbeddingEncoder on full dataset + 0%| | 0/20 [00:00= 15) +[VTJEPA] Early stopping triggered at epoch 46 + Best downstream validation score: 0.729730 at epoch 15 + No improvement for 15 probe evaluations +Restored best downstream-selected checkpoint from epoch 15. +Training completed. ++ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -W ignore -m src.downstream_pipeline --exp_path /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951 --num_runs 5 --epochs 50 --models mlp,dcnv2,resnet,autoint,fttransformer --no-filter_uncertainty +[Config] Loaded sigma_aggregation=mean from /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/vtjepa_config.yaml +[Config] Loaded training data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 from /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/vtjepa_config.yaml +Split statistics: + val: 999 samples + test: 2000 samples + train: 7001 samples +Running downstream for dataset 'covertype', model 'mlp'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp34bsiuiq/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=0 --np_seed=0 --output_file=/tmp/tmp34bsiuiq/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp34bsiuiq/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp34bsiuiq/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ covertype │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─Linear: 2-2 [128, 54] [128, 128] 7,040 -- 901,120 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +===================================================================================================================================================================== +Total params: 211,389 +Trainable params: 211,389 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 27.06 +===================================================================================================================================================================== +Input size (MB): 1.77 +Forward/backward pass size (MB): 0.46 +Params size (MB): 0.85 +Estimated Total Size (MB): 3.07 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 210 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 211 K +Non-trainable params: 0 +Total params: 211 K +Total estimated model params size (MB): 0.846 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 40/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 253.24it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7277277112007141 │ +│ covertype_val_f1-score │ 0.7182832956314087 │ +│ covertype_val_loss │ 0.6716936826705933 │ +│ covertype_val_precision │ 0.7335260510444641 │ +│ covertype_val_recall │ 0.7277277112007141 │ +│ covertype_val_score │ 0.7277277112007141 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0010000000474974513 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 489.72it/s +Validation metrics: [{'covertype_val_loss': 0.6716936826705933, 'lr': 0.0010000000474974513, 'covertype_val_accuracy': 0.7277277112007141, 'covertype_val_precision': 0.7335260510444641, 'covertype_val_recall': 0.7277277112007141, 'covertype_val_f1-score': 0.7182832956314087, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7277277112007141}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 1: 974 samples (48.70%) + Class 2: 750 samples (37.50%) + Class 3: 65 samples (3.25%) + Class 4: 187 samples (9.35%) + Class 5: 15 samples (0.75%) + Class 6: 9 samples (0.45%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.7114999890327454 │ +│ covertype_test_f1-score │ 0.7001162767410278 │ +│ covertype_test_loss │ 0.7074241042137146 │ +│ covertype_test_precision │ 0.6978535652160645 │ +│ covertype_test_recall │ 0.7114999890327454 │ +│ covertype_test_score │ 0.7114999890327454 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0010000001639127731 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 501.36it/s +Test metrics: [{'covertype_test_loss': 0.7074241042137146, 'lr': 0.0010000001639127731, 'covertype_test_accuracy': 0.7114999890327454, 'covertype_test_precision': 0.6978535652160645, 'covertype_test_recall': 0.7114999890327454, 'covertype_test_f1-score': 0.7001162767410278, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.7114999890327454}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp34bsiuiq/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=1 --np_seed=1 --output_file=/tmp/tmp34bsiuiq/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp34bsiuiq/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp34bsiuiq/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ covertype │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─Linear: 2-2 [128, 54] [128, 128] 7,040 -- 901,120 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +===================================================================================================================================================================== +Total params: 211,389 +Trainable params: 211,389 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 27.06 +===================================================================================================================================================================== +Input size (MB): 1.77 +Forward/backward pass size (MB): 0.46 +Params size (MB): 0.85 +Estimated Total Size (MB): 3.07 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 210 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 211 K +Non-trainable params: 0 +Total params: 211 K +Total estimated model params size (MB): 0.846 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 272.55it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7197197079658508 │ +│ covertype_val_f1-score │ 0.7083337306976318 │ +│ covertype_val_loss │ 0.747097373008728 │ +│ covertype_val_precision │ 0.7187588810920715 │ +│ covertype_val_recall │ 0.7197197079658508 │ +│ covertype_val_score │ 0.7197197079658508 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0010000000474974513 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 512.90it/s +Validation metrics: [{'covertype_val_loss': 0.747097373008728, 'lr': 0.0010000000474974513, 'covertype_val_accuracy': 0.7197197079658508, 'covertype_val_precision': 0.7187588810920715, 'covertype_val_recall': 0.7197197079658508, 'covertype_val_f1-score': 0.7083337306976318, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7197197079658508}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 1: 1158 samples (57.90%) + Class 2: 520 samples (26.00%) + Class 3: 56 samples (2.80%) + Class 4: 223 samples (11.15%) + Class 5: 43 samples (2.15%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.6945000290870667 │ +│ covertype_test_f1-score │ 0.6832040548324585 │ +│ covertype_test_loss │ 0.731846809387207 │ +│ covertype_test_precision │ 0.6960458159446716 │ +│ covertype_test_recall │ 0.6945000290870667 │ +│ covertype_test_score │ 0.6945000290870667 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0010000001639127731 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 511.13it/s +Test metrics: [{'covertype_test_loss': 0.731846809387207, 'lr': 0.0010000001639127731, 'covertype_test_accuracy': 0.6945000290870667, 'covertype_test_precision': 0.6960458159446716, 'covertype_test_recall': 0.6945000290870667, 'covertype_test_f1-score': 0.6832040548324585, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.6945000290870667}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp34bsiuiq/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=2 --np_seed=2 --output_file=/tmp/tmp34bsiuiq/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp34bsiuiq/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp34bsiuiq/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ covertype │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─Linear: 2-2 [128, 54] [128, 128] 7,040 -- 901,120 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +===================================================================================================================================================================== +Total params: 211,389 +Trainable params: 211,389 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 27.06 +===================================================================================================================================================================== +Input size (MB): 1.77 +Forward/backward pass size (MB): 0.46 +Params size (MB): 0.85 +Estimated Total Size (MB): 3.07 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 210 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 211 K +Non-trainable params: 0 +Total params: 211 K +Total estimated model params size (MB): 0.846 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 32/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 217.21it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7137137055397034 │ +│ covertype_val_f1-score │ 0.6883257031440735 │ +│ covertype_val_loss │ 0.6517016291618347 │ +│ covertype_val_precision │ 0.7004646062850952 │ +│ covertype_val_recall │ 0.7137137055397034 │ +│ covertype_val_score │ 0.7137137055397034 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0010000000474974513 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 418.37it/s +Validation metrics: [{'covertype_val_loss': 0.6517016291618347, 'lr': 0.0010000000474974513, 'covertype_val_accuracy': 0.7137137055397034, 'covertype_val_precision': 0.7004646062850952, 'covertype_val_recall': 0.7137137055397034, 'covertype_val_f1-score': 0.6883257031440735, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7137137055397034}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 1: 1269 samples (63.45%) + Class 2: 505 samples (25.25%) + Class 3: 34 samples (1.70%) + Class 4: 189 samples (9.45%) + Class 5: 3 samples (0.15%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.7074999809265137 │ +│ covertype_test_f1-score │ 0.6821770071983337 │ +│ covertype_test_loss │ 0.6506989598274231 │ +│ covertype_test_precision │ 0.6984860301017761 │ +│ covertype_test_recall │ 0.7074999809265137 │ +│ covertype_test_score │ 0.7074999809265137 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0010000001639127731 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 294.45it/s +Test metrics: [{'covertype_test_loss': 0.6506989598274231, 'lr': 0.0010000001639127731, 'covertype_test_accuracy': 0.7074999809265137, 'covertype_test_precision': 0.6984860301017761, 'covertype_test_recall': 0.7074999809265137, 'covertype_test_f1-score': 0.6821770071983337, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.7074999809265137}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp34bsiuiq/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=3 --np_seed=3 --output_file=/tmp/tmp34bsiuiq/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp34bsiuiq/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp34bsiuiq/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ covertype │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─Linear: 2-2 [128, 54] [128, 128] 7,040 -- 901,120 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +===================================================================================================================================================================== +Total params: 211,389 +Trainable params: 211,389 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 27.06 +===================================================================================================================================================================== +Input size (MB): 1.77 +Forward/backward pass size (MB): 0.46 +Params size (MB): 0.85 +Estimated Total Size (MB): 3.07 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 210 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 211 K +Non-trainable params: 0 +Total params: 211 K +Total estimated model params size (MB): 0.846 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 25/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 208.41it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.5125125050544739 │ +│ covertype_val_f1-score │ 0.4801228940486908 │ +│ covertype_val_loss │ 1.1511790752410889 │ +│ covertype_val_precision │ 0.5424755215644836 │ +│ covertype_val_recall │ 0.5125125050544739 │ +│ covertype_val_score │ 0.5125125050544739 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0010000000474974513 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 334.67it/s +Validation metrics: [{'covertype_val_loss': 1.1511790752410889, 'lr': 0.0010000000474974513, 'covertype_val_accuracy': 0.5125125050544739, 'covertype_val_precision': 0.5424755215644836, 'covertype_val_recall': 0.5125125050544739, 'covertype_val_f1-score': 0.4801228940486908, 'covertype_val_support': 999.0, 'covertype_val_score': 0.5125125050544739}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 1: 551 samples (27.55%) + Class 2: 1257 samples (62.85%) + Class 3: 150 samples (7.50%) + Class 4: 34 samples (1.70%) + Class 6: 8 samples (0.40%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.5264999866485596 │ +│ covertype_test_f1-score │ 0.4904860258102417 │ +│ covertype_test_loss │ 1.154642939567566 │ +│ covertype_test_precision │ 0.5562233924865723 │ +│ covertype_test_recall │ 0.5264999866485596 │ +│ covertype_test_score │ 0.5264999866485596 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0010000001639127731 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 542.30it/s +Test metrics: [{'covertype_test_loss': 1.154642939567566, 'lr': 0.0010000001639127731, 'covertype_test_accuracy': 0.5264999866485596, 'covertype_test_precision': 0.5562233924865723, 'covertype_test_recall': 0.5264999866485596, 'covertype_test_f1-score': 0.4904860258102417, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.5264999866485596}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp34bsiuiq/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=4 --np_seed=4 --output_file=/tmp/tmp34bsiuiq/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp34bsiuiq/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp34bsiuiq/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ covertype │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─Linear: 2-2 [128, 54] [128, 128] 7,040 -- 901,120 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +===================================================================================================================================================================== +Total params: 211,389 +Trainable params: 211,389 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 27.06 +===================================================================================================================================================================== +Input size (MB): 1.77 +Forward/backward pass size (MB): 0.46 +Params size (MB): 0.85 +Estimated Total Size (MB): 3.07 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 210 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 211 K +Non-trainable params: 0 +Total params: 211 K +Total estimated model params size (MB): 0.846 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 242.41it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.661661684513092 │ +│ covertype_val_f1-score │ 0.648414134979248 │ +│ covertype_val_loss │ 0.7696166634559631 │ +│ covertype_val_precision │ 0.6684898138046265 │ +│ covertype_val_recall │ 0.661661684513092 │ +│ covertype_val_score │ 0.661661684513092 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0010000000474974513 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 491.08it/s +Validation metrics: [{'covertype_val_loss': 0.7696166634559631, 'lr': 0.0010000000474974513, 'covertype_val_accuracy': 0.661661684513092, 'covertype_val_precision': 0.6684898138046265, 'covertype_val_recall': 0.661661684513092, 'covertype_val_f1-score': 0.648414134979248, 'covertype_val_support': 999.0, 'covertype_val_score': 0.661661684513092}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 1: 757 samples (37.85%) + Class 2: 1061 samples (53.05%) + Class 3: 55 samples (2.75%) + Class 4: 87 samples (4.35%) + Class 5: 34 samples (1.70%) + Class 6: 6 samples (0.30%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.6744999885559082 │ +│ covertype_test_f1-score │ 0.6627953052520752 │ +│ covertype_test_loss │ 0.7460566759109497 │ +│ covertype_test_precision │ 0.6900953650474548 │ +│ covertype_test_recall │ 0.6744999885559082 │ +│ covertype_test_score │ 0.6744999885559082 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0010000001639127731 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 533.65it/s +Test metrics: [{'covertype_test_loss': 0.7460566759109497, 'lr': 0.0010000001639127731, 'covertype_test_accuracy': 0.6744999885559082, 'covertype_test_precision': 0.6900953650474548, 'covertype_test_recall': 0.6744999885559082, 'covertype_test_f1-score': 0.6627953052520752, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.6744999885559082}] +Wrote results for mlp on covertype to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/downstream_results/mlp_covertype_downstream_results.json +Running downstream for dataset 'covertype', model 'dcnv2'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpaux3weei/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpaux3weei/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpaux3weei/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpaux3weei/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2750 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ covertype │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─EncoderDCNv2: 2-2 [128, 54] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 54] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 54] [128, 128] 7,040 -- 901,120 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +========================================================================================================================================================================== +Total params: 310,205 +Trainable params: 310,205 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 39.71 +========================================================================================================================================================================== +Input size (MB): 1.77 +Forward/backward pass size (MB): 1.11 +Params size (MB): 1.24 +Estimated Total Size (MB): 4.12 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 309 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 310 K +Non-trainable params: 0 +Total params: 310 K +Total estimated model params size (MB): 1.241 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 188.35it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.759759783744812 │ +│ covertype_val_f1-score │ 0.75389164686203 │ +│ covertype_val_loss │ 0.5759305357933044 │ +│ covertype_val_precision │ 0.7565332651138306 │ +│ covertype_val_recall │ 0.759759783744812 │ +│ covertype_val_score │ 0.759759783744812 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 374.77it/s +Validation metrics: [{'covertype_val_loss': 0.5759305357933044, 'lr': 0.0, 'covertype_val_accuracy': 0.759759783744812, 'covertype_val_precision': 0.7565332651138306, 'covertype_val_recall': 0.759759783744812, 'covertype_val_f1-score': 0.75389164686203, 'covertype_val_support': 999.0, 'covertype_val_score': 0.759759783744812}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 0: 12 samples (0.60%) + Class 1: 1027 samples (51.35%) + Class 2: 713 samples (35.65%) + Class 3: 56 samples (2.80%) + Class 4: 140 samples (7.00%) + Class 5: 49 samples (2.45%) + Class 6: 3 samples (0.15%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.7850000262260437 │ +│ covertype_test_f1-score │ 0.7826421856880188 │ +│ covertype_test_loss │ 0.5478729605674744 │ +│ covertype_test_precision │ 0.7864188551902771 │ +│ covertype_test_recall │ 0.7850000262260437 │ +│ covertype_test_score │ 0.7850000262260437 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 402.23it/s +Test metrics: [{'covertype_test_loss': 0.5478729605674744, 'lr': 0.0, 'covertype_test_accuracy': 0.7850000262260437, 'covertype_test_precision': 0.7864188551902771, 'covertype_test_recall': 0.7850000262260437, 'covertype_test_f1-score': 0.7826421856880188, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.7850000262260437}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpaux3weei/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpaux3weei/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpaux3weei/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpaux3weei/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2750 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ covertype │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─EncoderDCNv2: 2-2 [128, 54] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 54] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 54] [128, 128] 7,040 -- 901,120 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +========================================================================================================================================================================== +Total params: 310,205 +Trainable params: 310,205 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 39.71 +========================================================================================================================================================================== +Input size (MB): 1.77 +Forward/backward pass size (MB): 1.11 +Params size (MB): 1.24 +Estimated Total Size (MB): 4.12 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 309 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 310 K +Non-trainable params: 0 +Total params: 310 K +Total estimated model params size (MB): 1.241 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 140.93it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7667667865753174 │ +│ covertype_val_f1-score │ 0.7613160610198975 │ +│ covertype_val_loss │ 0.5856520533561707 │ +│ covertype_val_precision │ 0.7599090933799744 │ +│ covertype_val_recall │ 0.7667667865753174 │ +│ covertype_val_score │ 0.7667667865753174 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 434.55it/s +Validation metrics: [{'covertype_val_loss': 0.5856520533561707, 'lr': 0.0, 'covertype_val_accuracy': 0.7667667865753174, 'covertype_val_precision': 0.7599090933799744, 'covertype_val_recall': 0.7667667865753174, 'covertype_val_f1-score': 0.7613160610198975, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7667667865753174}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 0: 11 samples (0.55%) + Class 1: 995 samples (49.75%) + Class 2: 745 samples (37.25%) + Class 3: 54 samples (2.70%) + Class 4: 143 samples (7.15%) + Class 5: 49 samples (2.45%) + Class 6: 3 samples (0.15%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.7705000042915344 │ +│ covertype_test_f1-score │ 0.7678247094154358 │ +│ covertype_test_loss │ 0.5665362477302551 │ +│ covertype_test_precision │ 0.7719066143035889 │ +│ covertype_test_recall │ 0.7705000042915344 │ +│ covertype_test_score │ 0.7705000042915344 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 149.20it/s +Test metrics: [{'covertype_test_loss': 0.5665362477302551, 'lr': 0.0, 'covertype_test_accuracy': 0.7705000042915344, 'covertype_test_precision': 0.7719066143035889, 'covertype_test_recall': 0.7705000042915344, 'covertype_test_f1-score': 0.7678247094154358, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.7705000042915344}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpaux3weei/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpaux3weei/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpaux3weei/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpaux3weei/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2750 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ covertype │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─EncoderDCNv2: 2-2 [128, 54] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 54] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 54] [128, 128] 7,040 -- 901,120 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +========================================================================================================================================================================== +Total params: 310,205 +Trainable params: 310,205 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 39.71 +========================================================================================================================================================================== +Input size (MB): 1.77 +Forward/backward pass size (MB): 1.11 +Params size (MB): 1.24 +Estimated Total Size (MB): 4.12 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 309 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 310 K +Non-trainable params: 0 +Total params: 310 K +Total estimated model params size (MB): 1.241 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 178.14it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7727727890014648 │ +│ covertype_val_f1-score │ 0.7663441300392151 │ +│ covertype_val_loss │ 0.5612449049949646 │ +│ covertype_val_precision │ 0.767389714717865 │ +│ covertype_val_recall │ 0.7727727890014648 │ +│ covertype_val_score │ 0.7727727890014648 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 394.56it/s +Validation metrics: [{'covertype_val_loss': 0.5612449049949646, 'lr': 0.0, 'covertype_val_accuracy': 0.7727727890014648, 'covertype_val_precision': 0.767389714717865, 'covertype_val_recall': 0.7727727890014648, 'covertype_val_f1-score': 0.7663441300392151, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7727727890014648}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 0: 13 samples (0.65%) + Class 1: 1036 samples (51.80%) + Class 2: 704 samples (35.20%) + Class 3: 57 samples (2.85%) + Class 4: 146 samples (7.30%) + Class 5: 41 samples (2.05%) + Class 6: 3 samples (0.15%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.7770000100135803 │ +│ covertype_test_f1-score │ 0.7740790247917175 │ +│ covertype_test_loss │ 0.5588735938072205 │ +│ covertype_test_precision │ 0.7781260013580322 │ +│ covertype_test_recall │ 0.7770000100135803 │ +│ covertype_test_score │ 0.7770000100135803 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 398.61it/s +Test metrics: [{'covertype_test_loss': 0.5588735938072205, 'lr': 0.0, 'covertype_test_accuracy': 0.7770000100135803, 'covertype_test_precision': 0.7781260013580322, 'covertype_test_recall': 0.7770000100135803, 'covertype_test_f1-score': 0.7740790247917175, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.7770000100135803}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpaux3weei/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpaux3weei/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpaux3weei/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpaux3weei/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2750 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ covertype │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─EncoderDCNv2: 2-2 [128, 54] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 54] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 54] [128, 128] 7,040 -- 901,120 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +========================================================================================================================================================================== +Total params: 310,205 +Trainable params: 310,205 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 39.71 +========================================================================================================================================================================== +Input size (MB): 1.77 +Forward/backward pass size (MB): 1.11 +Params size (MB): 1.24 +Estimated Total Size (MB): 4.12 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 309 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 310 K +Non-trainable params: 0 +Total params: 310 K +Total estimated model params size (MB): 1.241 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 177.80it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7777777910232544 │ +│ covertype_val_f1-score │ 0.773967981338501 │ +│ covertype_val_loss │ 0.5603235960006714 │ +│ covertype_val_precision │ 0.778019368648529 │ +│ covertype_val_recall │ 0.7777777910232544 │ +│ covertype_val_score │ 0.7777777910232544 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 383.51it/s +Validation metrics: [{'covertype_val_loss': 0.5603235960006714, 'lr': 0.0, 'covertype_val_accuracy': 0.7777777910232544, 'covertype_val_precision': 0.778019368648529, 'covertype_val_recall': 0.7777777910232544, 'covertype_val_f1-score': 0.773967981338501, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7777777910232544}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 0: 16 samples (0.80%) + Class 1: 996 samples (49.80%) + Class 2: 732 samples (36.60%) + Class 3: 57 samples (2.85%) + Class 4: 138 samples (6.90%) + Class 5: 57 samples (2.85%) + Class 6: 4 samples (0.20%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.765999972820282 │ +│ covertype_test_f1-score │ 0.7647589445114136 │ +│ covertype_test_loss │ 0.5782790184020996 │ +│ covertype_test_precision │ 0.7669026255607605 │ +│ covertype_test_recall │ 0.765999972820282 │ +│ covertype_test_score │ 0.765999972820282 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 418.24it/s +Test metrics: [{'covertype_test_loss': 0.5782790184020996, 'lr': 0.0, 'covertype_test_accuracy': 0.765999972820282, 'covertype_test_precision': 0.7669026255607605, 'covertype_test_recall': 0.765999972820282, 'covertype_test_f1-score': 0.7647589445114136, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.765999972820282}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpaux3weei/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpaux3weei/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpaux3weei/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpaux3weei/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2750 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ covertype │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─EncoderDCNv2: 2-2 [128, 54] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 54] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 54] [128, 128] 7,040 -- 901,120 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +========================================================================================================================================================================== +Total params: 310,205 +Trainable params: 310,205 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 39.71 +========================================================================================================================================================================== +Input size (MB): 1.77 +Forward/backward pass size (MB): 1.11 +Params size (MB): 1.24 +Estimated Total Size (MB): 4.12 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 309 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 310 K +Non-trainable params: 0 +Total params: 310 K +Total estimated model params size (MB): 1.241 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 129.17it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7727727890014648 │ +│ covertype_val_f1-score │ 0.766056478023529 │ +│ covertype_val_loss │ 0.5580820441246033 │ +│ covertype_val_precision │ 0.7668650150299072 │ +│ covertype_val_recall │ 0.7727727890014648 │ +│ covertype_val_score │ 0.7727727890014648 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 466.80it/s +Validation metrics: [{'covertype_val_loss': 0.5580820441246033, 'lr': 0.0, 'covertype_val_accuracy': 0.7727727890014648, 'covertype_val_precision': 0.7668650150299072, 'covertype_val_recall': 0.7727727890014648, 'covertype_val_f1-score': 0.766056478023529, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7727727890014648}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 0: 17 samples (0.85%) + Class 1: 979 samples (48.95%) + Class 2: 752 samples (37.60%) + Class 3: 56 samples (2.80%) + Class 4: 149 samples (7.45%) + Class 5: 44 samples (2.20%) + Class 6: 3 samples (0.15%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.7764999866485596 │ +│ covertype_test_f1-score │ 0.7748356461524963 │ +│ covertype_test_loss │ 0.5508959889411926 │ +│ covertype_test_precision │ 0.7792395353317261 │ +│ covertype_test_recall │ 0.7764999866485596 │ +│ covertype_test_score │ 0.7764999866485596 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 146.73it/s +Test metrics: [{'covertype_test_loss': 0.5508959889411926, 'lr': 0.0, 'covertype_test_accuracy': 0.7764999866485596, 'covertype_test_precision': 0.7792395353317261, 'covertype_test_recall': 0.7764999866485596, 'covertype_test_f1-score': 0.7748356461524963, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.7764999866485596}] +Wrote results for dcnv2 on covertype to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/downstream_results/dcnv2_covertype_downstream_results.json +Running downstream for dataset 'covertype', model 'resnet'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpz5bs6mob/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpz5bs6mob/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpz5bs6mob/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpz5bs6mob/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─ResNet: 2-2 [128, 54] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 54] [128, 128] 7,040 -- 901,120 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +========================================================================================================================================================================== +Total params: 409,917 +Trainable params: 409,917 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 52.47 +========================================================================================================================================================================== +Input size (MB): 1.77 +Forward/backward pass size (MB): 2.03 +Params size (MB): 1.64 +Estimated Total Size (MB): 5.44 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 409 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 409 K +Non-trainable params: 0 +Total params: 409 K +Total estimated model params size (MB): 1.640 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 179.71it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7477477192878723 │ +│ covertype_val_f1-score │ 0.7452293634414673 │ +│ covertype_val_loss │ 0.5991008281707764 │ +│ covertype_val_precision │ 0.7439250349998474 │ +│ covertype_val_recall │ 0.7477477192878723 │ +│ covertype_val_score │ 0.7477477192878723 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 398.75it/s +Validation metrics: [{'covertype_val_loss': 0.5991008281707764, 'lr': 0.0, 'covertype_val_accuracy': 0.7477477192878723, 'covertype_val_precision': 0.7439250349998474, 'covertype_val_recall': 0.7477477192878723, 'covertype_val_f1-score': 0.7452293634414673, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7477477192878723}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 0: 25 samples (1.25%) + Class 1: 1050 samples (52.50%) + Class 2: 671 samples (33.55%) + Class 3: 61 samples (3.05%) + Class 4: 127 samples (6.35%) + Class 5: 61 samples (3.05%) + Class 6: 5 samples (0.25%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.7739999890327454 │ +│ covertype_test_f1-score │ 0.7733880281448364 │ +│ covertype_test_loss │ 0.5643165111541748 │ +│ covertype_test_precision │ 0.774341881275177 │ +│ covertype_test_recall │ 0.7739999890327454 │ +│ covertype_test_score │ 0.7739999890327454 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 420.28it/s +Test metrics: [{'covertype_test_loss': 0.5643165111541748, 'lr': 0.0, 'covertype_test_accuracy': 0.7739999890327454, 'covertype_test_precision': 0.774341881275177, 'covertype_test_recall': 0.7739999890327454, 'covertype_test_f1-score': 0.7733880281448364, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.7739999890327454}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpz5bs6mob/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpz5bs6mob/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpz5bs6mob/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpz5bs6mob/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─ResNet: 2-2 [128, 54] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 54] [128, 128] 7,040 -- 901,120 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +========================================================================================================================================================================== +Total params: 409,917 +Trainable params: 409,917 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 52.47 +========================================================================================================================================================================== +Input size (MB): 1.77 +Forward/backward pass size (MB): 2.03 +Params size (MB): 1.64 +Estimated Total Size (MB): 5.44 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 409 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 409 K +Non-trainable params: 0 +Total params: 409 K +Total estimated model params size (MB): 1.640 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 168.71it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7557557821273804 │ +│ covertype_val_f1-score │ 0.7491279244422913 │ +│ covertype_val_loss │ 0.6245357394218445 │ +│ covertype_val_precision │ 0.74927818775177 │ +│ covertype_val_recall │ 0.7557557821273804 │ +│ covertype_val_score │ 0.7557557821273804 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 396.60it/s +Validation metrics: [{'covertype_val_loss': 0.6245357394218445, 'lr': 0.0, 'covertype_val_accuracy': 0.7557557821273804, 'covertype_val_precision': 0.74927818775177, 'covertype_val_recall': 0.7557557821273804, 'covertype_val_f1-score': 0.7491279244422913, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7557557821273804}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 0: 51 samples (2.55%) + Class 1: 1098 samples (54.90%) + Class 2: 626 samples (31.30%) + Class 3: 49 samples (2.45%) + Class 4: 136 samples (6.80%) + Class 5: 35 samples (1.75%) + Class 6: 5 samples (0.25%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.7789999842643738 │ +│ covertype_test_f1-score │ 0.777698278427124 │ +│ covertype_test_loss │ 0.5583758354187012 │ +│ covertype_test_precision │ 0.7850139141082764 │ +│ covertype_test_recall │ 0.7789999842643738 │ +│ covertype_test_score │ 0.7789999842643738 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 432.66it/s +Test metrics: [{'covertype_test_loss': 0.5583758354187012, 'lr': 0.0, 'covertype_test_accuracy': 0.7789999842643738, 'covertype_test_precision': 0.7850139141082764, 'covertype_test_recall': 0.7789999842643738, 'covertype_test_f1-score': 0.777698278427124, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.7789999842643738}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpz5bs6mob/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpz5bs6mob/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpz5bs6mob/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpz5bs6mob/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─ResNet: 2-2 [128, 54] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 54] [128, 128] 7,040 -- 901,120 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +========================================================================================================================================================================== +Total params: 409,917 +Trainable params: 409,917 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 52.47 +========================================================================================================================================================================== +Input size (MB): 1.77 +Forward/backward pass size (MB): 2.03 +Params size (MB): 1.64 +Estimated Total Size (MB): 5.44 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 409 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 409 K +Non-trainable params: 0 +Total params: 409 K +Total estimated model params size (MB): 1.640 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 153.08it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7577577829360962 │ +│ covertype_val_f1-score │ 0.7501642107963562 │ +│ covertype_val_loss │ 0.6009353399276733 │ +│ covertype_val_precision │ 0.7495850324630737 │ +│ covertype_val_recall │ 0.7577577829360962 │ +│ covertype_val_score │ 0.7577577829360962 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 218.52it/s +Validation metrics: [{'covertype_val_loss': 0.6009353399276733, 'lr': 0.0, 'covertype_val_accuracy': 0.7577577829360962, 'covertype_val_precision': 0.7495850324630737, 'covertype_val_recall': 0.7577577829360962, 'covertype_val_f1-score': 0.7501642107963562, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7577577829360962}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 0: 21 samples (1.05%) + Class 1: 954 samples (47.70%) + Class 2: 790 samples (39.50%) + Class 3: 57 samples (2.85%) + Class 4: 133 samples (6.65%) + Class 5: 41 samples (2.05%) + Class 6: 4 samples (0.20%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.7705000042915344 │ +│ covertype_test_f1-score │ 0.7685503959655762 │ +│ covertype_test_loss │ 0.5516045093536377 │ +│ covertype_test_precision │ 0.7724647521972656 │ +│ covertype_test_recall │ 0.7705000042915344 │ +│ covertype_test_score │ 0.7705000042915344 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 457.18it/s +Test metrics: [{'covertype_test_loss': 0.5516045093536377, 'lr': 0.0, 'covertype_test_accuracy': 0.7705000042915344, 'covertype_test_precision': 0.7724647521972656, 'covertype_test_recall': 0.7705000042915344, 'covertype_test_f1-score': 0.7685503959655762, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.7705000042915344}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpz5bs6mob/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpz5bs6mob/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpz5bs6mob/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpz5bs6mob/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─ResNet: 2-2 [128, 54] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 54] [128, 128] 7,040 -- 901,120 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +========================================================================================================================================================================== +Total params: 409,917 +Trainable params: 409,917 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 52.47 +========================================================================================================================================================================== +Input size (MB): 1.77 +Forward/backward pass size (MB): 2.03 +Params size (MB): 1.64 +Estimated Total Size (MB): 5.44 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 409 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 409 K +Non-trainable params: 0 +Total params: 409 K +Total estimated model params size (MB): 1.640 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 181.86it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7657657861709595 │ +│ covertype_val_f1-score │ 0.7634060382843018 │ +│ covertype_val_loss │ 0.5863093137741089 │ +│ covertype_val_precision │ 0.7628076076507568 │ +│ covertype_val_recall │ 0.7657657861709595 │ +│ covertype_val_score │ 0.7657657861709595 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 437.50it/s +Validation metrics: [{'covertype_val_loss': 0.5863093137741089, 'lr': 0.0, 'covertype_val_accuracy': 0.7657657861709595, 'covertype_val_precision': 0.7628076076507568, 'covertype_val_recall': 0.7657657861709595, 'covertype_val_f1-score': 0.7634060382843018, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7657657861709595}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 0: 33 samples (1.65%) + Class 1: 980 samples (49.00%) + Class 2: 751 samples (37.55%) + Class 3: 62 samples (3.10%) + Class 4: 110 samples (5.50%) + Class 5: 59 samples (2.95%) + Class 6: 5 samples (0.25%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.7820000052452087 │ +│ covertype_test_f1-score │ 0.7818713784217834 │ +│ covertype_test_loss │ 0.5828087329864502 │ +│ covertype_test_precision │ 0.7827760577201843 │ +│ covertype_test_recall │ 0.7820000052452087 │ +│ covertype_test_score │ 0.7820000052452087 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 428.53it/s +Test metrics: [{'covertype_test_loss': 0.5828087329864502, 'lr': 0.0, 'covertype_test_accuracy': 0.7820000052452087, 'covertype_test_precision': 0.7827760577201843, 'covertype_test_recall': 0.7820000052452087, 'covertype_test_f1-score': 0.7818713784217834, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.7820000052452087}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpz5bs6mob/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpz5bs6mob/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpz5bs6mob/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpz5bs6mob/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─ResNet: 2-2 [128, 54] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 54] [128, 128] 7,040 -- 901,120 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +========================================================================================================================================================================== +Total params: 409,917 +Trainable params: 409,917 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 52.47 +========================================================================================================================================================================== +Input size (MB): 1.77 +Forward/backward pass size (MB): 2.03 +Params size (MB): 1.64 +Estimated Total Size (MB): 5.44 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 409 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 409 K +Non-trainable params: 0 +Total params: 409 K +Total estimated model params size (MB): 1.640 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 190.88it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7627627849578857 │ +│ covertype_val_f1-score │ 0.7572939395904541 │ +│ covertype_val_loss │ 0.5974613428115845 │ +│ covertype_val_precision │ 0.753847062587738 │ +│ covertype_val_recall │ 0.7627627849578857 │ +│ covertype_val_score │ 0.7627627849578857 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 433.24it/s +Validation metrics: [{'covertype_val_loss': 0.5974613428115845, 'lr': 0.0, 'covertype_val_accuracy': 0.7627627849578857, 'covertype_val_precision': 0.753847062587738, 'covertype_val_recall': 0.7627627849578857, 'covertype_val_f1-score': 0.7572939395904541, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7627627849578857}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 0: 44 samples (2.20%) + Class 1: 1000 samples (50.00%) + Class 2: 719 samples (35.95%) + Class 3: 67 samples (3.35%) + Class 4: 129 samples (6.45%) + Class 5: 38 samples (1.90%) + Class 6: 3 samples (0.15%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.7689999938011169 │ +│ covertype_test_f1-score │ 0.7680985331535339 │ +│ covertype_test_loss │ 0.5822221636772156 │ +│ covertype_test_precision │ 0.7705439925193787 │ +│ covertype_test_recall │ 0.7689999938011169 │ +│ covertype_test_score │ 0.7689999938011169 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 415.66it/s +Test metrics: [{'covertype_test_loss': 0.5822221636772156, 'lr': 0.0, 'covertype_test_accuracy': 0.7689999938011169, 'covertype_test_precision': 0.7705439925193787, 'covertype_test_recall': 0.7689999938011169, 'covertype_test_f1-score': 0.7680985331535339, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.7689999938011169}] +Wrote results for resnet on covertype to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/downstream_results/resnet_covertype_downstream_results.json +Running downstream for dataset 'covertype', model 'autoint'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpn8rehn80/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpn8rehn80/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpn8rehn80/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpn8rehn80/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2750 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ covertype │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─EncoderAutoIntBase: 2-2 [128, 54] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 54] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 54] [128, 54, 128] 6,912 -- 884,736 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 54, 54] [256, 54, 54] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 54, 128] [128, 54, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 54, 128] [128, 54, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 54, 54] [256, 54, 54] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 54, 128] [128, 54, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 54, 128] [128, 54, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 54, 54] [256, 54, 54] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 54, 128] [128, 54, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 54, 128] [128, 54, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 6912] [128, 128] 884,864 -- 113,262,592 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +==================================================================================================================================================================================== +Total params: 1,277,885 +Trainable params: 1,277,885 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 163.57 +==================================================================================================================================================================================== +Input size (MB): 1.77 +Forward/backward pass size (MB): 113.44 +Params size (MB): 5.11 +Estimated Total Size (MB): 120.32 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 1.3 M │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 1.3 M +Non-trainable params: 0 +Total params: 1.3 M +Total estimated model params size (MB): 5.112 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 109.05it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7527527809143066 │ +│ covertype_val_f1-score │ 0.748886227607727 │ +│ covertype_val_loss │ 0.6210604906082153 │ +│ covertype_val_precision │ 0.7481434345245361 │ +│ covertype_val_recall │ 0.7527527809143066 │ +│ covertype_val_score │ 0.7527527809143066 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 345.68it/s +Validation metrics: [{'covertype_val_loss': 0.6210604906082153, 'lr': 0.0, 'covertype_val_accuracy': 0.7527527809143066, 'covertype_val_precision': 0.7481434345245361, 'covertype_val_recall': 0.7527527809143066, 'covertype_val_f1-score': 0.748886227607727, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7527527809143066}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 0: 18 samples (0.90%) + Class 1: 1022 samples (51.10%) + Class 2: 715 samples (35.75%) + Class 3: 49 samples (2.45%) + Class 4: 131 samples (6.55%) + Class 5: 61 samples (3.05%) + Class 6: 4 samples (0.20%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.753000020980835 │ +│ covertype_test_f1-score │ 0.7517149448394775 │ +│ covertype_test_loss │ 0.5905119776725769 │ +│ covertype_test_precision │ 0.7534472942352295 │ +│ covertype_test_recall │ 0.753000020980835 │ +│ covertype_test_score │ 0.753000020980835 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 198.56it/s +Test metrics: [{'covertype_test_loss': 0.5905119776725769, 'lr': 0.0, 'covertype_test_accuracy': 0.753000020980835, 'covertype_test_precision': 0.7534472942352295, 'covertype_test_recall': 0.753000020980835, 'covertype_test_f1-score': 0.7517149448394775, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.753000020980835}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpn8rehn80/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpn8rehn80/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpn8rehn80/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpn8rehn80/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2750 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ covertype │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─EncoderAutoIntBase: 2-2 [128, 54] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 54] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 54] [128, 54, 128] 6,912 -- 884,736 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 54, 54] [256, 54, 54] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 54, 128] [128, 54, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 54, 128] [128, 54, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 54, 54] [256, 54, 54] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 54, 128] [128, 54, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 54, 128] [128, 54, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 54, 54] [256, 54, 54] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 54, 128] [128, 54, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 54, 128] [128, 54, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 6912] [128, 128] 884,864 -- 113,262,592 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +==================================================================================================================================================================================== +Total params: 1,277,885 +Trainable params: 1,277,885 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 163.57 +==================================================================================================================================================================================== +Input size (MB): 1.77 +Forward/backward pass size (MB): 113.44 +Params size (MB): 5.11 +Estimated Total Size (MB): 120.32 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 1.3 M │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 1.3 M +Non-trainable params: 0 +Total params: 1.3 M +Total estimated model params size (MB): 5.112 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 110.86it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7467467188835144 │ +│ covertype_val_f1-score │ 0.7426261901855469 │ +│ covertype_val_loss │ 0.6148595213890076 │ +│ covertype_val_precision │ 0.7431632876396179 │ +│ covertype_val_recall │ 0.7467467188835144 │ +│ covertype_val_score │ 0.7467467188835144 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 337.23it/s +Validation metrics: [{'covertype_val_loss': 0.6148595213890076, 'lr': 0.0, 'covertype_val_accuracy': 0.7467467188835144, 'covertype_val_precision': 0.7431632876396179, 'covertype_val_recall': 0.7467467188835144, 'covertype_val_f1-score': 0.7426261901855469, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7467467188835144}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 0: 14 samples (0.70%) + Class 1: 985 samples (49.25%) + Class 2: 750 samples (37.50%) + Class 3: 56 samples (2.80%) + Class 4: 138 samples (6.90%) + Class 5: 52 samples (2.60%) + Class 6: 5 samples (0.25%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.7484999895095825 │ +│ covertype_test_f1-score │ 0.7457489371299744 │ +│ covertype_test_loss │ 0.5982586741447449 │ +│ covertype_test_precision │ 0.7465230822563171 │ +│ covertype_test_recall │ 0.7484999895095825 │ +│ covertype_test_score │ 0.7484999895095825 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 327.58it/s +Test metrics: [{'covertype_test_loss': 0.5982586741447449, 'lr': 0.0, 'covertype_test_accuracy': 0.7484999895095825, 'covertype_test_precision': 0.7465230822563171, 'covertype_test_recall': 0.7484999895095825, 'covertype_test_f1-score': 0.7457489371299744, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.7484999895095825}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpn8rehn80/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpn8rehn80/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpn8rehn80/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpn8rehn80/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2750 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ covertype │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─EncoderAutoIntBase: 2-2 [128, 54] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 54] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 54] [128, 54, 128] 6,912 -- 884,736 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 54, 54] [256, 54, 54] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 54, 128] [128, 54, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 54, 128] [128, 54, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 54, 54] [256, 54, 54] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 54, 128] [128, 54, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 54, 128] [128, 54, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 54, 54] [256, 54, 54] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 54, 128] [128, 54, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 54, 128] [128, 54, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 6912] [128, 128] 884,864 -- 113,262,592 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +==================================================================================================================================================================================== +Total params: 1,277,885 +Trainable params: 1,277,885 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 163.57 +==================================================================================================================================================================================== +Input size (MB): 1.77 +Forward/backward pass size (MB): 113.44 +Params size (MB): 5.11 +Estimated Total Size (MB): 120.32 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 1.3 M │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 1.3 M +Non-trainable params: 0 +Total params: 1.3 M +Total estimated model params size (MB): 5.112 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 48/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 123.99it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7577577829360962 │ +│ covertype_val_f1-score │ 0.7554540634155273 │ +│ covertype_val_loss │ 0.5999589562416077 │ +│ covertype_val_precision │ 0.7574031949043274 │ +│ covertype_val_recall │ 0.7577577829360962 │ +│ covertype_val_score │ 0.7577577829360962 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 9.866357686405536e-07 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 305.69it/s +Validation metrics: [{'covertype_val_loss': 0.5999589562416077, 'lr': 9.866357686405536e-07, 'covertype_val_accuracy': 0.7577577829360962, 'covertype_val_precision': 0.7574031949043274, 'covertype_val_recall': 0.7577577829360962, 'covertype_val_f1-score': 0.7554540634155273, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7577577829360962}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 0: 11 samples (0.55%) + Class 1: 1002 samples (50.10%) + Class 2: 728 samples (36.40%) + Class 3: 59 samples (2.95%) + Class 4: 134 samples (6.70%) + Class 5: 62 samples (3.10%) + Class 6: 4 samples (0.20%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.7574999928474426 │ +│ covertype_test_f1-score │ 0.7547937035560608 │ +│ covertype_test_loss │ 0.5876102447509766 │ +│ covertype_test_precision │ 0.7553665637969971 │ +│ covertype_test_recall │ 0.7574999928474426 │ +│ covertype_test_score │ 0.7574999928474426 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 9.866358823273913e-07 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 317.40it/s +Test metrics: [{'covertype_test_loss': 0.5876102447509766, 'lr': 9.866358823273913e-07, 'covertype_test_accuracy': 0.7574999928474426, 'covertype_test_precision': 0.7553665637969971, 'covertype_test_recall': 0.7574999928474426, 'covertype_test_f1-score': 0.7547937035560608, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.7574999928474426}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpn8rehn80/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpn8rehn80/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpn8rehn80/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpn8rehn80/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2750 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ covertype │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─EncoderAutoIntBase: 2-2 [128, 54] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 54] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 54] [128, 54, 128] 6,912 -- 884,736 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 54, 54] [256, 54, 54] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 54, 128] [128, 54, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 54, 128] [128, 54, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 54, 54] [256, 54, 54] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 54, 128] [128, 54, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 54, 128] [128, 54, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 54, 54] [256, 54, 54] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 54, 128] [128, 54, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 54, 128] [128, 54, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 6912] [128, 128] 884,864 -- 113,262,592 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +==================================================================================================================================================================================== +Total params: 1,277,885 +Trainable params: 1,277,885 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 163.57 +==================================================================================================================================================================================== +Input size (MB): 1.77 +Forward/backward pass size (MB): 113.44 +Params size (MB): 5.11 +Estimated Total Size (MB): 120.32 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 1.3 M │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 1.3 M +Non-trainable params: 0 +Total params: 1.3 M +Total estimated model params size (MB): 5.112 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 110.97it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7267267107963562 │ +│ covertype_val_f1-score │ 0.7243560552597046 │ +│ covertype_val_loss │ 0.6177465319633484 │ +│ covertype_val_precision │ 0.7270073890686035 │ +│ covertype_val_recall │ 0.7267267107963562 │ +│ covertype_val_score │ 0.7267267107963562 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 348.94it/s +Validation metrics: [{'covertype_val_loss': 0.6177465319633484, 'lr': 0.0, 'covertype_val_accuracy': 0.7267267107963562, 'covertype_val_precision': 0.7270073890686035, 'covertype_val_recall': 0.7267267107963562, 'covertype_val_f1-score': 0.7243560552597046, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7267267107963562}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 0: 19 samples (0.95%) + Class 1: 986 samples (49.30%) + Class 2: 749 samples (37.45%) + Class 3: 58 samples (2.90%) + Class 4: 135 samples (6.75%) + Class 5: 48 samples (2.40%) + Class 6: 5 samples (0.25%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.7565000057220459 │ +│ covertype_test_f1-score │ 0.7551760673522949 │ +│ covertype_test_loss │ 0.5888822078704834 │ +│ covertype_test_precision │ 0.7572526335716248 │ +│ covertype_test_recall │ 0.7565000057220459 │ +│ covertype_test_score │ 0.7565000057220459 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 183.22it/s +Test metrics: [{'covertype_test_loss': 0.5888822078704834, 'lr': 0.0, 'covertype_test_accuracy': 0.7565000057220459, 'covertype_test_precision': 0.7572526335716248, 'covertype_test_recall': 0.7565000057220459, 'covertype_test_f1-score': 0.7551760673522949, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.7565000057220459}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpn8rehn80/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpn8rehn80/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpn8rehn80/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpn8rehn80/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2750 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ covertype │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─EncoderAutoIntBase: 2-2 [128, 54] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 54] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 54] [128, 54, 128] 6,912 -- 884,736 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 54, 54] [256, 54, 54] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 54, 128] [128, 54, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 54, 128] [128, 54, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 54, 54] [256, 54, 54] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 54, 128] [128, 54, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 54, 128] [128, 54, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 54, 128] [128, 54, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 54, 54] [256, 54, 54] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 54, 128] [128, 54, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 54, 128] [128, 54, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 54, 128] [128, 54, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 6912] [128, 128] 884,864 -- 113,262,592 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +==================================================================================================================================================================================== +Total params: 1,277,885 +Trainable params: 1,277,885 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 163.57 +==================================================================================================================================================================================== +Input size (MB): 1.77 +Forward/backward pass size (MB): 113.44 +Params size (MB): 5.11 +Estimated Total Size (MB): 120.32 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 1.3 M │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 1.3 M +Non-trainable params: 0 +Total params: 1.3 M +Total estimated model params size (MB): 5.112 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 43/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 115.87it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7437437176704407 │ +│ covertype_val_f1-score │ 0.7389692664146423 │ +│ covertype_val_loss │ 0.6296648979187012 │ +│ covertype_val_precision │ 0.7408604025840759 │ +│ covertype_val_recall │ 0.7437437176704407 │ +│ covertype_val_score │ 0.7437437176704407 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 3.511175964376889e-05 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 316.15it/s +Validation metrics: [{'covertype_val_loss': 0.6296648979187012, 'lr': 3.511175964376889e-05, 'covertype_val_accuracy': 0.7437437176704407, 'covertype_val_precision': 0.7408604025840759, 'covertype_val_recall': 0.7437437176704407, 'covertype_val_f1-score': 0.7389692664146423, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7437437176704407}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 0: 8 samples (0.40%) + Class 1: 978 samples (48.90%) + Class 2: 760 samples (38.00%) + Class 3: 56 samples (2.80%) + Class 4: 143 samples (7.15%) + Class 5: 49 samples (2.45%) + Class 6: 6 samples (0.30%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.7455000281333923 │ +│ covertype_test_f1-score │ 0.7421423196792603 │ +│ covertype_test_loss │ 0.5885452628135681 │ +│ covertype_test_precision │ 0.7455663681030273 │ +│ covertype_test_recall │ 0.7455000281333923 │ +│ covertype_test_score │ 0.7455000281333923 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 3.511175964376889e-05 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 319.38it/s +Test metrics: [{'covertype_test_loss': 0.5885452628135681, 'lr': 3.511175964376889e-05, 'covertype_test_accuracy': 0.7455000281333923, 'covertype_test_precision': 0.7455663681030273, 'covertype_test_recall': 0.7455000281333923, 'covertype_test_f1-score': 0.7421423196792603, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.7455000281333923}] +Wrote results for autoint on covertype to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/downstream_results/autoint_covertype_downstream_results.json +Running downstream for dataset 'covertype', model 'fttransformer'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpumvcdv10/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpumvcdv10/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpumvcdv10/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpumvcdv10/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2750 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ covertype │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─EncoderFTTransformer: 2-2 [128, 54] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 54] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 54] [128, 54, 128] 13,824 -- 1,769,472 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 55, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 55, 55] [512, 55, 55] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 55, 128] [128, 55, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 55, 512] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 55, 256] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 55, 256] [128, 55, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 55, 55] [512, 55, 55] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 55, 128] [128, 55, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 55, 512] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 55, 256] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 55, 256] [128, 55, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 55] [512, 1, 55] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 55, 128] [128, 55, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 55, 512] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 55, 256] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 55, 256] [128, 55, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +========================================================================================================================================================================================= +Total params: 714,557 +Trainable params: 714,557 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 91.46 +========================================================================================================================================================================================= +Input size (MB): 1.77 +Forward/backward pass size (MB): 224.06 +Params size (MB): 2.86 +Estimated Total Size (MB): 228.69 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 713 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 714 K +Non-trainable params: 0 +Total params: 714 K +Total estimated model params size (MB): 2.858 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 72.00it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7477477192878723 │ +│ covertype_val_f1-score │ 0.7361925840377808 │ +│ covertype_val_loss │ 0.6155408620834351 │ +│ covertype_val_precision │ 0.7297002077102661 │ +│ covertype_val_recall │ 0.7477477192878723 │ +│ covertype_val_score │ 0.7477477192878723 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 181.57it/s +Validation metrics: [{'covertype_val_loss': 0.6155408620834351, 'lr': 0.0, 'covertype_val_accuracy': 0.7477477192878723, 'covertype_val_precision': 0.7297002077102661, 'covertype_val_recall': 0.7477477192878723, 'covertype_val_f1-score': 0.7361925840377808, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7477477192878723}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 1: 1030 samples (51.50%) + Class 2: 737 samples (36.85%) + Class 3: 50 samples (2.50%) + Class 4: 157 samples (7.85%) + Class 5: 21 samples (1.05%) + Class 6: 5 samples (0.25%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.7390000224113464 │ +│ covertype_test_f1-score │ 0.7285676598548889 │ +│ covertype_test_loss │ 0.5945818424224854 │ +│ covertype_test_precision │ 0.7274198532104492 │ +│ covertype_test_recall │ 0.7390000224113464 │ +│ covertype_test_score │ 0.7390000224113464 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 200.73it/s +Test metrics: [{'covertype_test_loss': 0.5945818424224854, 'lr': 0.0, 'covertype_test_accuracy': 0.7390000224113464, 'covertype_test_precision': 0.7274198532104492, 'covertype_test_recall': 0.7390000224113464, 'covertype_test_f1-score': 0.7285676598548889, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.7390000224113464}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpumvcdv10/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpumvcdv10/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpumvcdv10/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpumvcdv10/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2750 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ covertype │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─EncoderFTTransformer: 2-2 [128, 54] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 54] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 54] [128, 54, 128] 13,824 -- 1,769,472 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 55, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 55, 55] [512, 55, 55] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 55, 128] [128, 55, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 55, 512] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 55, 256] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 55, 256] [128, 55, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 55, 55] [512, 55, 55] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 55, 128] [128, 55, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 55, 512] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 55, 256] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 55, 256] [128, 55, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 55] [512, 1, 55] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 55, 128] [128, 55, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 55, 512] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 55, 256] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 55, 256] [128, 55, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +========================================================================================================================================================================================= +Total params: 714,557 +Trainable params: 714,557 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 91.46 +========================================================================================================================================================================================= +Input size (MB): 1.77 +Forward/backward pass size (MB): 224.06 +Params size (MB): 2.86 +Estimated Total Size (MB): 228.69 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 713 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 714 K +Non-trainable params: 0 +Total params: 714 K +Total estimated model params size (MB): 2.858 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 80.98it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7457457184791565 │ +│ covertype_val_f1-score │ 0.7360843420028687 │ +│ covertype_val_loss │ 0.6016526222229004 │ +│ covertype_val_precision │ 0.7388072609901428 │ +│ covertype_val_recall │ 0.7457457184791565 │ +│ covertype_val_score │ 0.7457457184791565 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 261.24it/s +Validation metrics: [{'covertype_val_loss': 0.6016526222229004, 'lr': 0.0, 'covertype_val_accuracy': 0.7457457184791565, 'covertype_val_precision': 0.7388072609901428, 'covertype_val_recall': 0.7457457184791565, 'covertype_val_f1-score': 0.7360843420028687, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7457457184791565}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 0: 4 samples (0.20%) + Class 1: 1008 samples (50.40%) + Class 2: 751 samples (37.55%) + Class 3: 56 samples (2.80%) + Class 4: 154 samples (7.70%) + Class 5: 24 samples (1.20%) + Class 6: 3 samples (0.15%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.7404999732971191 │ +│ covertype_test_f1-score │ 0.7323582172393799 │ +│ covertype_test_loss │ 0.5934622883796692 │ +│ covertype_test_precision │ 0.7396467924118042 │ +│ covertype_test_recall │ 0.7404999732971191 │ +│ covertype_test_score │ 0.7404999732971191 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 258.29it/s +Test metrics: [{'covertype_test_loss': 0.5934622883796692, 'lr': 0.0, 'covertype_test_accuracy': 0.7404999732971191, 'covertype_test_precision': 0.7396467924118042, 'covertype_test_recall': 0.7404999732971191, 'covertype_test_f1-score': 0.7323582172393799, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.7404999732971191}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpumvcdv10/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpumvcdv10/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpumvcdv10/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpumvcdv10/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2750 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ covertype │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─EncoderFTTransformer: 2-2 [128, 54] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 54] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 54] [128, 54, 128] 13,824 -- 1,769,472 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 55, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 55, 55] [512, 55, 55] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 55, 128] [128, 55, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 55, 512] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 55, 256] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 55, 256] [128, 55, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 55, 55] [512, 55, 55] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 55, 128] [128, 55, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 55, 512] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 55, 256] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 55, 256] [128, 55, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 55] [512, 1, 55] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 55, 128] [128, 55, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 55, 512] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 55, 256] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 55, 256] [128, 55, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +========================================================================================================================================================================================= +Total params: 714,557 +Trainable params: 714,557 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 91.46 +========================================================================================================================================================================================= +Input size (MB): 1.77 +Forward/backward pass size (MB): 224.06 +Params size (MB): 2.86 +Estimated Total Size (MB): 228.69 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 713 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 714 K +Non-trainable params: 0 +Total params: 714 K +Total estimated model params size (MB): 2.858 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 75.68it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7427427172660828 │ +│ covertype_val_f1-score │ 0.7351154685020447 │ +│ covertype_val_loss │ 0.5965853929519653 │ +│ covertype_val_precision │ 0.7358342409133911 │ +│ covertype_val_recall │ 0.7427427172660828 │ +│ covertype_val_score │ 0.7427427172660828 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 239.35it/s +Validation metrics: [{'covertype_val_loss': 0.5965853929519653, 'lr': 0.0, 'covertype_val_accuracy': 0.7427427172660828, 'covertype_val_precision': 0.7358342409133911, 'covertype_val_recall': 0.7427427172660828, 'covertype_val_f1-score': 0.7351154685020447, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7427427172660828}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 0: 5 samples (0.25%) + Class 1: 1022 samples (51.10%) + Class 2: 729 samples (36.45%) + Class 3: 53 samples (2.65%) + Class 4: 146 samples (7.30%) + Class 5: 42 samples (2.10%) + Class 6: 3 samples (0.15%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.7415000200271606 │ +│ covertype_test_f1-score │ 0.7356262803077698 │ +│ covertype_test_loss │ 0.5890328884124756 │ +│ covertype_test_precision │ 0.7395013570785522 │ +│ covertype_test_recall │ 0.7415000200271606 │ +│ covertype_test_score │ 0.7415000200271606 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 250.75it/s +Test metrics: [{'covertype_test_loss': 0.5890328884124756, 'lr': 0.0, 'covertype_test_accuracy': 0.7415000200271606, 'covertype_test_precision': 0.7395013570785522, 'covertype_test_recall': 0.7415000200271606, 'covertype_test_f1-score': 0.7356262803077698, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.7415000200271606}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpumvcdv10/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpumvcdv10/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpumvcdv10/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpumvcdv10/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2750 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ covertype │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─EncoderFTTransformer: 2-2 [128, 54] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 54] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 54] [128, 54, 128] 13,824 -- 1,769,472 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 55, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 55, 55] [512, 55, 55] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 55, 128] [128, 55, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 55, 512] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 55, 256] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 55, 256] [128, 55, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 55, 55] [512, 55, 55] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 55, 128] [128, 55, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 55, 512] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 55, 256] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 55, 256] [128, 55, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 55] [512, 1, 55] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 55, 128] [128, 55, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 55, 512] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 55, 256] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 55, 256] [128, 55, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +========================================================================================================================================================================================= +Total params: 714,557 +Trainable params: 714,557 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 91.46 +========================================================================================================================================================================================= +Input size (MB): 1.77 +Forward/backward pass size (MB): 224.06 +Params size (MB): 2.86 +Estimated Total Size (MB): 228.69 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 713 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 714 K +Non-trainable params: 0 +Total params: 714 K +Total estimated model params size (MB): 2.858 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 87.52it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7527527809143066 │ +│ covertype_val_f1-score │ 0.7450913190841675 │ +│ covertype_val_loss │ 0.6058993935585022 │ +│ covertype_val_precision │ 0.7523147463798523 │ +│ covertype_val_recall │ 0.7527527809143066 │ +│ covertype_val_score │ 0.7527527809143066 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 205.84it/s +Validation metrics: [{'covertype_val_loss': 0.6058993935585022, 'lr': 0.0, 'covertype_val_accuracy': 0.7527527809143066, 'covertype_val_precision': 0.7523147463798523, 'covertype_val_recall': 0.7527527809143066, 'covertype_val_f1-score': 0.7450913190841675, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7527527809143066}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 0: 6 samples (0.30%) + Class 1: 971 samples (48.55%) + Class 2: 778 samples (38.90%) + Class 3: 52 samples (2.60%) + Class 4: 140 samples (7.00%) + Class 5: 51 samples (2.55%) + Class 6: 2 samples (0.10%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.75 │ +│ covertype_test_f1-score │ 0.7460992336273193 │ +│ covertype_test_loss │ 0.5745090246200562 │ +│ covertype_test_precision │ 0.7542448043823242 │ +│ covertype_test_recall │ 0.75 │ +│ covertype_test_score │ 0.75 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 181.19it/s +Test metrics: [{'covertype_test_loss': 0.5745090246200562, 'lr': 0.0, 'covertype_test_accuracy': 0.75, 'covertype_test_precision': 0.7542448043823242, 'covertype_test_recall': 0.75, 'covertype_test_f1-score': 0.7460992336273193, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.75}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 --data_set=covertype --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpumvcdv10/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpumvcdv10/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ covertype │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpumvcdv10/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpumvcdv10/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 54 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 55 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 7 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 54, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 54 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2750 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ covertype │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 54, 64] [128, 7] -- -- -- +├─Sequential: 1-1 [128, 54, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 54, 64] [128, 54] -- -- -- +│ │ └─Linear: 3-1 [128, 3456] [128, 54] 186,678 -- 23,894,784 +│ └─EncoderFTTransformer: 2-2 [128, 54] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 54] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 54] [128, 54, 128] 13,824 -- 1,769,472 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 55, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 55, 55] [512, 55, 55] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 55, 128] [128, 55, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 55, 512] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 55, 256] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 55, 256] [128, 55, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 55, 55] [512, 55, 55] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 55, 128] [128, 55, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 55, 512] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 55, 256] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 55, 256] [128, 55, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 55, 128] [128, 55, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 55] [512, 1, 55] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 55, 128] [128, 55, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 55, 128] [128, 55, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 55, 512] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 55, 256] [128, 55, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 55, 256] [128, 55, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 55, 128] [128, 55, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 7] 903 -- 115,584 +========================================================================================================================================================================================= +Total params: 714,557 +Trainable params: 714,557 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 91.46 +========================================================================================================================================================================================= +Input size (MB): 1.77 +Forward/backward pass size (MB): 224.06 +Params size (MB): 2.86 +Estimated Total Size (MB): 228.69 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 903 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 713 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 714 K +Non-trainable params: 0 +Total params: 714 K +Total estimated model params size (MB): 2.858 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 55/55 0:00:00 • 0:00:00 86.11it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_val_accuracy │ 0.7527527809143066 │ +│ covertype_val_f1-score │ 0.7474658489227295 │ +│ covertype_val_loss │ 0.5903725028038025 │ +│ covertype_val_precision │ 0.7500065565109253 │ +│ covertype_val_recall │ 0.7527527809143066 │ +│ covertype_val_score │ 0.7527527809143066 │ +│ covertype_val_support │ 999.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 135.00it/s +Validation metrics: [{'covertype_val_loss': 0.5903725028038025, 'lr': 0.0, 'covertype_val_accuracy': 0.7527527809143066, 'covertype_val_precision': 0.7500065565109253, 'covertype_val_recall': 0.7527527809143066, 'covertype_val_f1-score': 0.7474658489227295, 'covertype_val_support': 999.0, 'covertype_val_score': 0.7527527809143066}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 32 samples (1.60%) + Class 1: 1017 samples (50.85%) + Class 2: 708 samples (35.40%) + Class 3: 64 samples (3.20%) + Class 4: 108 samples (5.40%) + Class 5: 63 samples (3.15%) + Class 6: 8 samples (0.40%) +Predicted class frequencies: + Class 0: 9 samples (0.45%) + Class 1: 1000 samples (50.00%) + Class 2: 736 samples (36.80%) + Class 3: 62 samples (3.10%) + Class 4: 148 samples (7.40%) + Class 5: 42 samples (2.10%) + Class 6: 3 samples (0.15%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ covertype_test_accuracy │ 0.7494999766349792 │ +│ covertype_test_f1-score │ 0.7457274794578552 │ +│ covertype_test_loss │ 0.5884156227111816 │ +│ covertype_test_precision │ 0.7507426738739014 │ +│ covertype_test_recall │ 0.7494999766349792 │ +│ covertype_test_score │ 0.7494999766349792 │ +│ covertype_test_support │ 2000.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 181.98it/s +Test metrics: [{'covertype_test_loss': 0.5884156227111816, 'lr': 0.0, 'covertype_test_accuracy': 0.7494999766349792, 'covertype_test_precision': 0.7507426738739014, 'covertype_test_recall': 0.7494999766349792, 'covertype_test_f1-score': 0.7457274794578552, 'covertype_test_support': 2000.0, 'covertype_test_score': 0.7494999766349792}] +Wrote results for fttransformer on covertype to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/downstream_results/fttransformer_covertype_downstream_results.json + +Downstream pipeline complete. Results under /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_182951/downstream_results diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu3-credit.exit_code b/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu3-credit.exit_code new file mode 100644 index 0000000000000000000000000000000000000000..573541ac9702dd3969c9bc859d2b91ec1f7e6e56 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu3-credit.exit_code @@ -0,0 +1 @@ +0 diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu3-credit.log b/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu3-credit.log new file mode 100644 index 0000000000000000000000000000000000000000..e672418c5a46f684d6c1dae3b4b2bec77175e81e --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu3-credit.log @@ -0,0 +1,12252 @@ +[gpu3-credit] Starting credit_card_default on physical GPU 3 at Tue Sep 15 18:28:13 UTC 2026 ++ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py --dataset credit_card_default --results-root /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42 --epochs 500 --objective w_jepa_joint --latent-architecture deterministic_no_z --sw-full-end-weight 0.00125 --torch-seed 42 --np-seed 42 --mask-seed 0 --sw-seed 0 --sw-full-num-projections 2048 --sw-marginal-num-projections 1024 --sw-p 2 --sw-s-x-end-weight 0.001 --sw-z-end-weight 0.0 --sw-s-y-end-weight 0.0001 --model-selection unfiltered --deterministic-mode warn --sw-include-marginals --batch-size 512 +{ + "dataset": "credit_card_default", + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "credit_card_default", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.00125", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.001", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "0.0001", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} +Not SLURM job. +╒══════════════════════════════════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ mock │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ activation_checkpointing │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 42 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 42 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_seed │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_seed │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ deterministic_mode │ warn │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pre_evaluation │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 512 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 500 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ improvement │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_val_cache_cadence │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ init_type │ trunc_normal │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_start_lr │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_lr │ 0.001 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_final_lr │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 10 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_final_weight_decay │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_ipe_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_gradient_clipping │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_feature_type_embedding │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_feature_index_embedding │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dropout_prob │ 0.001 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_layer_norm_eps │ 1e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dim_hidden │ 64 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dim_feedforward │ 256 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_num_heads │ 4 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_num_layers │ 8 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_act_func │ relu │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ probe_cadence │ 2 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ probe_model │ linear_probe │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ n_cls_tokens │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_type │ transformer │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_num_layers │ 16 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_embed_dim │ 16 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_num_heads │ 4 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_p_dropout │ 0.002 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_layer_norm_eps │ 1e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_activation │ relu │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_dim_feedforward │ 256 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_allow_overlap │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_min_ctx_share │ 0.1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_max_ctx_share │ 0.3 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_min_trgt_share │ 0.1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_max_trgt_share │ 0.6 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_num_preds │ 4 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_num_encs │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mu_init_std │ 0.03 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ logvar_init_std │ 0.005 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ logvar_init_bias │ -4.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_sigma_x_init │ -2.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_sigma_y_init │ -2.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ downstream_evaluation_epochs │ 50 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ downstream_probe_patience │ 20 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ probe_representation_type │ mean │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ probe_encoder_type │ target_posterior │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pooled_context_representation_size_for_target_posterior │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pooled_context_representation_size_for_auxiliary_encoder │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_end_weight │ 0.0001 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_end_weight │ 1e-06 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_end_weight │ 1e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_annealing_epochs │ 15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_annealing_epochs │ 15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_annealing_epochs │ 15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ recon_x_weight │ 0.1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ recon_y_weight │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ latent_architecture │ deterministic_no_z │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ jepa_objective │ w_jepa_joint │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_full_num_projections │ 2048 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_marginal_num_projections │ 1024 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_p │ 2 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_include_marginals │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_s_x_weight │ 0.001 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_z_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_s_y_weight │ 0.0001 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_end_weight │ 0.00125 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_annealing_epochs │ 15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_marginal_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_marginal_end_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_end_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_start_epoch │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_annealing_epochs │ 30 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_end_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_start_epoch │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_annealing_epochs │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_end_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_start_epoch │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_annealing_epochs │ 30 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ safe_exit │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_selection │ unfiltered │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sigma_aggregation │ mean │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ subset_size │ -1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ is_batchlearning │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ iteration │ 0 │ +╘══════════════════════════════════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ + +CUDA Device Information: +CUDA_VISIBLE_DEVICES: 3 +Number of CUDA devices: 1 +Current CUDA device: 0 +CUDA device name: NVIDIA A100-SXM4-80GB + +Using device: cuda:0 +Device name: NVIDIA A100-SXM4-80GB +Encoder Module parameters: ++-------------------+------------------------------------------------------------------------------------+ +| Parameter | Value | ++===================+====================================================================================+ +| hidden_dim | 64 | ++-------------------+------------------------------------------------------------------------------------+ +| idx_num_features | [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22] | ++-------------------+------------------------------------------------------------------------------------+ +| idx_cat_features | [] | ++-------------------+------------------------------------------------------------------------------------+ +| cardinalities | [] | ++-------------------+------------------------------------------------------------------------------------+ +| n_input_features | 23 | ++-------------------+------------------------------------------------------------------------------------+ +| device | cuda:0 | ++-------------------+------------------------------------------------------------------------------------+ +| num_layers | 8 | ++-------------------+------------------------------------------------------------------------------------+ +| num_heads | 4 | ++-------------------+------------------------------------------------------------------------------------+ +| p_dropout | 0.001 | ++-------------------+------------------------------------------------------------------------------------+ +| layer_norm_eps | 1e-05 | ++-------------------+------------------------------------------------------------------------------------+ +| gradient_clipping | 1.0 | ++-------------------+------------------------------------------------------------------------------------+ +| dim_feedforward | 256 | ++-------------------+------------------------------------------------------------------------------------+ +Encoder Module parameters: ++-------------------+------------------------------------------------------------------------------------+ +| Parameter | Value | ++===================+====================================================================================+ +| hidden_dim | 64 | ++-------------------+------------------------------------------------------------------------------------+ +| idx_num_features | [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22] | ++-------------------+------------------------------------------------------------------------------------+ +| idx_cat_features | [] | ++-------------------+------------------------------------------------------------------------------------+ +| cardinalities | [] | ++-------------------+------------------------------------------------------------------------------------+ +| n_input_features | 23 | ++-------------------+------------------------------------------------------------------------------------+ +| device | cuda:0 | ++-------------------+------------------------------------------------------------------------------------+ +| num_layers | 8 | ++-------------------+------------------------------------------------------------------------------------+ +| num_heads | 4 | ++-------------------+------------------------------------------------------------------------------------+ +| p_dropout | 0.001 | ++-------------------+------------------------------------------------------------------------------------+ +| layer_norm_eps | 1e-05 | ++-------------------+------------------------------------------------------------------------------------+ +| gradient_clipping | 1.0 | ++-------------------+------------------------------------------------------------------------------------+ +| dim_feedforward | 256 | ++-------------------+------------------------------------------------------------------------------------+ +DJEPA parameters: ++-------------------------+---------+ +| Parameter | Value | ++=========================+=========+ +| hidden_dim | 64 | ++-------------------------+---------+ +| num_layers | 8 | ++-------------------------+---------+ +| num_heads | 4 | ++-------------------------+---------+ +| p_dropout | 0.001 | ++-------------------------+---------+ +| layer_norm_eps | 1e-05 | ++-------------------------+---------+ +| feature_type_embedding | True | ++-------------------------+---------+ +| feature_index_embedding | True | ++-------------------------+---------+ +| dim_feedforward | 256 | ++-------------------------+---------+ +| use_reg_token | True | ++-------------------------+---------+ +| gradient_clipping | 1.0 | ++-------------------------+---------+ +| activation | gelu | ++-------------------------+---------+ +| num_features | 23 | ++-------------------------+---------+ +| n_cls_tokens | 1 | ++-------------------------+---------+ +MaskCollator params: +----------------- ----- +Allow overlap False +Min context share 0.1 +Max context share 0.3 +Min target share 0.1 +Max target share 0.6 +Min context 2 +Max context 6 +Min target 2 +Max target 13 +Num preds 4 +Num encs 1 +Num features 23 +Cardinalities [] +Base mask seed 0 +----------------- ----- +KL-x annealing: start=0.0 for 59 steps, then anneal to 0.0001 over 885 steps. +KL-z annealing: start=0.0 for 59 steps, then anneal to 1e-06 over 885 steps. +KL-y annealing: start=0.0 for 59 steps, then anneal to 1e-05 over 885 steps. +W-JEPA marginal-scale annealing (s_x): start=0.0, end=1.0, start_epoch=0, anneal_epochs=30. +W-JEPA marginal-scale annealing (z): start=0.0, end=0.0, start_epoch=0, anneal_epochs=0. +W-JEPA marginal-scale annealing (s_y): start=0.0, end=1.0, start_epoch=0, anneal_epochs=30. +Configuration saved to: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/vtjepa_config.yaml +All results will be saved to: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951 +[VTJEPA] Running downstream probe before training begins +Using full dataset for embedding generation: credit_card_default +Generating embedding using: JEPAEmbeddingEncoder on full dataset + 0%| | 0/59 [00:00= 15) +[VTJEPA] Early stopping triggered at epoch 36 + Best downstream validation score: 0.809943 at epoch 5 + No improvement for 15 probe evaluations +Restored best downstream-selected checkpoint from epoch 5. +Training completed. ++ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -W ignore -m src.downstream_pipeline --exp_path /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951 --num_runs 5 --epochs 50 --models mlp,dcnv2,resnet,autoint,fttransformer --no-filter_uncertainty +[Config] Loaded sigma_aggregation=mean from /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/vtjepa_config.yaml +[Config] Loaded training data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets from /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/vtjepa_config.yaml +Split statistics: + val: 2957 samples + test: 5956 samples + train: 21087 samples +Running downstream for dataset 'credit_card_default', model 'mlp'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp7lkoj9si/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=0 --np_seed=0 --output_file=/tmp/tmp7lkoj9si/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp7lkoj9si/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp7lkoj9si/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤═════════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪═════════════════════╡ +│ emb_dim │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dataset_name │ credit_card_default │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧═════════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─Linear: 2-2 [128, 23] [128, 128] 3,072 -- 393,216 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +===================================================================================================================================================================== +Total params: 53,848 +Trainable params: 53,848 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 6.89 +===================================================================================================================================================================== +Input size (MB): 0.75 +Forward/backward pass size (MB): 0.42 +Params size (MB): 0.22 +Estimated Total Size (MB): 1.39 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 53.7 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 53.8 K +Non-trainable params: 0 +Total params: 53.8 K +Total estimated model params size (MB): 0.215 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 33/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:00 • 0:00:00 244.78it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8123098015785217 │ +│ credit_card_default_val_f1-score │ 0.7886457443237305 │ +│ credit_card_default_val_loss │ 0.456789493560791 │ +│ credit_card_default_val_precision │ 0.7928820848464966 │ +│ credit_card_default_val_recall │ 0.8123098015785217 │ +│ credit_card_default_val_roc_auc │ 0.73887699842453 │ +│ credit_card_default_val_score │ 0.8123098015785217 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 0.0010000000474974513 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 543.98it/s +Validation metrics: [{'credit_card_default_val_loss': 0.456789493560791, 'lr': 0.0010000000474974513, 'credit_card_default_val_accuracy': 0.8123098015785217, 'credit_card_default_val_precision': 0.7928820848464966, 'credit_card_default_val_recall': 0.8123098015785217, 'credit_card_default_val_f1-score': 0.7886457443237305, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.73887699842453, 'credit_card_default_val_score': 0.8123098015785217}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 765 samples (12.84%) + Class 1: 5191 samples (87.16%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8183344602584839 │ +│ credit_card_default_test_f1-score │ 0.7986555695533752 │ +│ credit_card_default_test_loss │ 0.45091870427131653 │ +│ credit_card_default_test_precision │ 0.8021538257598877 │ +│ credit_card_default_test_recall │ 0.8183344602584839 │ +│ credit_card_default_test_roc_auc │ 0.7522221207618713 │ +│ credit_card_default_test_score │ 0.8183344602584839 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 0.0009999994654208422 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 472.20it/s +Test metrics: [{'credit_card_default_test_loss': 0.45091870427131653, 'lr': 0.0009999994654208422, 'credit_card_default_test_accuracy': 0.8183344602584839, 'credit_card_default_test_precision': 0.8021538257598877, 'credit_card_default_test_recall': 0.8183344602584839, 'credit_card_default_test_f1-score': 0.7986555695533752, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.7522221207618713, 'credit_card_default_test_score': 0.8183344602584839}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp7lkoj9si/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=1 --np_seed=1 --output_file=/tmp/tmp7lkoj9si/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp7lkoj9si/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp7lkoj9si/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤═════════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪═════════════════════╡ +│ emb_dim │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dataset_name │ credit_card_default │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧═════════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─Linear: 2-2 [128, 23] [128, 128] 3,072 -- 393,216 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +===================================================================================================================================================================== +Total params: 53,848 +Trainable params: 53,848 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 6.89 +===================================================================================================================================================================== +Input size (MB): 0.75 +Forward/backward pass size (MB): 0.42 +Params size (MB): 0.22 +Estimated Total Size (MB): 1.39 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 53.7 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 53.8 K +Non-trainable params: 0 +Total params: 53.8 K +Total estimated model params size (MB): 0.215 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 27/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:00 • 0:00:00 266.24it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8099424839019775 │ +│ credit_card_default_val_f1-score │ 0.773453950881958 │ +│ credit_card_default_val_loss │ 0.4575861692428589 │ +│ credit_card_default_val_precision │ 0.7934061884880066 │ +│ credit_card_default_val_recall │ 0.8099424839019775 │ +│ credit_card_default_val_roc_auc │ 0.7348690629005432 │ +│ credit_card_default_val_score │ 0.8099424839019775 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 0.0010000000474974513 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 514.10it/s +Validation metrics: [{'credit_card_default_val_loss': 0.4575861692428589, 'lr': 0.0010000000474974513, 'credit_card_default_val_accuracy': 0.8099424839019775, 'credit_card_default_val_precision': 0.7934061884880066, 'credit_card_default_val_recall': 0.8099424839019775, 'credit_card_default_val_f1-score': 0.773453950881958, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7348690629005432, 'credit_card_default_val_score': 0.8099424839019775}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 554 samples (9.30%) + Class 1: 5402 samples (90.70%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8107790350914001 │ +│ credit_card_default_test_f1-score │ 0.7787691950798035 │ +│ credit_card_default_test_loss │ 0.45597708225250244 │ +│ credit_card_default_test_precision │ 0.7945290803909302 │ +│ credit_card_default_test_recall │ 0.8107790350914001 │ +│ credit_card_default_test_roc_auc │ 0.7376885414123535 │ +│ credit_card_default_test_score │ 0.8107790350914001 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 0.0009999994654208422 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 517.08it/s +Test metrics: [{'credit_card_default_test_loss': 0.45597708225250244, 'lr': 0.0009999994654208422, 'credit_card_default_test_accuracy': 0.8107790350914001, 'credit_card_default_test_precision': 0.7945290803909302, 'credit_card_default_test_recall': 0.8107790350914001, 'credit_card_default_test_f1-score': 0.7787691950798035, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.7376885414123535, 'credit_card_default_test_score': 0.8107790350914001}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp7lkoj9si/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=2 --np_seed=2 --output_file=/tmp/tmp7lkoj9si/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp7lkoj9si/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp7lkoj9si/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤═════════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪═════════════════════╡ +│ emb_dim │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dataset_name │ credit_card_default │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧═════════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─Linear: 2-2 [128, 23] [128, 128] 3,072 -- 393,216 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +===================================================================================================================================================================== +Total params: 53,848 +Trainable params: 53,848 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 6.89 +===================================================================================================================================================================== +Input size (MB): 0.75 +Forward/backward pass size (MB): 0.42 +Params size (MB): 0.22 +Estimated Total Size (MB): 1.39 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 53.7 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 53.8 K +Non-trainable params: 0 +Total params: 53.8 K +Total estimated model params size (MB): 0.215 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 23/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:00 • 0:00:00 262.04it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8133243322372437 │ +│ credit_card_default_val_f1-score │ 0.7892600297927856 │ +│ credit_card_default_val_loss │ 0.4570556879043579 │ +│ credit_card_default_val_precision │ 0.7943274974822998 │ +│ credit_card_default_val_recall │ 0.8133243322372437 │ +│ credit_card_default_val_roc_auc │ 0.7351929545402527 │ +│ credit_card_default_val_score │ 0.8133243322372437 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 0.0010000000474974513 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 497.01it/s +Validation metrics: [{'credit_card_default_val_loss': 0.4570556879043579, 'lr': 0.0010000000474974513, 'credit_card_default_val_accuracy': 0.8133243322372437, 'credit_card_default_val_precision': 0.7943274974822998, 'credit_card_default_val_recall': 0.8133243322372437, 'credit_card_default_val_f1-score': 0.7892600297927856, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7351929545402527, 'credit_card_default_val_score': 0.8133243322372437}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 761 samples (12.78%) + Class 1: 5195 samples (87.22%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8176628351211548 │ +│ credit_card_default_test_f1-score │ 0.7977256178855896 │ +│ credit_card_default_test_loss │ 0.4498947262763977 │ +│ credit_card_default_test_precision │ 0.8012663125991821 │ +│ credit_card_default_test_recall │ 0.8176628351211548 │ +│ credit_card_default_test_roc_auc │ 0.7465294599533081 │ +│ credit_card_default_test_score │ 0.8176628351211548 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 0.0009999994654208422 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 499.84it/s +Test metrics: [{'credit_card_default_test_loss': 0.4498947262763977, 'lr': 0.0009999994654208422, 'credit_card_default_test_accuracy': 0.8176628351211548, 'credit_card_default_test_precision': 0.8012663125991821, 'credit_card_default_test_recall': 0.8176628351211548, 'credit_card_default_test_f1-score': 0.7977256178855896, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.7465294599533081, 'credit_card_default_test_score': 0.8176628351211548}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp7lkoj9si/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=3 --np_seed=3 --output_file=/tmp/tmp7lkoj9si/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp7lkoj9si/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp7lkoj9si/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤═════════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪═════════════════════╡ +│ emb_dim │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dataset_name │ credit_card_default │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧═════════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─Linear: 2-2 [128, 23] [128, 128] 3,072 -- 393,216 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +===================================================================================================================================================================== +Total params: 53,848 +Trainable params: 53,848 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 6.89 +===================================================================================================================================================================== +Input size (MB): 0.75 +Forward/backward pass size (MB): 0.42 +Params size (MB): 0.22 +Estimated Total Size (MB): 1.39 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 53.7 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 53.8 K +Non-trainable params: 0 +Total params: 53.8 K +Total estimated model params size (MB): 0.215 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 27/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:00 • 0:00:00 205.91it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8089279532432556 │ +│ credit_card_default_val_f1-score │ 0.7692520022392273 │ +│ credit_card_default_val_loss │ 0.4659510552883148 │ +│ credit_card_default_val_precision │ 0.7939008474349976 │ +│ credit_card_default_val_recall │ 0.8089279532432556 │ +│ credit_card_default_val_roc_auc │ 0.7287012338638306 │ +│ credit_card_default_val_score │ 0.8089279532432556 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 0.0010000000474974513 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 404.11it/s +Validation metrics: [{'credit_card_default_val_loss': 0.4659510552883148, 'lr': 0.0010000000474974513, 'credit_card_default_val_accuracy': 0.8089279532432556, 'credit_card_default_val_precision': 0.7939008474349976, 'credit_card_default_val_recall': 0.8089279532432556, 'credit_card_default_val_f1-score': 0.7692520022392273, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7287012338638306, 'credit_card_default_val_score': 0.8089279532432556}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 517 samples (8.68%) + Class 1: 5439 samples (91.32%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.809603750705719 │ +│ credit_card_default_test_f1-score │ 0.7750416398048401 │ +│ credit_card_default_test_loss │ 0.46061089634895325 │ +│ credit_card_default_test_precision │ 0.7940508723258972 │ +│ credit_card_default_test_recall │ 0.809603750705719 │ +│ credit_card_default_test_roc_auc │ 0.7370550632476807 │ +│ credit_card_default_test_score │ 0.809603750705719 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 0.0009999994654208422 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 312.04it/s +Test metrics: [{'credit_card_default_test_loss': 0.46061089634895325, 'lr': 0.0009999994654208422, 'credit_card_default_test_accuracy': 0.809603750705719, 'credit_card_default_test_precision': 0.7940508723258972, 'credit_card_default_test_recall': 0.809603750705719, 'credit_card_default_test_f1-score': 0.7750416398048401, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.7370550632476807, 'credit_card_default_test_score': 0.809603750705719}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp7lkoj9si/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=4 --np_seed=4 --output_file=/tmp/tmp7lkoj9si/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp7lkoj9si/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp7lkoj9si/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤═════════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪═════════════════════╡ +│ emb_dim │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dataset_name │ credit_card_default │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧═════════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─Linear: 2-2 [128, 23] [128, 128] 3,072 -- 393,216 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +===================================================================================================================================================================== +Total params: 53,848 +Trainable params: 53,848 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 6.89 +===================================================================================================================================================================== +Input size (MB): 0.75 +Forward/backward pass size (MB): 0.42 +Params size (MB): 0.22 +Estimated Total Size (MB): 1.39 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 53.7 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 53.8 K +Non-trainable params: 0 +Total params: 53.8 K +Total estimated model params size (MB): 0.215 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 40/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:00 • 0:00:00 220.56it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8116334080696106 │ +│ credit_card_default_val_f1-score │ 0.774224579334259 │ +│ credit_card_default_val_loss │ 0.47154441475868225 │ +│ credit_card_default_val_precision │ 0.7974916100502014 │ +│ credit_card_default_val_recall │ 0.8116334080696106 │ +│ credit_card_default_val_roc_auc │ 0.7265719771385193 │ +│ credit_card_default_val_score │ 0.8116334080696106 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 0.0010000000474974513 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 504.81it/s +Validation metrics: [{'credit_card_default_val_loss': 0.47154441475868225, 'lr': 0.0010000000474974513, 'credit_card_default_val_accuracy': 0.8116334080696106, 'credit_card_default_val_precision': 0.7974916100502014, 'credit_card_default_val_recall': 0.8116334080696106, 'credit_card_default_val_f1-score': 0.774224579334259, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7265719771385193, 'credit_card_default_val_score': 0.8116334080696106}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 518 samples (8.70%) + Class 1: 5438 samples (91.30%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8101074695587158 │ +│ credit_card_default_test_f1-score │ 0.775701642036438 │ +│ credit_card_default_test_loss │ 0.462413489818573 │ +│ credit_card_default_test_precision │ 0.7948746085166931 │ +│ credit_card_default_test_recall │ 0.8101074695587158 │ +│ credit_card_default_test_roc_auc │ 0.74432772397995 │ +│ credit_card_default_test_score │ 0.8101074695587158 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 0.0009999994654208422 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 500.20it/s +Test metrics: [{'credit_card_default_test_loss': 0.462413489818573, 'lr': 0.0009999994654208422, 'credit_card_default_test_accuracy': 0.8101074695587158, 'credit_card_default_test_precision': 0.7948746085166931, 'credit_card_default_test_recall': 0.8101074695587158, 'credit_card_default_test_f1-score': 0.775701642036438, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.74432772397995, 'credit_card_default_test_score': 0.8101074695587158}] +Wrote results for mlp on credit_card_default to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/downstream_results/mlp_credit_card_default_downstream_results.json +Running downstream for dataset 'credit_card_default', model 'dcnv2'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpbp_c49mn/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpbp_c49mn/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpbp_c49mn/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpbp_c49mn/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 23 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤═════════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪═════════════════════╡ +│ input_dim │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ T_max │ 8250 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dataset_name │ credit_card_default │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧═════════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─EncoderDCNv2: 2-2 [128, 23] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 23] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 23] [128, 128] 3,072 -- 393,216 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 152,664 +Trainable params: 152,664 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 19.54 +========================================================================================================================================================================== +Input size (MB): 0.75 +Forward/backward pass size (MB): 1.07 +Params size (MB): 0.61 +Estimated Total Size (MB): 2.44 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 152 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 152 K +Non-trainable params: 0 +Total params: 152 K +Total estimated model params size (MB): 0.611 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 43/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:01 • 0:00:00 139.92it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8092661499977112 │ +│ credit_card_default_val_f1-score │ 0.7855390310287476 │ +│ credit_card_default_val_loss │ 0.45164215564727783 │ +│ credit_card_default_val_precision │ 0.7886391878128052 │ +│ credit_card_default_val_recall │ 0.8092661499977112 │ +│ credit_card_default_val_roc_auc │ 0.7485185265541077 │ +│ credit_card_default_val_score │ 0.8092661499977112 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 3.511174872983247e-05 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 391.09it/s +Validation metrics: [{'credit_card_default_val_loss': 0.45164215564727783, 'lr': 3.511174872983247e-05, 'credit_card_default_val_accuracy': 0.8092661499977112, 'credit_card_default_val_precision': 0.7886391878128052, 'credit_card_default_val_recall': 0.8092661499977112, 'credit_card_default_val_f1-score': 0.7855390310287476, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7485185265541077, 'credit_card_default_val_score': 0.8092661499977112}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 767 samples (12.88%) + Class 1: 5189 samples (87.12%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8136333227157593 │ +│ credit_card_default_test_f1-score │ 0.7935397028923035 │ +│ credit_card_default_test_loss │ 0.4460753798484802 │ +│ credit_card_default_test_precision │ 0.7959308624267578 │ +│ credit_card_default_test_recall │ 0.8136333227157593 │ +│ credit_card_default_test_roc_auc │ 0.7620080709457397 │ +│ credit_card_default_test_score │ 0.8136333227157593 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 3.51117632817477e-05 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 378.07it/s +Test metrics: [{'credit_card_default_test_loss': 0.4460753798484802, 'lr': 3.51117632817477e-05, 'credit_card_default_test_accuracy': 0.8136333227157593, 'credit_card_default_test_precision': 0.7959308624267578, 'credit_card_default_test_recall': 0.8136333227157593, 'credit_card_default_test_f1-score': 0.7935397028923035, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.7620080709457397, 'credit_card_default_test_score': 0.8136333227157593}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpbp_c49mn/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpbp_c49mn/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpbp_c49mn/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpbp_c49mn/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 23 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤═════════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪═════════════════════╡ +│ input_dim │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ T_max │ 8250 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dataset_name │ credit_card_default │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧═════════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─EncoderDCNv2: 2-2 [128, 23] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 23] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 23] [128, 128] 3,072 -- 393,216 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 152,664 +Trainable params: 152,664 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 19.54 +========================================================================================================================================================================== +Input size (MB): 0.75 +Forward/backward pass size (MB): 1.07 +Params size (MB): 0.61 +Estimated Total Size (MB): 2.44 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 152 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 152 K +Non-trainable params: 0 +Total params: 152 K +Total estimated model params size (MB): 0.611 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 46/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:00 • 0:00:00 187.71it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8116334080696106 │ +│ credit_card_default_val_f1-score │ 0.787671685218811 │ +│ credit_card_default_val_loss │ 0.45115557312965393 │ +│ credit_card_default_val_precision │ 0.7919310331344604 │ +│ credit_card_default_val_recall │ 0.8116334080696106 │ +│ credit_card_default_val_roc_auc │ 0.7514497637748718 │ +│ credit_card_default_val_score │ 0.8116334080696106 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 8.856372005539015e-06 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 393.10it/s +Validation metrics: [{'credit_card_default_val_loss': 0.45115557312965393, 'lr': 8.856372005539015e-06, 'credit_card_default_val_accuracy': 0.8116334080696106, 'credit_card_default_val_precision': 0.7919310331344604, 'credit_card_default_val_recall': 0.8116334080696106, 'credit_card_default_val_f1-score': 0.787671685218811, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7514497637748718, 'credit_card_default_val_score': 0.8116334080696106}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 736 samples (12.36%) + Class 1: 5220 samples (87.64%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8158159852027893 │ +│ credit_card_default_test_f1-score │ 0.7944850921630859 │ +│ credit_card_default_test_loss │ 0.4473575949668884 │ +│ credit_card_default_test_precision │ 0.7988327741622925 │ +│ credit_card_default_test_recall │ 0.8158159852027893 │ +│ credit_card_default_test_roc_auc │ 0.7607871294021606 │ +│ credit_card_default_test_score │ 0.8158159852027893 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 8.856375643517822e-06 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 387.73it/s +Test metrics: [{'credit_card_default_test_loss': 0.4473575949668884, 'lr': 8.856375643517822e-06, 'credit_card_default_test_accuracy': 0.8158159852027893, 'credit_card_default_test_precision': 0.7988327741622925, 'credit_card_default_test_recall': 0.8158159852027893, 'credit_card_default_test_f1-score': 0.7944850921630859, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.7607871294021606, 'credit_card_default_test_score': 0.8158159852027893}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpbp_c49mn/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpbp_c49mn/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpbp_c49mn/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpbp_c49mn/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 23 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤═════════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪═════════════════════╡ +│ input_dim │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ T_max │ 8250 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dataset_name │ credit_card_default │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧═════════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─EncoderDCNv2: 2-2 [128, 23] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 23] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 23] [128, 128] 3,072 -- 393,216 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 152,664 +Trainable params: 152,664 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 19.54 +========================================================================================================================================================================== +Input size (MB): 0.75 +Forward/backward pass size (MB): 1.07 +Params size (MB): 0.61 +Estimated Total Size (MB): 2.44 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 152 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 152 K +Non-trainable params: 0 +Total params: 152 K +Total estimated model params size (MB): 0.611 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 47/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:01 • 0:00:00 144.35it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8109570741653442 │ +│ credit_card_default_val_f1-score │ 0.7879649996757507 │ +│ credit_card_default_val_loss │ 0.4507676362991333 │ +│ credit_card_default_val_precision │ 0.7910326719284058 │ +│ credit_card_default_val_recall │ 0.8109570741653442 │ +│ credit_card_default_val_roc_auc │ 0.7503932118415833 │ +│ credit_card_default_val_score │ 0.8109570741653442 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 3.942649982491275e-06 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 235.50it/s +Validation metrics: [{'credit_card_default_val_loss': 0.4507676362991333, 'lr': 3.942649982491275e-06, 'credit_card_default_val_accuracy': 0.8109570741653442, 'credit_card_default_val_precision': 0.7910326719284058, 'credit_card_default_val_recall': 0.8109570741653442, 'credit_card_default_val_f1-score': 0.7879649996757507, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7503932118415833, 'credit_card_default_val_score': 0.8109570741653442}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 764 samples (12.83%) + Class 1: 5192 samples (87.17%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8144727945327759 │ +│ credit_card_default_test_f1-score │ 0.794328510761261 │ +│ credit_card_default_test_loss │ 0.44750547409057617 │ +│ credit_card_default_test_precision │ 0.7970324158668518 │ +│ credit_card_default_test_recall │ 0.8144727945327759 │ +│ credit_card_default_test_roc_auc │ 0.7594409584999084 │ +│ credit_card_default_test_score │ 0.8144727945327759 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 3.942650891985977e-06 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 203.53it/s +Test metrics: [{'credit_card_default_test_loss': 0.44750547409057617, 'lr': 3.942650891985977e-06, 'credit_card_default_test_accuracy': 0.8144727945327759, 'credit_card_default_test_precision': 0.7970324158668518, 'credit_card_default_test_recall': 0.8144727945327759, 'credit_card_default_test_f1-score': 0.794328510761261, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.7594409584999084, 'credit_card_default_test_score': 0.8144727945327759}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpbp_c49mn/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpbp_c49mn/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpbp_c49mn/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpbp_c49mn/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 23 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤═════════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪═════════════════════╡ +│ input_dim │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ T_max │ 8250 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dataset_name │ credit_card_default │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧═════════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─EncoderDCNv2: 2-2 [128, 23] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 23] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 23] [128, 128] 3,072 -- 393,216 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 152,664 +Trainable params: 152,664 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 19.54 +========================================================================================================================================================================== +Input size (MB): 0.75 +Forward/backward pass size (MB): 1.07 +Params size (MB): 0.61 +Estimated Total Size (MB): 2.44 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 152 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 152 K +Non-trainable params: 0 +Total params: 152 K +Total estimated model params size (MB): 0.611 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:01 • 0:00:00 151.15it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8106188774108887 │ +│ credit_card_default_val_f1-score │ 0.7881050705909729 │ +│ credit_card_default_val_loss │ 0.4513669013977051 │ +│ credit_card_default_val_precision │ 0.7906190752983093 │ +│ credit_card_default_val_recall │ 0.8106188774108887 │ +│ credit_card_default_val_roc_auc │ 0.7520156502723694 │ +│ credit_card_default_val_score │ 0.8106188774108887 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 0.0 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 237.95it/s +Validation metrics: [{'credit_card_default_val_loss': 0.4513669013977051, 'lr': 0.0, 'credit_card_default_val_accuracy': 0.8106188774108887, 'credit_card_default_val_precision': 0.7906190752983093, 'credit_card_default_val_recall': 0.8106188774108887, 'credit_card_default_val_f1-score': 0.7881050705909729, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7520156502723694, 'credit_card_default_val_score': 0.8106188774108887}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 775 samples (13.01%) + Class 1: 5181 samples (86.99%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8126259446144104 │ +│ credit_card_default_test_f1-score │ 0.7928016185760498 │ +│ credit_card_default_test_loss │ 0.44810912013053894 │ +│ credit_card_default_test_precision │ 0.7946444153785706 │ +│ credit_card_default_test_recall │ 0.8126259446144104 │ +│ credit_card_default_test_roc_auc │ 0.7593835592269897 │ +│ credit_card_default_test_score │ 0.8126259446144104 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 0.0 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 260.06it/s +Test metrics: [{'credit_card_default_test_loss': 0.44810912013053894, 'lr': 0.0, 'credit_card_default_test_accuracy': 0.8126259446144104, 'credit_card_default_test_precision': 0.7946444153785706, 'credit_card_default_test_recall': 0.8126259446144104, 'credit_card_default_test_f1-score': 0.7928016185760498, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.7593835592269897, 'credit_card_default_test_score': 0.8126259446144104}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpbp_c49mn/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpbp_c49mn/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpbp_c49mn/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpbp_c49mn/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 23 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤═════════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪═════════════════════╡ +│ input_dim │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ T_max │ 8250 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dataset_name │ credit_card_default │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧═════════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─EncoderDCNv2: 2-2 [128, 23] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 23] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 23] [128, 128] 3,072 -- 393,216 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 152,664 +Trainable params: 152,664 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 19.54 +========================================================================================================================================================================== +Input size (MB): 0.75 +Forward/backward pass size (MB): 1.07 +Params size (MB): 0.61 +Estimated Total Size (MB): 2.44 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 152 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 152 K +Non-trainable params: 0 +Total params: 152 K +Total estimated model params size (MB): 0.611 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 48/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:00 • 0:00:00 181.28it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8129861354827881 │ +│ credit_card_default_val_f1-score │ 0.7889845967292786 │ +│ credit_card_default_val_loss │ 0.45220452547073364 │ +│ credit_card_default_val_precision │ 0.7938436269760132 │ +│ credit_card_default_val_recall │ 0.8129861354827881 │ +│ credit_card_default_val_roc_auc │ 0.7494129538536072 │ +│ credit_card_default_val_score │ 0.8129861354827881 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 9.86635996014229e-07 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 399.11it/s +Validation metrics: [{'credit_card_default_val_loss': 0.45220452547073364, 'lr': 9.86635996014229e-07, 'credit_card_default_val_accuracy': 0.8129861354827881, 'credit_card_default_val_precision': 0.7938436269760132, 'credit_card_default_val_recall': 0.8129861354827881, 'credit_card_default_val_f1-score': 0.7889845967292786, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7494129538536072, 'credit_card_default_val_score': 0.8129861354827881}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 754 samples (12.66%) + Class 1: 5202 samples (87.34%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8158159852027893 │ +│ credit_card_default_test_f1-score │ 0.7953466176986694 │ +│ credit_card_default_test_loss │ 0.44776108860969543 │ +│ credit_card_default_test_precision │ 0.7988069653511047 │ +│ credit_card_default_test_recall │ 0.8158159852027893 │ +│ credit_card_default_test_roc_auc │ 0.7596734166145325 │ +│ credit_card_default_test_score │ 0.8158159852027893 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 9.866355412668781e-07 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 396.01it/s +Test metrics: [{'credit_card_default_test_loss': 0.44776108860969543, 'lr': 9.866355412668781e-07, 'credit_card_default_test_accuracy': 0.8158159852027893, 'credit_card_default_test_precision': 0.7988069653511047, 'credit_card_default_test_recall': 0.8158159852027893, 'credit_card_default_test_f1-score': 0.7953466176986694, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.7596734166145325, 'credit_card_default_test_score': 0.8158159852027893}] +Wrote results for dcnv2 on credit_card_default to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/downstream_results/dcnv2_credit_card_default_downstream_results.json +Running downstream for dataset 'credit_card_default', model 'resnet'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmphon_xfhk/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=0 --np_seed=0 --output_file=/tmp/tmphon_xfhk/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmphon_xfhk/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmphon_xfhk/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 23 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─ResNet: 2-2 [128, 23] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 23] [128, 128] 3,072 -- 393,216 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 252,376 +Trainable params: 252,376 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 32.30 +========================================================================================================================================================================== +Input size (MB): 0.75 +Forward/backward pass size (MB): 1.99 +Params size (MB): 1.01 +Estimated Total Size (MB): 3.75 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 252 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 252 K +Non-trainable params: 0 +Total params: 252 K +Total estimated model params size (MB): 1.010 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 39/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:00 • 0:00:00 171.26it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8068988919258118 │ +│ credit_card_default_val_f1-score │ 0.7888129949569702 │ +│ credit_card_default_val_loss │ 0.45115408301353455 │ +│ credit_card_default_val_precision │ 0.7871233224868774 │ +│ credit_card_default_val_recall │ 0.8068988919258118 │ +│ credit_card_default_val_roc_auc │ 0.7570883631706238 │ +│ credit_card_default_val_score │ 0.8068988919258118 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 0.0010000000474974513 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 415.62it/s +Validation metrics: [{'credit_card_default_val_loss': 0.45115408301353455, 'lr': 0.0010000000474974513, 'credit_card_default_val_accuracy': 0.8068988919258118, 'credit_card_default_val_precision': 0.7871233224868774, 'credit_card_default_val_recall': 0.8068988919258118, 'credit_card_default_val_f1-score': 0.7888129949569702, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7570883631706238, 'credit_card_default_val_score': 0.8068988919258118}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 858 samples (14.41%) + Class 1: 5098 samples (85.59%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8171591758728027 │ +│ credit_card_default_test_f1-score │ 0.8014456629753113 │ +│ credit_card_default_test_loss │ 0.443339079618454 │ +│ credit_card_default_test_precision │ 0.8012056350708008 │ +│ credit_card_default_test_recall │ 0.8171591758728027 │ +│ credit_card_default_test_roc_auc │ 0.7628201842308044 │ +│ credit_card_default_test_score │ 0.8171591758728027 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 0.0009999994654208422 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 411.51it/s +Test metrics: [{'credit_card_default_test_loss': 0.443339079618454, 'lr': 0.0009999994654208422, 'credit_card_default_test_accuracy': 0.8171591758728027, 'credit_card_default_test_precision': 0.8012056350708008, 'credit_card_default_test_recall': 0.8171591758728027, 'credit_card_default_test_f1-score': 0.8014456629753113, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.7628201842308044, 'credit_card_default_test_score': 0.8171591758728027}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmphon_xfhk/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=1 --np_seed=1 --output_file=/tmp/tmphon_xfhk/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmphon_xfhk/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmphon_xfhk/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 23 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─ResNet: 2-2 [128, 23] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 23] [128, 128] 3,072 -- 393,216 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 252,376 +Trainable params: 252,376 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 32.30 +========================================================================================================================================================================== +Input size (MB): 0.75 +Forward/backward pass size (MB): 1.99 +Params size (MB): 1.01 +Estimated Total Size (MB): 3.75 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 252 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 252 K +Non-trainable params: 0 +Total params: 252 K +Total estimated model params size (MB): 1.010 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 32/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:01 • 0:00:00 161.38it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8048697710037231 │ +│ credit_card_default_val_f1-score │ 0.7854407429695129 │ +│ credit_card_default_val_loss │ 0.4531085193157196 │ +│ credit_card_default_val_precision │ 0.7840221524238586 │ +│ credit_card_default_val_recall │ 0.8048697710037231 │ +│ credit_card_default_val_roc_auc │ 0.7518139481544495 │ +│ credit_card_default_val_score │ 0.8048697710037231 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 0.0005000000237487257 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 298.19it/s +Validation metrics: [{'credit_card_default_val_loss': 0.4531085193157196, 'lr': 0.0005000000237487257, 'credit_card_default_val_accuracy': 0.8048697710037231, 'credit_card_default_val_precision': 0.7840221524238586, 'credit_card_default_val_recall': 0.8048697710037231, 'credit_card_default_val_f1-score': 0.7854407429695129, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7518139481544495, 'credit_card_default_val_score': 0.8048697710037231}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 871 samples (14.62%) + Class 1: 5085 samples (85.38%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8166554570198059 │ +│ credit_card_default_test_f1-score │ 0.8014382123947144 │ +│ credit_card_default_test_loss │ 0.4463523030281067 │ +│ credit_card_default_test_precision │ 0.8007574677467346 │ +│ credit_card_default_test_recall │ 0.8166554570198059 │ +│ credit_card_default_test_roc_auc │ 0.7634868025779724 │ +│ credit_card_default_test_score │ 0.8166554570198059 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 0.0004999997327104211 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 287.56it/s +Test metrics: [{'credit_card_default_test_loss': 0.4463523030281067, 'lr': 0.0004999997327104211, 'credit_card_default_test_accuracy': 0.8166554570198059, 'credit_card_default_test_precision': 0.8007574677467346, 'credit_card_default_test_recall': 0.8166554570198059, 'credit_card_default_test_f1-score': 0.8014382123947144, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.7634868025779724, 'credit_card_default_test_score': 0.8166554570198059}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmphon_xfhk/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=2 --np_seed=2 --output_file=/tmp/tmphon_xfhk/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmphon_xfhk/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmphon_xfhk/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 23 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─ResNet: 2-2 [128, 23] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 23] [128, 128] 3,072 -- 393,216 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 252,376 +Trainable params: 252,376 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 32.30 +========================================================================================================================================================================== +Input size (MB): 0.75 +Forward/backward pass size (MB): 1.99 +Params size (MB): 1.01 +Estimated Total Size (MB): 3.75 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 252 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 252 K +Non-trainable params: 0 +Total params: 252 K +Total estimated model params size (MB): 1.010 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 32/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:00 • 0:00:00 167.65it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8109570741653442 │ +│ credit_card_default_val_f1-score │ 0.7891989350318909 │ +│ credit_card_default_val_loss │ 0.4580335021018982 │ +│ credit_card_default_val_precision │ 0.791175901889801 │ +│ credit_card_default_val_recall │ 0.8109570741653442 │ +│ credit_card_default_val_roc_auc │ 0.7497589588165283 │ +│ credit_card_default_val_score │ 0.8109570741653442 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 0.0005000000237487257 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 414.46it/s +Validation metrics: [{'credit_card_default_val_loss': 0.4580335021018982, 'lr': 0.0005000000237487257, 'credit_card_default_val_accuracy': 0.8109570741653442, 'credit_card_default_val_precision': 0.791175901889801, 'credit_card_default_val_recall': 0.8109570741653442, 'credit_card_default_val_f1-score': 0.7891989350318909, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7497589588165283, 'credit_card_default_val_score': 0.8109570741653442}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 818 samples (13.73%) + Class 1: 5138 samples (86.27%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8174949884414673 │ +│ credit_card_default_test_f1-score │ 0.8001068234443665 │ +│ credit_card_default_test_loss │ 0.44910746812820435 │ +│ credit_card_default_test_precision │ 0.8012334108352661 │ +│ credit_card_default_test_recall │ 0.8174949884414673 │ +│ credit_card_default_test_roc_auc │ 0.7649736404418945 │ +│ credit_card_default_test_score │ 0.8174949884414673 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 0.0004999997327104211 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 417.30it/s +Test metrics: [{'credit_card_default_test_loss': 0.44910746812820435, 'lr': 0.0004999997327104211, 'credit_card_default_test_accuracy': 0.8174949884414673, 'credit_card_default_test_precision': 0.8012334108352661, 'credit_card_default_test_recall': 0.8174949884414673, 'credit_card_default_test_f1-score': 0.8001068234443665, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.7649736404418945, 'credit_card_default_test_score': 0.8174949884414673}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmphon_xfhk/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=3 --np_seed=3 --output_file=/tmp/tmphon_xfhk/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmphon_xfhk/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmphon_xfhk/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 23 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─ResNet: 2-2 [128, 23] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 23] [128, 128] 3,072 -- 393,216 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 252,376 +Trainable params: 252,376 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 32.30 +========================================================================================================================================================================== +Input size (MB): 0.75 +Forward/backward pass size (MB): 1.99 +Params size (MB): 1.01 +Estimated Total Size (MB): 3.75 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 252 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 252 K +Non-trainable params: 0 +Total params: 252 K +Total estimated model params size (MB): 1.010 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 31/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:00 • 0:00:00 187.01it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8116334080696106 │ +│ credit_card_default_val_f1-score │ 0.7891373634338379 │ +│ credit_card_default_val_loss │ 0.45125502347946167 │ +│ credit_card_default_val_precision │ 0.791999340057373 │ +│ credit_card_default_val_recall │ 0.8116334080696106 │ +│ credit_card_default_val_roc_auc │ 0.7552090287208557 │ +│ credit_card_default_val_score │ 0.8116334080696106 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 0.0010000000474974513 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 446.91it/s +Validation metrics: [{'credit_card_default_val_loss': 0.45125502347946167, 'lr': 0.0010000000474974513, 'credit_card_default_val_accuracy': 0.8116334080696106, 'credit_card_default_val_precision': 0.791999340057373, 'credit_card_default_val_recall': 0.8116334080696106, 'credit_card_default_val_f1-score': 0.7891373634338379, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7552090287208557, 'credit_card_default_val_score': 0.8116334080696106}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 782 samples (13.13%) + Class 1: 5174 samples (86.87%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8174949884414673 │ +│ credit_card_default_test_f1-score │ 0.7985050678253174 │ +│ credit_card_default_test_loss │ 0.4443984627723694 │ +│ credit_card_default_test_precision │ 0.8010596632957458 │ +│ credit_card_default_test_recall │ 0.8174949884414673 │ +│ credit_card_default_test_roc_auc │ 0.764964759349823 │ +│ credit_card_default_test_score │ 0.8174949884414673 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 0.0009999994654208422 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 442.53it/s +Test metrics: [{'credit_card_default_test_loss': 0.4443984627723694, 'lr': 0.0009999994654208422, 'credit_card_default_test_accuracy': 0.8174949884414673, 'credit_card_default_test_precision': 0.8010596632957458, 'credit_card_default_test_recall': 0.8174949884414673, 'credit_card_default_test_f1-score': 0.7985050678253174, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.764964759349823, 'credit_card_default_test_score': 0.8174949884414673}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmphon_xfhk/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=4 --np_seed=4 --output_file=/tmp/tmphon_xfhk/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmphon_xfhk/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmphon_xfhk/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 23 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─ResNet: 2-2 [128, 23] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 23] [128, 128] 3,072 -- 393,216 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 252,376 +Trainable params: 252,376 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 32.30 +========================================================================================================================================================================== +Input size (MB): 0.75 +Forward/backward pass size (MB): 1.99 +Params size (MB): 1.01 +Estimated Total Size (MB): 3.75 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 252 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 252 K +Non-trainable params: 0 +Total params: 252 K +Total estimated model params size (MB): 1.010 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 29/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:00 • 0:00:00 181.61it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8099424839019775 │ +│ credit_card_default_val_f1-score │ 0.783901572227478 │ +│ credit_card_default_val_loss │ 0.45458972454071045 │ +│ credit_card_default_val_precision │ 0.789559543132782 │ +│ credit_card_default_val_recall │ 0.8099424839019775 │ +│ credit_card_default_val_roc_auc │ 0.7507177591323853 │ +│ credit_card_default_val_score │ 0.8099424839019775 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 0.0 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 423.07it/s +Validation metrics: [{'credit_card_default_val_loss': 0.45458972454071045, 'lr': 0.0, 'credit_card_default_val_accuracy': 0.8099424839019775, 'credit_card_default_val_precision': 0.789559543132782, 'credit_card_default_val_recall': 0.8099424839019775, 'credit_card_default_val_f1-score': 0.783901572227478, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7507177591323853, 'credit_card_default_val_score': 0.8099424839019775}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 702 samples (11.79%) + Class 1: 5254 samples (88.21%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8168233633041382 │ +│ credit_card_default_test_f1-score │ 0.7939406037330627 │ +│ credit_card_default_test_loss │ 0.44696947932243347 │ +│ credit_card_default_test_precision │ 0.8004374504089355 │ +│ credit_card_default_test_recall │ 0.8168233633041382 │ +│ credit_card_default_test_roc_auc │ 0.7622969746589661 │ +│ credit_card_default_test_score │ 0.8168233633041382 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 0.0 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 413.97it/s +Test metrics: [{'credit_card_default_test_loss': 0.44696947932243347, 'lr': 0.0, 'credit_card_default_test_accuracy': 0.8168233633041382, 'credit_card_default_test_precision': 0.8004374504089355, 'credit_card_default_test_recall': 0.8168233633041382, 'credit_card_default_test_f1-score': 0.7939406037330627, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.7622969746589661, 'credit_card_default_test_score': 0.8168233633041382}] +Wrote results for resnet on credit_card_default to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/downstream_results/resnet_credit_card_default_downstream_results.json +Running downstream for dataset 'credit_card_default', model 'autoint'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpz4osx5es/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpz4osx5es/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpz4osx5es/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpz4osx5es/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 23 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤═════════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪═════════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_dim │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ T_max │ 8250 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dataset_name │ credit_card_default │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧═════════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─EncoderAutoIntBase: 2-2 [128, 23] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 23] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 23] [128, 23, 128] 2,944 -- 376,832 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 23, 23] [256, 23, 23] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 23, 128] [128, 23, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 23, 128] [128, 23, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 23, 23] [256, 23, 23] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 23, 128] [128, 23, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 23, 128] [128, 23, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 23, 23] [256, 23, 23] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 23, 128] [128, 23, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 23, 128] [128, 23, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 2944] [128, 128] 376,960 -- 48,250,880 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +==================================================================================================================================================================================== +Total params: 612,440 +Trainable params: 612,440 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 78.39 +==================================================================================================================================================================================== +Input size (MB): 0.75 +Forward/backward pass size (MB): 48.39 +Params size (MB): 2.45 +Estimated Total Size (MB): 51.59 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 612 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 612 K +Non-trainable params: 0 +Total params: 612 K +Total estimated model params size (MB): 2.450 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:01 • 0:00:00 121.22it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8150151968002319 │ +│ credit_card_default_val_f1-score │ 0.7918998599052429 │ +│ credit_card_default_val_loss │ 0.44546744227409363 │ +│ credit_card_default_val_precision │ 0.7966643571853638 │ +│ credit_card_default_val_recall │ 0.8150151968002319 │ +│ credit_card_default_val_roc_auc │ 0.7653359770774841 │ +│ credit_card_default_val_score │ 0.8150151968002319 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 0.0 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 321.00it/s +Validation metrics: [{'credit_card_default_val_loss': 0.44546744227409363, 'lr': 0.0, 'credit_card_default_val_accuracy': 0.8150151968002319, 'credit_card_default_val_precision': 0.7966643571853638, 'credit_card_default_val_recall': 0.8150151968002319, 'credit_card_default_val_f1-score': 0.7918998599052429, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7653359770774841, 'credit_card_default_val_score': 0.8150151968002319}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 737 samples (12.37%) + Class 1: 5219 samples (87.63%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8183344602584839 │ +│ credit_card_default_test_f1-score │ 0.7973429560661316 │ +│ credit_card_default_test_loss │ 0.4415125846862793 │ +│ credit_card_default_test_precision │ 0.8022464513778687 │ +│ credit_card_default_test_recall │ 0.8183344602584839 │ +│ credit_card_default_test_roc_auc │ 0.7686120867729187 │ +│ credit_card_default_test_score │ 0.8183344602584839 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 0.0 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 319.83it/s +Test metrics: [{'credit_card_default_test_loss': 0.4415125846862793, 'lr': 0.0, 'credit_card_default_test_accuracy': 0.8183344602584839, 'credit_card_default_test_precision': 0.8022464513778687, 'credit_card_default_test_recall': 0.8183344602584839, 'credit_card_default_test_f1-score': 0.7973429560661316, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.7686120867729187, 'credit_card_default_test_score': 0.8183344602584839}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpz4osx5es/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpz4osx5es/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpz4osx5es/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpz4osx5es/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 23 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤═════════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪═════════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_dim │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ T_max │ 8250 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dataset_name │ credit_card_default │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧═════════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─EncoderAutoIntBase: 2-2 [128, 23] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 23] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 23] [128, 23, 128] 2,944 -- 376,832 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 23, 23] [256, 23, 23] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 23, 128] [128, 23, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 23, 128] [128, 23, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 23, 23] [256, 23, 23] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 23, 128] [128, 23, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 23, 128] [128, 23, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 23, 23] [256, 23, 23] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 23, 128] [128, 23, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 23, 128] [128, 23, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 2944] [128, 128] 376,960 -- 48,250,880 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +==================================================================================================================================================================================== +Total params: 612,440 +Trainable params: 612,440 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 78.39 +==================================================================================================================================================================================== +Input size (MB): 0.75 +Forward/backward pass size (MB): 48.39 +Params size (MB): 2.45 +Estimated Total Size (MB): 51.59 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 612 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 612 K +Non-trainable params: 0 +Total params: 612 K +Total estimated model params size (MB): 2.450 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 41/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:01 • 0:00:00 119.14it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8153533935546875 │ +│ credit_card_default_val_f1-score │ 0.7927940487861633 │ +│ credit_card_default_val_loss │ 0.4460710883140564 │ +│ credit_card_default_val_precision │ 0.7971121072769165 │ +│ credit_card_default_val_recall │ 0.8153533935546875 │ +│ credit_card_default_val_roc_auc │ 0.7558637857437134 │ +│ credit_card_default_val_score │ 0.8153533935546875 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 6.184668018249795e-05 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 317.51it/s +Validation metrics: [{'credit_card_default_val_loss': 0.4460710883140564, 'lr': 6.184668018249795e-05, 'credit_card_default_val_accuracy': 0.8153533935546875, 'credit_card_default_val_precision': 0.7971121072769165, 'credit_card_default_val_recall': 0.8153533935546875, 'credit_card_default_val_f1-score': 0.7927940487861633, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7558637857437134, 'credit_card_default_val_score': 0.8153533935546875}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 768 samples (12.89%) + Class 1: 5188 samples (87.11%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8171591758728027 │ +│ credit_card_default_test_f1-score │ 0.7974919676780701 │ +│ credit_card_default_test_loss │ 0.4429546296596527 │ +│ credit_card_default_test_precision │ 0.8005972504615784 │ +│ credit_card_default_test_recall │ 0.8171591758728027 │ +│ credit_card_default_test_roc_auc │ 0.7640913128852844 │ +│ credit_card_default_test_score │ 0.8171591758728027 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 6.184665107866749e-05 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 321.73it/s +Test metrics: [{'credit_card_default_test_loss': 0.4429546296596527, 'lr': 6.184665107866749e-05, 'credit_card_default_test_accuracy': 0.8171591758728027, 'credit_card_default_test_precision': 0.8005972504615784, 'credit_card_default_test_recall': 0.8171591758728027, 'credit_card_default_test_f1-score': 0.7974919676780701, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.7640913128852844, 'credit_card_default_test_score': 0.8171591758728027}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpz4osx5es/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpz4osx5es/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpz4osx5es/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpz4osx5es/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 23 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤═════════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪═════════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_dim │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ T_max │ 8250 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dataset_name │ credit_card_default │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧═════════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─EncoderAutoIntBase: 2-2 [128, 23] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 23] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 23] [128, 23, 128] 2,944 -- 376,832 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 23, 23] [256, 23, 23] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 23, 128] [128, 23, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 23, 128] [128, 23, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 23, 23] [256, 23, 23] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 23, 128] [128, 23, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 23, 128] [128, 23, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 23, 23] [256, 23, 23] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 23, 128] [128, 23, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 23, 128] [128, 23, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 2944] [128, 128] 376,960 -- 48,250,880 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +==================================================================================================================================================================================== +Total params: 612,440 +Trainable params: 612,440 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 78.39 +==================================================================================================================================================================================== +Input size (MB): 0.75 +Forward/backward pass size (MB): 48.39 +Params size (MB): 2.45 +Estimated Total Size (MB): 51.59 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 612 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 612 K +Non-trainable params: 0 +Total params: 612 K +Total estimated model params size (MB): 2.450 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:01 • 0:00:00 116.99it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8129861354827881 │ +│ credit_card_default_val_f1-score │ 0.7904467582702637 │ +│ credit_card_default_val_loss │ 0.44784149527549744 │ +│ credit_card_default_val_precision │ 0.7938451170921326 │ +│ credit_card_default_val_recall │ 0.8129861354827881 │ +│ credit_card_default_val_roc_auc │ 0.7518703937530518 │ +│ credit_card_default_val_score │ 0.8129861354827881 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 0.0 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 313.51it/s +Validation metrics: [{'credit_card_default_val_loss': 0.44784149527549744, 'lr': 0.0, 'credit_card_default_val_accuracy': 0.8129861354827881, 'credit_card_default_val_precision': 0.7938451170921326, 'credit_card_default_val_recall': 0.8129861354827881, 'credit_card_default_val_f1-score': 0.7904467582702637, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7518703937530518, 'credit_card_default_val_score': 0.8129861354827881}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 772 samples (12.96%) + Class 1: 5184 samples (87.04%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8171591758728027 │ +│ credit_card_default_test_f1-score │ 0.7976765632629395 │ +│ credit_card_default_test_loss │ 0.44063782691955566 │ +│ credit_card_default_test_precision │ 0.8006010055541992 │ +│ credit_card_default_test_recall │ 0.8171591758728027 │ +│ credit_card_default_test_roc_auc │ 0.767403781414032 │ +│ credit_card_default_test_score │ 0.8171591758728027 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 0.0 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 315.91it/s +Test metrics: [{'credit_card_default_test_loss': 0.44063782691955566, 'lr': 0.0, 'credit_card_default_test_accuracy': 0.8171591758728027, 'credit_card_default_test_precision': 0.8006010055541992, 'credit_card_default_test_recall': 0.8171591758728027, 'credit_card_default_test_f1-score': 0.7976765632629395, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.767403781414032, 'credit_card_default_test_score': 0.8171591758728027}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpz4osx5es/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpz4osx5es/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpz4osx5es/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpz4osx5es/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 23 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤═════════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪═════════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_dim │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ T_max │ 8250 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dataset_name │ credit_card_default │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧═════════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─EncoderAutoIntBase: 2-2 [128, 23] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 23] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 23] [128, 23, 128] 2,944 -- 376,832 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 23, 23] [256, 23, 23] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 23, 128] [128, 23, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 23, 128] [128, 23, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 23, 23] [256, 23, 23] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 23, 128] [128, 23, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 23, 128] [128, 23, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 23, 23] [256, 23, 23] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 23, 128] [128, 23, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 23, 128] [128, 23, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 2944] [128, 128] 376,960 -- 48,250,880 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +==================================================================================================================================================================================== +Total params: 612,440 +Trainable params: 612,440 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 78.39 +==================================================================================================================================================================================== +Input size (MB): 0.75 +Forward/backward pass size (MB): 48.39 +Params size (MB): 2.45 +Estimated Total Size (MB): 51.59 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 612 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 612 K +Non-trainable params: 0 +Total params: 612 K +Total estimated model params size (MB): 2.450 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:01 • 0:00:00 97.03it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8123098015785217 │ +│ credit_card_default_val_f1-score │ 0.7882214188575745 │ +│ credit_card_default_val_loss │ 0.4451896846294403 │ +│ credit_card_default_val_precision │ 0.7928875684738159 │ +│ credit_card_default_val_recall │ 0.8123098015785217 │ +│ credit_card_default_val_roc_auc │ 0.7605947256088257 │ +│ credit_card_default_val_score │ 0.8123098015785217 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 0.0 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 96.61it/s +Validation metrics: [{'credit_card_default_val_loss': 0.4451896846294403, 'lr': 0.0, 'credit_card_default_val_accuracy': 0.8123098015785217, 'credit_card_default_val_precision': 0.7928875684738159, 'credit_card_default_val_recall': 0.8123098015785217, 'credit_card_default_val_f1-score': 0.7882214188575745, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7605947256088257, 'credit_card_default_val_score': 0.8123098015785217}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 743 samples (12.47%) + Class 1: 5213 samples (87.53%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8149765133857727 │ +│ credit_card_default_test_f1-score │ 0.7938871383666992 │ +│ credit_card_default_test_loss │ 0.44132500886917114 │ +│ credit_card_default_test_precision │ 0.7976836562156677 │ +│ credit_card_default_test_recall │ 0.8149765133857727 │ +│ credit_card_default_test_roc_auc │ 0.7687815427780151 │ +│ credit_card_default_test_score │ 0.8149765133857727 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 0.0 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 264.80it/s +Test metrics: [{'credit_card_default_test_loss': 0.44132500886917114, 'lr': 0.0, 'credit_card_default_test_accuracy': 0.8149765133857727, 'credit_card_default_test_precision': 0.7976836562156677, 'credit_card_default_test_recall': 0.8149765133857727, 'credit_card_default_test_f1-score': 0.7938871383666992, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.7687815427780151, 'credit_card_default_test_score': 0.8149765133857727}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpz4osx5es/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpz4osx5es/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpz4osx5es/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpz4osx5es/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 23 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤═════════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪═════════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_dim │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ T_max │ 8250 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dataset_name │ credit_card_default │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧═════════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─EncoderAutoIntBase: 2-2 [128, 23] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 23] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 23] [128, 23, 128] 2,944 -- 376,832 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 23, 23] [256, 23, 23] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 23, 128] [128, 23, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 23, 128] [128, 23, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 23, 23] [256, 23, 23] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 23, 128] [128, 23, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 23, 128] [128, 23, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 23, 128] [128, 23, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 23, 23] [256, 23, 23] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 23, 128] [128, 23, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 23, 128] [128, 23, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 23, 128] [128, 23, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 2944] [128, 128] 376,960 -- 48,250,880 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +==================================================================================================================================================================================== +Total params: 612,440 +Trainable params: 612,440 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 78.39 +==================================================================================================================================================================================== +Input size (MB): 0.75 +Forward/backward pass size (MB): 48.39 +Params size (MB): 2.45 +Estimated Total Size (MB): 51.59 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 612 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 612 K +Non-trainable params: 0 +Total params: 612 K +Total estimated model params size (MB): 2.450 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 39/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:01 • 0:00:00 91.47it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8150151968002319 │ +│ credit_card_default_val_f1-score │ 0.7910634279251099 │ +│ credit_card_default_val_loss │ 0.4478146731853485 │ +│ credit_card_default_val_precision │ 0.7967334389686584 │ +│ credit_card_default_val_recall │ 0.8150151968002319 │ +│ credit_card_default_val_roc_auc │ 0.7499752640724182 │ +│ credit_card_default_val_score │ 0.8150151968002319 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 9.549151582177728e-05 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 138.57it/s +Validation metrics: [{'credit_card_default_val_loss': 0.4478146731853485, 'lr': 9.549151582177728e-05, 'credit_card_default_val_accuracy': 0.8150151968002319, 'credit_card_default_val_precision': 0.7967334389686584, 'credit_card_default_val_recall': 0.8150151968002319, 'credit_card_default_val_f1-score': 0.7910634279251099, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7499752640724182, 'credit_card_default_val_score': 0.8150151968002319}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 748 samples (12.56%) + Class 1: 5208 samples (87.44%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.815144419670105 │ +│ credit_card_default_test_f1-score │ 0.7943142056465149 │ +│ credit_card_default_test_loss │ 0.4436773359775543 │ +│ credit_card_default_test_precision │ 0.797907292842865 │ +│ credit_card_default_test_recall │ 0.815144419670105 │ +│ credit_card_default_test_roc_auc │ 0.7638025879859924 │ +│ credit_card_default_test_score │ 0.815144419670105 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 9.549150126986206e-05 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 139.80it/s +Test metrics: [{'credit_card_default_test_loss': 0.4436773359775543, 'lr': 9.549150126986206e-05, 'credit_card_default_test_accuracy': 0.815144419670105, 'credit_card_default_test_precision': 0.797907292842865, 'credit_card_default_test_recall': 0.815144419670105, 'credit_card_default_test_f1-score': 0.7943142056465149, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.7638025879859924, 'credit_card_default_test_score': 0.815144419670105}] +Wrote results for autoint on credit_card_default to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/downstream_results/autoint_credit_card_default_downstream_results.json +Running downstream for dataset 'credit_card_default', model 'fttransformer'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpjq29br_h/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpjq29br_h/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpjq29br_h/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpjq29br_h/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 23 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤═════════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪═════════════════════╡ +│ input_dim │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_cont_features │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ T_max │ 8250 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dataset_name │ credit_card_default │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧═════════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─EncoderFTTransformer: 2-2 [128, 23] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 23] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 23] [128, 23, 128] 5,888 -- 753,664 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 24, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 24, 24] [512, 24, 24] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 24, 128] [128, 24, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 24, 512] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 24, 256] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 24, 256] [128, 24, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 24, 24] [512, 24, 24] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 24, 128] [128, 24, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 24, 512] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 24, 256] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 24, 256] [128, 24, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 24] [512, 1, 24] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 24, 128] [128, 24, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 24, 512] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 24, 256] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 24, 256] [128, 24, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================================= +Total params: 553,048 +Trainable params: 553,048 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 70.79 +========================================================================================================================================================================================= +Input size (MB): 0.75 +Forward/backward pass size (MB): 98.07 +Params size (MB): 2.21 +Estimated Total Size (MB): 101.03 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 552 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 553 K +Non-trainable params: 0 +Total params: 553 K +Total estimated model params size (MB): 2.212 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:02 • 0:00:00 80.82it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8143388628959656 │ +│ credit_card_default_val_f1-score │ 0.791758120059967 │ +│ credit_card_default_val_loss │ 0.44726452231407166 │ +│ credit_card_default_val_precision │ 0.7957068681716919 │ +│ credit_card_default_val_recall │ 0.8143388628959656 │ +│ credit_card_default_val_roc_auc │ 0.7507679462432861 │ +│ credit_card_default_val_score │ 0.8143388628959656 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 0.0 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 208.67it/s +Validation metrics: [{'credit_card_default_val_loss': 0.44726452231407166, 'lr': 0.0, 'credit_card_default_val_accuracy': 0.8143388628959656, 'credit_card_default_val_precision': 0.7957068681716919, 'credit_card_default_val_recall': 0.8143388628959656, 'credit_card_default_val_f1-score': 0.791758120059967, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7507679462432861, 'credit_card_default_val_score': 0.8143388628959656}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 767 samples (12.88%) + Class 1: 5189 samples (87.12%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8190060257911682 │ +│ credit_card_default_test_f1-score │ 0.7994917035102844 │ +│ credit_card_default_test_loss │ 0.4400160014629364 │ +│ credit_card_default_test_precision │ 0.8030407428741455 │ +│ credit_card_default_test_recall │ 0.8190060257911682 │ +│ credit_card_default_test_roc_auc │ 0.7647790312767029 │ +│ credit_card_default_test_score │ 0.8190060257911682 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 0.0 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 211.72it/s +Test metrics: [{'credit_card_default_test_loss': 0.4400160014629364, 'lr': 0.0, 'credit_card_default_test_accuracy': 0.8190060257911682, 'credit_card_default_test_precision': 0.8030407428741455, 'credit_card_default_test_recall': 0.8190060257911682, 'credit_card_default_test_f1-score': 0.7994917035102844, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.7647790312767029, 'credit_card_default_test_score': 0.8190060257911682}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpjq29br_h/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpjq29br_h/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpjq29br_h/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpjq29br_h/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 23 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤═════════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪═════════════════════╡ +│ input_dim │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_cont_features │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ T_max │ 8250 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dataset_name │ credit_card_default │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧═════════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─EncoderFTTransformer: 2-2 [128, 23] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 23] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 23] [128, 23, 128] 5,888 -- 753,664 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 24, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 24, 24] [512, 24, 24] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 24, 128] [128, 24, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 24, 512] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 24, 256] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 24, 256] [128, 24, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 24, 24] [512, 24, 24] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 24, 128] [128, 24, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 24, 512] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 24, 256] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 24, 256] [128, 24, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 24] [512, 1, 24] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 24, 128] [128, 24, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 24, 512] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 24, 256] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 24, 256] [128, 24, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================================= +Total params: 553,048 +Trainable params: 553,048 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 70.79 +========================================================================================================================================================================================= +Input size (MB): 0.75 +Forward/backward pass size (MB): 98.07 +Params size (MB): 2.21 +Estimated Total Size (MB): 101.03 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 552 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 553 K +Non-trainable params: 0 +Total params: 553 K +Total estimated model params size (MB): 2.212 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:01 • 0:00:00 82.59it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8150151968002319 │ +│ credit_card_default_val_f1-score │ 0.7914837002754211 │ +│ credit_card_default_val_loss │ 0.44776108860969543 │ +│ credit_card_default_val_precision │ 0.7966932058334351 │ +│ credit_card_default_val_recall │ 0.8150151968002319 │ +│ credit_card_default_val_roc_auc │ 0.7525211572647095 │ +│ credit_card_default_val_score │ 0.8150151968002319 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 0.0 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 246.94it/s +Validation metrics: [{'credit_card_default_val_loss': 0.44776108860969543, 'lr': 0.0, 'credit_card_default_val_accuracy': 0.8150151968002319, 'credit_card_default_val_precision': 0.7966932058334351, 'credit_card_default_val_recall': 0.8150151968002319, 'credit_card_default_val_f1-score': 0.7914837002754211, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7525211572647095, 'credit_card_default_val_score': 0.8150151968002319}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 718 samples (12.06%) + Class 1: 5238 samples (87.94%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8185023665428162 │ +│ credit_card_default_test_f1-score │ 0.7966155409812927 │ +│ credit_card_default_test_loss │ 0.44099584221839905 │ +│ credit_card_default_test_precision │ 0.8026161789894104 │ +│ credit_card_default_test_recall │ 0.8185023665428162 │ +│ credit_card_default_test_roc_auc │ 0.7639510631561279 │ +│ credit_card_default_test_score │ 0.8185023665428162 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 0.0 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 209.42it/s +Test metrics: [{'credit_card_default_test_loss': 0.44099584221839905, 'lr': 0.0, 'credit_card_default_test_accuracy': 0.8185023665428162, 'credit_card_default_test_precision': 0.8026161789894104, 'credit_card_default_test_recall': 0.8185023665428162, 'credit_card_default_test_f1-score': 0.7966155409812927, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.7639510631561279, 'credit_card_default_test_score': 0.8185023665428162}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpjq29br_h/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpjq29br_h/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpjq29br_h/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpjq29br_h/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 23 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤═════════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪═════════════════════╡ +│ input_dim │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_cont_features │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ T_max │ 8250 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dataset_name │ credit_card_default │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧═════════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─EncoderFTTransformer: 2-2 [128, 23] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 23] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 23] [128, 23, 128] 5,888 -- 753,664 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 24, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 24, 24] [512, 24, 24] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 24, 128] [128, 24, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 24, 512] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 24, 256] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 24, 256] [128, 24, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 24, 24] [512, 24, 24] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 24, 128] [128, 24, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 24, 512] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 24, 256] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 24, 256] [128, 24, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 24] [512, 1, 24] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 24, 128] [128, 24, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 24, 512] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 24, 256] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 24, 256] [128, 24, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================================= +Total params: 553,048 +Trainable params: 553,048 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 70.79 +========================================================================================================================================================================================= +Input size (MB): 0.75 +Forward/backward pass size (MB): 98.07 +Params size (MB): 2.21 +Estimated Total Size (MB): 101.03 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 552 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 553 K +Non-trainable params: 0 +Total params: 553 K +Total estimated model params size (MB): 2.212 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:02 • 0:00:00 80.56it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8160297870635986 │ +│ credit_card_default_val_f1-score │ 0.7916812300682068 │ +│ credit_card_default_val_loss │ 0.4494698643684387 │ +│ credit_card_default_val_precision │ 0.7982540726661682 │ +│ credit_card_default_val_recall │ 0.8160297870635986 │ +│ credit_card_default_val_roc_auc │ 0.7508789300918579 │ +│ credit_card_default_val_score │ 0.8160297870635986 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 0.0 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 260.50it/s +Validation metrics: [{'credit_card_default_val_loss': 0.4494698643684387, 'lr': 0.0, 'credit_card_default_val_accuracy': 0.8160297870635986, 'credit_card_default_val_precision': 0.7982540726661682, 'credit_card_default_val_recall': 0.8160297870635986, 'credit_card_default_val_f1-score': 0.7916812300682068, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7508789300918579, 'credit_card_default_val_score': 0.8160297870635986}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 741 samples (12.44%) + Class 1: 5215 samples (87.56%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8176628351211548 │ +│ credit_card_default_test_f1-score │ 0.7967845797538757 │ +│ credit_card_default_test_loss │ 0.4420795738697052 │ +│ credit_card_default_test_precision │ 0.8013173937797546 │ +│ credit_card_default_test_recall │ 0.8176628351211548 │ +│ credit_card_default_test_roc_auc │ 0.761913001537323 │ +│ credit_card_default_test_score │ 0.8176628351211548 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 0.0 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 228.30it/s +Test metrics: [{'credit_card_default_test_loss': 0.4420795738697052, 'lr': 0.0, 'credit_card_default_test_accuracy': 0.8176628351211548, 'credit_card_default_test_precision': 0.8013173937797546, 'credit_card_default_test_recall': 0.8176628351211548, 'credit_card_default_test_f1-score': 0.7967845797538757, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.761913001537323, 'credit_card_default_test_score': 0.8176628351211548}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpjq29br_h/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpjq29br_h/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpjq29br_h/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpjq29br_h/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 23 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤═════════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪═════════════════════╡ +│ input_dim │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_cont_features │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ T_max │ 8250 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dataset_name │ credit_card_default │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧═════════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─EncoderFTTransformer: 2-2 [128, 23] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 23] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 23] [128, 23, 128] 5,888 -- 753,664 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 24, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 24, 24] [512, 24, 24] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 24, 128] [128, 24, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 24, 512] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 24, 256] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 24, 256] [128, 24, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 24, 24] [512, 24, 24] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 24, 128] [128, 24, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 24, 512] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 24, 256] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 24, 256] [128, 24, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 24] [512, 1, 24] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 24, 128] [128, 24, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 24, 512] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 24, 256] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 24, 256] [128, 24, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================================= +Total params: 553,048 +Trainable params: 553,048 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 70.79 +========================================================================================================================================================================================= +Input size (MB): 0.75 +Forward/backward pass size (MB): 98.07 +Params size (MB): 2.21 +Estimated Total Size (MB): 101.03 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 552 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 553 K +Non-trainable params: 0 +Total params: 553 K +Total estimated model params size (MB): 2.212 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:01 • 0:00:00 87.35it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8156915903091431 │ +│ credit_card_default_val_f1-score │ 0.7905467748641968 │ +│ credit_card_default_val_loss │ 0.44779202342033386 │ +│ credit_card_default_val_precision │ 0.7979090213775635 │ +│ credit_card_default_val_recall │ 0.8156915903091431 │ +│ credit_card_default_val_roc_auc │ 0.7489837408065796 │ +│ credit_card_default_val_score │ 0.8156915903091431 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 0.0 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 206.10it/s +Validation metrics: [{'credit_card_default_val_loss': 0.44779202342033386, 'lr': 0.0, 'credit_card_default_val_accuracy': 0.8156915903091431, 'credit_card_default_val_precision': 0.7979090213775635, 'credit_card_default_val_recall': 0.8156915903091431, 'credit_card_default_val_f1-score': 0.7905467748641968, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.7489837408065796, 'credit_card_default_val_score': 0.8156915903091431}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 718 samples (12.06%) + Class 1: 5238 samples (87.94%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8181665539741516 │ +│ credit_card_default_test_f1-score │ 0.7962392568588257 │ +│ credit_card_default_test_loss │ 0.44313955307006836 │ +│ credit_card_default_test_precision │ 0.8021530508995056 │ +│ credit_card_default_test_recall │ 0.8181665539741516 │ +│ credit_card_default_test_roc_auc │ 0.759540319442749 │ +│ credit_card_default_test_score │ 0.8181665539741516 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 0.0 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 205.02it/s +Test metrics: [{'credit_card_default_test_loss': 0.44313955307006836, 'lr': 0.0, 'credit_card_default_test_accuracy': 0.8181665539741516, 'credit_card_default_test_precision': 0.8021530508995056, 'credit_card_default_test_recall': 0.8181665539741516, 'credit_card_default_test_f1-score': 0.7962392568588257, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.759540319442749, 'credit_card_default_test_score': 0.8181665539741516}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=credit_card_default --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpjq29br_h/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpjq29br_h/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ credit_card_default │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpjq29br_h/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpjq29br_h/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 23 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 165 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 23 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 23, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤═════════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪═════════════════════╡ +│ input_dim │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_cont_features │ 23 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ T_max │ 8250 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ dataset_name │ credit_card_default │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼─────────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧═════════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 23, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 23, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 23, 64] [128, 23] -- -- -- +│ │ └─Linear: 3-1 [128, 1472] [128, 23] 33,879 -- 4,336,512 +│ └─EncoderFTTransformer: 2-2 [128, 23] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 23] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 23] [128, 23, 128] 5,888 -- 753,664 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 24, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 24, 24] [512, 24, 24] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 24, 128] [128, 24, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 24, 512] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 24, 256] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 24, 256] [128, 24, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 24, 24] [512, 24, 24] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 24, 128] [128, 24, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 24, 512] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 24, 256] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 24, 256] [128, 24, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 24, 128] [128, 24, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 24] [512, 1, 24] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 24, 128] [128, 24, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 24, 128] [128, 24, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 24, 512] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 24, 256] [128, 24, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 24, 256] [128, 24, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 24, 128] [128, 24, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================================= +Total params: 553,048 +Trainable params: 553,048 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 70.79 +========================================================================================================================================================================================= +Input size (MB): 0.75 +Forward/backward pass size (MB): 98.07 +Params size (MB): 2.21 +Estimated Total Size (MB): 101.03 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 552 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 553 K +Non-trainable params: 0 +Total params: 553 K +Total estimated model params size (MB): 2.212 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 165/165 0:00:01 • 0:00:00 88.24it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_val_accuracy │ 0.8133243322372437 │ +│ credit_card_default_val_f1-score │ 0.7903100252151489 │ +│ credit_card_default_val_loss │ 0.44690796732902527 │ +│ credit_card_default_val_precision │ 0.7943000197410583 │ +│ credit_card_default_val_recall │ 0.8133243322372437 │ +│ credit_card_default_val_roc_auc │ 0.752513587474823 │ +│ credit_card_default_val_score │ 0.8133243322372437 │ +│ credit_card_default_val_support │ 2957.0 │ +│ lr │ 0.0 │ +└───────────────────────────────────┴───────────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24/24 0:00:00 • 0:00:00 213.71it/s +Validation metrics: [{'credit_card_default_val_loss': 0.44690796732902527, 'lr': 0.0, 'credit_card_default_val_accuracy': 0.8133243322372437, 'credit_card_default_val_precision': 0.7943000197410583, 'credit_card_default_val_recall': 0.8133243322372437, 'credit_card_default_val_f1-score': 0.7903100252151489, 'credit_card_default_val_support': 2957.0, 'credit_card_default_val_roc_auc': 0.752513587474823, 'credit_card_default_val_score': 0.8133243322372437}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 1349 samples (22.65%) + Class 1.0: 4607 samples (77.35%) +Predicted class frequencies: + Class 0: 733 samples (12.31%) + Class 1: 5223 samples (87.69%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ credit_card_default_test_accuracy │ 0.8179986476898193 │ +│ credit_card_default_test_f1-score │ 0.7967769503593445 │ +│ credit_card_default_test_loss │ 0.43984168767929077 │ +│ credit_card_default_test_precision │ 0.8018132448196411 │ +│ credit_card_default_test_recall │ 0.8179986476898193 │ +│ credit_card_default_test_roc_auc │ 0.7656611800193787 │ +│ credit_card_default_test_score │ 0.8179986476898193 │ +│ credit_card_default_test_support │ 5956.0 │ +│ lr │ 0.0 │ +└────────────────────────────────────┴────────────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47/47 0:00:00 • 0:00:00 207.07it/s +Test metrics: [{'credit_card_default_test_loss': 0.43984168767929077, 'lr': 0.0, 'credit_card_default_test_accuracy': 0.8179986476898193, 'credit_card_default_test_precision': 0.8018132448196411, 'credit_card_default_test_recall': 0.8179986476898193, 'credit_card_default_test_f1-score': 0.7967769503593445, 'credit_card_default_test_support': 5956.0, 'credit_card_default_test_roc_auc': 0.7656611800193787, 'credit_card_default_test_score': 0.8179986476898193}] +Wrote results for fttransformer on credit_card_default to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/downstream_results/fttransformer_credit_card_default_downstream_results.json + +Downstream pipeline complete. Results under /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_182951/downstream_results diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu3-electricity.exit_code b/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu3-electricity.exit_code new file mode 100644 index 0000000000000000000000000000000000000000..573541ac9702dd3969c9bc859d2b91ec1f7e6e56 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu3-electricity.exit_code @@ -0,0 +1 @@ +0 diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu3-electricity.log b/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu3-electricity.log new file mode 100644 index 0000000000000000000000000000000000000000..5d3119b886e48f3ecb976bdcdd4a0c1a8017f199 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/gpu3-electricity.log @@ -0,0 +1,13155 @@ +[gpu3-electricity] Starting electricity on physical GPU 3 at Tue Sep 15 18:28:13 UTC 2026 ++ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py --dataset electricity --results-root /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42 --epochs 500 --objective w_jepa_joint --latent-architecture deterministic_no_z --sw-full-end-weight 0.00025 --torch-seed 42 --np-seed 42 --mask-seed 0 --sw-seed 0 --sw-full-num-projections 2048 --sw-marginal-num-projections 1024 --sw-p 2 --sw-s-x-end-weight 5e-05 --sw-z-end-weight 0.0 --sw-s-y-end-weight 1e-05 --model-selection unfiltered --deterministic-mode warn --sw-include-marginals --batch-size 512 +{ + "dataset": "electricity", + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "electricity", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "64", + "--model_num_layers", + "6", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.15", + "--mask_max_ctx_share", + "0.6", + "--mask_min_trgt_share", + "0.15", + "--mask_max_trgt_share", + "0.9", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "128", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.0005", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "1e-06", + "--vtjepa_kl_x_annealing_epochs", + "40", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-05", + "--vtjepa_kl_z_annealing_epochs", + "40", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-06", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.001", + "--recon_y_weight", + "0.25", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.00025", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "40", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "5e-05", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "1e-05", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} +Not SLURM job. +╒══════════════════════════════════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ mock │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ activation_checkpointing │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 42 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 42 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_seed │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_seed │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ deterministic_mode │ warn │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pre_evaluation │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 512 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 500 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ improvement │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_val_cache_cadence │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ init_type │ trunc_normal │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_start_lr │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_lr │ 0.0005 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_final_lr │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 10 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_final_weight_decay │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_ipe_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_gradient_clipping │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_feature_type_embedding │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_feature_index_embedding │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dropout_prob │ 0.001 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_layer_norm_eps │ 1e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dim_hidden │ 64 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dim_feedforward │ 64 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_num_heads │ 4 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_num_layers │ 6 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_act_func │ relu │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ probe_cadence │ 2 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ probe_model │ linear_probe │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ n_cls_tokens │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_type │ transformer │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_num_layers │ 16 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_embed_dim │ 16 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_num_heads │ 4 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_p_dropout │ 0.002 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_layer_norm_eps │ 1e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_activation │ relu │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_dim_feedforward │ 128 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_allow_overlap │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_min_ctx_share │ 0.15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_max_ctx_share │ 0.6 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_min_trgt_share │ 0.15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_max_trgt_share │ 0.9 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_num_preds │ 4 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_num_encs │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mu_init_std │ 0.03 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ logvar_init_std │ 0.005 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ logvar_init_bias │ -4.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_sigma_x_init │ -2.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_sigma_y_init │ -2.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ downstream_evaluation_epochs │ 50 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ downstream_probe_patience │ 20 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ probe_representation_type │ mean │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ probe_encoder_type │ target_posterior │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pooled_context_representation_size_for_target_posterior │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pooled_context_representation_size_for_auxiliary_encoder │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_end_weight │ 1e-06 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_end_weight │ 1e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_end_weight │ 1e-06 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_annealing_epochs │ 40 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_annealing_epochs │ 40 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_annealing_epochs │ 15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ recon_x_weight │ 0.001 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ recon_y_weight │ 0.25 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ latent_architecture │ deterministic_no_z │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ jepa_objective │ w_jepa_joint │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_full_num_projections │ 2048 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_marginal_num_projections │ 1024 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_p │ 2 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_include_marginals │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_s_x_weight │ 5e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_z_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_s_y_weight │ 1e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_end_weight │ 0.00025 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_annealing_epochs │ 40 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_marginal_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_marginal_end_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_end_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_start_epoch │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_annealing_epochs │ 30 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_end_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_start_epoch │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_annealing_epochs │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_end_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_start_epoch │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_annealing_epochs │ 30 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ safe_exit │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_selection │ unfiltered │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sigma_aggregation │ mean │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ subset_size │ -1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ is_batchlearning │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ iteration │ 0 │ +╘══════════════════════════════════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ + +CUDA Device Information: +CUDA_VISIBLE_DEVICES: 3 +Number of CUDA devices: 1 +Current CUDA device: 0 +CUDA device name: NVIDIA A100-SXM4-80GB + +Using device: cuda:0 +Device name: NVIDIA A100-SXM4-80GB +Encoder Module parameters: ++-------------------+--------------------------+ +| Parameter | Value | ++===================+==========================+ +| hidden_dim | 64 | ++-------------------+--------------------------+ +| idx_num_features | [0, 1, 2, 3, 4, 5, 6, 7] | ++-------------------+--------------------------+ +| idx_cat_features | [] | ++-------------------+--------------------------+ +| cardinalities | [] | ++-------------------+--------------------------+ +| n_input_features | 8 | ++-------------------+--------------------------+ +| device | cuda:0 | ++-------------------+--------------------------+ +| num_layers | 6 | ++-------------------+--------------------------+ +| num_heads | 4 | ++-------------------+--------------------------+ +| p_dropout | 0.001 | ++-------------------+--------------------------+ +| layer_norm_eps | 1e-05 | ++-------------------+--------------------------+ +| gradient_clipping | 1.0 | ++-------------------+--------------------------+ +| dim_feedforward | 64 | ++-------------------+--------------------------+ +Encoder Module parameters: ++-------------------+--------------------------+ +| Parameter | Value | ++===================+==========================+ +| hidden_dim | 64 | ++-------------------+--------------------------+ +| idx_num_features | [0, 1, 2, 3, 4, 5, 6, 7] | ++-------------------+--------------------------+ +| idx_cat_features | [] | ++-------------------+--------------------------+ +| cardinalities | [] | ++-------------------+--------------------------+ +| n_input_features | 8 | ++-------------------+--------------------------+ +| device | cuda:0 | ++-------------------+--------------------------+ +| num_layers | 6 | ++-------------------+--------------------------+ +| num_heads | 4 | ++-------------------+--------------------------+ +| p_dropout | 0.001 | ++-------------------+--------------------------+ +| layer_norm_eps | 1e-05 | ++-------------------+--------------------------+ +| gradient_clipping | 1.0 | ++-------------------+--------------------------+ +| dim_feedforward | 64 | ++-------------------+--------------------------+ +DJEPA parameters: ++-------------------------+---------+ +| Parameter | Value | ++=========================+=========+ +| hidden_dim | 64 | ++-------------------------+---------+ +| num_layers | 6 | ++-------------------------+---------+ +| num_heads | 4 | ++-------------------------+---------+ +| p_dropout | 0.001 | ++-------------------------+---------+ +| layer_norm_eps | 1e-05 | ++-------------------------+---------+ +| feature_type_embedding | True | ++-------------------------+---------+ +| feature_index_embedding | True | ++-------------------------+---------+ +| dim_feedforward | 64 | ++-------------------------+---------+ +| use_reg_token | True | ++-------------------------+---------+ +| gradient_clipping | 1.0 | ++-------------------------+---------+ +| activation | gelu | ++-------------------------+---------+ +| num_features | 8 | ++-------------------------+---------+ +| n_cls_tokens | 1 | ++-------------------------+---------+ +MaskCollator params: +----------------- ----- +Allow overlap False +Min context share 0.15 +Max context share 0.6 +Min target share 0.15 +Max target share 0.9 +Min context 1 +Max context 4 +Min target 1 +Max target 7 +Num preds 4 +Num encs 1 +Num features 8 +Cardinalities [] +Base mask seed 0 +----------------- ----- +KL-x annealing: start=0.0 for 89 steps, then anneal to 1e-06 over 3560 steps. +KL-z annealing: start=0.0 for 89 steps, then anneal to 1e-05 over 3560 steps. +KL-y annealing: start=0.0 for 89 steps, then anneal to 1e-06 over 1335 steps. +W-JEPA marginal-scale annealing (s_x): start=0.0, end=1.0, start_epoch=0, anneal_epochs=30. +W-JEPA marginal-scale annealing (z): start=0.0, end=0.0, start_epoch=0, anneal_epochs=0. +W-JEPA marginal-scale annealing (s_y): start=0.0, end=1.0, start_epoch=0, anneal_epochs=30. +Configuration saved to: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/vtjepa_config.yaml +All results will be saved to: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946 +[VTJEPA] Running downstream probe before training begins +Using full dataset for embedding generation: electricity +Generating embedding using: JEPAEmbeddingEncoder on full dataset + 0%| | 0/89 [00:00= 15) +[VTJEPA] Early stopping triggered at epoch 44 + Best downstream validation score: 0.769838 at epoch 13 + No improvement for 15 probe evaluations +Restored best downstream-selected checkpoint from epoch 13. +Training completed. ++ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -W ignore -m src.downstream_pipeline --exp_path /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946 --num_runs 5 --epochs 50 --models mlp,dcnv2,resnet,autoint,fttransformer --no-filter_uncertainty +[Config] Loaded sigma_aggregation=mean from /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/vtjepa_config.yaml +[Config] Loaded training data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets from /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/vtjepa_config.yaml +Split statistics: + val: 4436 samples + test: 9031 samples + train: 31845 samples +Running downstream for dataset 'electricity', model 'mlp'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpo8zc5dpm/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpo8zc5dpm/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpo8zc5dpm/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpo8zc5dpm/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ electricity │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─Linear: 2-2 [128, 8] [128, 128] 1,152 -- 147,456 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +===================================================================================================================================================================== +Total params: 22,153 +Trainable params: 22,153 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 2.84 +===================================================================================================================================================================== +Input size (MB): 0.26 +Forward/backward pass size (MB): 0.40 +Params size (MB): 0.09 +Estimated Total Size (MB): 0.75 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 22.0 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 22.2 K +Non-trainable params: 0 +Total params: 22.2 K +Total estimated model params size (MB): 0.089 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 234.98it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.7842651009559631 │ +│ electricity_val_f1-score │ 0.7781946659088135 │ +│ electricity_val_loss │ 0.45377448201179504 │ +│ electricity_val_precision │ 0.7901862859725952 │ +│ electricity_val_recall │ 0.7842651009559631 │ +│ electricity_val_roc_auc │ 0.8673897981643677 │ +│ electricity_val_score │ 0.7842651009559631 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0009999999310821295 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 581.13it/s +Validation metrics: [{'electricity_val_loss': 0.45377448201179504, 'lr': 0.0009999999310821295, 'electricity_val_accuracy': 0.7842651009559631, 'electricity_val_precision': 0.7901862859725952, 'electricity_val_recall': 0.7842651009559631, 'electricity_val_f1-score': 0.7781946659088135, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.8673897981643677, 'electricity_val_score': 0.7842651009559631}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 2901 samples (32.12%) + Class 1: 6130 samples (67.88%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.7785406112670898 │ +│ electricity_test_f1-score │ 0.7723544239997864 │ +│ electricity_test_loss │ 0.4639207720756531 │ +│ electricity_test_precision │ 0.783967912197113 │ +│ electricity_test_recall │ 0.7785406112670898 │ +│ electricity_test_roc_auc │ 0.8621642589569092 │ +│ electricity_test_score │ 0.7785406112670898 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0009999994654208422 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 590.78it/s +Test metrics: [{'electricity_test_loss': 0.4639207720756531, 'lr': 0.0009999994654208422, 'electricity_test_accuracy': 0.7785406112670898, 'electricity_test_precision': 0.783967912197113, 'electricity_test_recall': 0.7785406112670898, 'electricity_test_f1-score': 0.7723544239997864, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.8621642589569092, 'electricity_test_score': 0.7785406112670898}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpo8zc5dpm/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpo8zc5dpm/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpo8zc5dpm/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpo8zc5dpm/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ electricity │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─Linear: 2-2 [128, 8] [128, 128] 1,152 -- 147,456 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +===================================================================================================================================================================== +Total params: 22,153 +Trainable params: 22,153 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 2.84 +===================================================================================================================================================================== +Input size (MB): 0.26 +Forward/backward pass size (MB): 0.40 +Params size (MB): 0.09 +Estimated Total Size (MB): 0.75 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 22.0 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 22.2 K +Non-trainable params: 0 +Total params: 22.2 K +Total estimated model params size (MB): 0.089 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 220.11it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.7824617028236389 │ +│ electricity_val_f1-score │ 0.7759427428245544 │ +│ electricity_val_loss │ 0.4568510353565216 │ +│ electricity_val_precision │ 0.7891604900360107 │ +│ electricity_val_recall │ 0.7824617028236389 │ +│ electricity_val_roc_auc │ 0.8643500804901123 │ +│ electricity_val_score │ 0.7824617028236389 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0009999999310821295 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 488.40it/s +Validation metrics: [{'electricity_val_loss': 0.4568510353565216, 'lr': 0.0009999999310821295, 'electricity_val_accuracy': 0.7824617028236389, 'electricity_val_precision': 0.7891604900360107, 'electricity_val_recall': 0.7824617028236389, 'electricity_val_f1-score': 0.7759427428245544, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.8643500804901123, 'electricity_val_score': 0.7824617028236389}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 2843 samples (31.48%) + Class 1: 6188 samples (68.52%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.775218665599823 │ +│ electricity_test_f1-score │ 0.7683533430099487 │ +│ electricity_test_loss │ 0.46997353434562683 │ +│ electricity_test_precision │ 0.781640887260437 │ +│ electricity_test_recall │ 0.775218665599823 │ +│ electricity_test_roc_auc │ 0.8589786887168884 │ +│ electricity_test_score │ 0.775218665599823 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0009999994654208422 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 491.40it/s +Test metrics: [{'electricity_test_loss': 0.46997353434562683, 'lr': 0.0009999994654208422, 'electricity_test_accuracy': 0.775218665599823, 'electricity_test_precision': 0.781640887260437, 'electricity_test_recall': 0.775218665599823, 'electricity_test_f1-score': 0.7683533430099487, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.8589786887168884, 'electricity_test_score': 0.775218665599823}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpo8zc5dpm/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpo8zc5dpm/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpo8zc5dpm/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpo8zc5dpm/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ electricity │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─Linear: 2-2 [128, 8] [128, 128] 1,152 -- 147,456 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +===================================================================================================================================================================== +Total params: 22,153 +Trainable params: 22,153 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 2.84 +===================================================================================================================================================================== +Input size (MB): 0.26 +Forward/backward pass size (MB): 0.40 +Params size (MB): 0.09 +Estimated Total Size (MB): 0.75 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 22.0 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 22.2 K +Non-trainable params: 0 +Total params: 22.2 K +Total estimated model params size (MB): 0.089 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 230.15it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.7880973815917969 │ +│ electricity_val_f1-score │ 0.7856864333152771 │ +│ electricity_val_loss │ 0.446729838848114 │ +│ electricity_val_precision │ 0.7878503799438477 │ +│ electricity_val_recall │ 0.7880973815917969 │ +│ electricity_val_roc_auc │ 0.8661814332008362 │ +│ electricity_val_score │ 0.7880973815917969 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0009999999310821295 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 495.12it/s +Validation metrics: [{'electricity_val_loss': 0.446729838848114, 'lr': 0.0009999999310821295, 'electricity_val_accuracy': 0.7880973815917969, 'electricity_val_precision': 0.7878503799438477, 'electricity_val_recall': 0.7880973815917969, 'electricity_val_f1-score': 0.7856864333152771, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.8661814332008362, 'electricity_val_score': 0.7880973815917969}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3325 samples (36.82%) + Class 1: 5706 samples (63.18%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.7847414612770081 │ +│ electricity_test_f1-score │ 0.7821466326713562 │ +│ electricity_test_loss │ 0.45572444796562195 │ +│ electricity_test_precision │ 0.7845763564109802 │ +│ electricity_test_recall │ 0.7847414612770081 │ +│ electricity_test_roc_auc │ 0.8619311451911926 │ +│ electricity_test_score │ 0.7847414612770081 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0009999994654208422 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 497.26it/s +Test metrics: [{'electricity_test_loss': 0.45572444796562195, 'lr': 0.0009999994654208422, 'electricity_test_accuracy': 0.7847414612770081, 'electricity_test_precision': 0.7845763564109802, 'electricity_test_recall': 0.7847414612770081, 'electricity_test_f1-score': 0.7821466326713562, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.8619311451911926, 'electricity_test_score': 0.7847414612770081}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpo8zc5dpm/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpo8zc5dpm/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpo8zc5dpm/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpo8zc5dpm/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ electricity │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─Linear: 2-2 [128, 8] [128, 128] 1,152 -- 147,456 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +===================================================================================================================================================================== +Total params: 22,153 +Trainable params: 22,153 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 2.84 +===================================================================================================================================================================== +Input size (MB): 0.26 +Forward/backward pass size (MB): 0.40 +Params size (MB): 0.09 +Estimated Total Size (MB): 0.75 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 22.0 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 22.2 K +Non-trainable params: 0 +Total params: 22.2 K +Total estimated model params size (MB): 0.089 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:00 • 0:00:00 256.69it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.7840396761894226 │ +│ electricity_val_f1-score │ 0.782926619052887 │ +│ electricity_val_loss │ 0.45924338698387146 │ +│ electricity_val_precision │ 0.7830287218093872 │ +│ electricity_val_recall │ 0.7840396761894226 │ +│ electricity_val_roc_auc │ 0.8588030934333801 │ +│ electricity_val_score │ 0.7840396761894226 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0009999999310821295 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 492.83it/s +Validation metrics: [{'electricity_val_loss': 0.45924338698387146, 'lr': 0.0009999999310821295, 'electricity_val_accuracy': 0.7840396761894226, 'electricity_val_precision': 0.7830287218093872, 'electricity_val_recall': 0.7840396761894226, 'electricity_val_f1-score': 0.782926619052887, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.8588030934333801, 'electricity_val_score': 0.7840396761894226}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3586 samples (39.71%) + Class 1: 5445 samples (60.29%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.7766581773757935 │ +│ electricity_test_f1-score │ 0.7755626440048218 │ +│ electricity_test_loss │ 0.4670161306858063 │ +│ electricity_test_precision │ 0.7755798697471619 │ +│ electricity_test_recall │ 0.7766581773757935 │ +│ electricity_test_roc_auc │ 0.8529821634292603 │ +│ electricity_test_score │ 0.7766581773757935 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0009999994654208422 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 496.16it/s +Test metrics: [{'electricity_test_loss': 0.4670161306858063, 'lr': 0.0009999994654208422, 'electricity_test_accuracy': 0.7766581773757935, 'electricity_test_precision': 0.7755798697471619, 'electricity_test_recall': 0.7766581773757935, 'electricity_test_f1-score': 0.7755626440048218, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.8529821634292603, 'electricity_test_score': 0.7766581773757935}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpo8zc5dpm/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpo8zc5dpm/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpo8zc5dpm/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpo8zc5dpm/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ electricity │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─Linear: 2-2 [128, 8] [128, 128] 1,152 -- 147,456 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +===================================================================================================================================================================== +Total params: 22,153 +Trainable params: 22,153 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 2.84 +===================================================================================================================================================================== +Input size (MB): 0.26 +Forward/backward pass size (MB): 0.40 +Params size (MB): 0.09 +Estimated Total Size (MB): 0.75 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 22.0 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 22.2 K +Non-trainable params: 0 +Total params: 22.2 K +Total estimated model params size (MB): 0.089 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 44/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:00 • 0:00:00 258.46it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.7790802717208862 │ +│ electricity_val_f1-score │ 0.771767258644104 │ +│ electricity_val_loss │ 0.47163620591163635 │ +│ electricity_val_precision │ 0.787118673324585 │ +│ electricity_val_recall │ 0.7790802717208862 │ +│ electricity_val_roc_auc │ 0.8560341596603394 │ +│ electricity_val_score │ 0.7790802717208862 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0009999999310821295 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 566.62it/s +Validation metrics: [{'electricity_val_loss': 0.47163620591163635, 'lr': 0.0009999999310821295, 'electricity_val_accuracy': 0.7790802717208862, 'electricity_val_precision': 0.787118673324585, 'electricity_val_recall': 0.7790802717208862, 'electricity_val_f1-score': 0.771767258644104, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.8560341596603394, 'electricity_val_score': 0.7790802717208862}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 2771 samples (30.68%) + Class 1: 6260 samples (69.32%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.7685748934745789 │ +│ electricity_test_f1-score │ 0.7607215046882629 │ +│ electricity_test_loss │ 0.48404136300086975 │ +│ electricity_test_precision │ 0.7760655283927917 │ +│ electricity_test_recall │ 0.7685748934745789 │ +│ electricity_test_roc_auc │ 0.8485464453697205 │ +│ electricity_test_score │ 0.7685748934745789 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0009999994654208422 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 577.45it/s +Test metrics: [{'electricity_test_loss': 0.48404136300086975, 'lr': 0.0009999994654208422, 'electricity_test_accuracy': 0.7685748934745789, 'electricity_test_precision': 0.7760655283927917, 'electricity_test_recall': 0.7685748934745789, 'electricity_test_f1-score': 0.7607215046882629, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.8485464453697205, 'electricity_test_score': 0.7685748934745789}] +Wrote results for mlp on electricity to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results/mlp_electricity_downstream_results.json +Running downstream for dataset 'electricity', model 'dcnv2'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpj6fuqx4a/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpj6fuqx4a/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpj6fuqx4a/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpj6fuqx4a/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ electricity │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─EncoderDCNv2: 2-2 [128, 8] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 8] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 8] [128, 128] 1,152 -- 147,456 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 120,969 +Trainable params: 120,969 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 15.48 +========================================================================================================================================================================== +Input size (MB): 0.26 +Forward/backward pass size (MB): 1.06 +Params size (MB): 0.48 +Estimated Total Size (MB): 1.80 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 120 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 120 K +Non-trainable params: 0 +Total params: 120 K +Total estimated model params size (MB): 0.484 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 147.50it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.8331830501556396 │ +│ electricity_val_f1-score │ 0.8325451612472534 │ +│ electricity_val_loss │ 0.36828869581222534 │ +│ electricity_val_precision │ 0.832675039768219 │ +│ electricity_val_recall │ 0.8331830501556396 │ +│ electricity_val_roc_auc │ 0.9150740504264832 │ +│ electricity_val_score │ 0.8331830501556396 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 376.40it/s +Validation metrics: [{'electricity_val_loss': 0.36828869581222534, 'lr': 0.0, 'electricity_val_accuracy': 0.8331830501556396, 'electricity_val_precision': 0.832675039768219, 'electricity_val_recall': 0.8331830501556396, 'electricity_val_f1-score': 0.8325451612472534, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.9150740504264832, 'electricity_val_score': 0.8331830501556396}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3724 samples (41.24%) + Class 1: 5307 samples (58.76%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.8351234793663025 │ +│ electricity_test_f1-score │ 0.8348125219345093 │ +│ electricity_test_loss │ 0.37148943543434143 │ +│ electricity_test_precision │ 0.8347146511077881 │ +│ electricity_test_recall │ 0.8351234793663025 │ +│ electricity_test_roc_auc │ 0.913567841053009 │ +│ electricity_test_score │ 0.8351234793663025 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 402.30it/s +Test metrics: [{'electricity_test_loss': 0.37148943543434143, 'lr': 0.0, 'electricity_test_accuracy': 0.8351234793663025, 'electricity_test_precision': 0.8347146511077881, 'electricity_test_recall': 0.8351234793663025, 'electricity_test_f1-score': 0.8348125219345093, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.913567841053009, 'electricity_test_score': 0.8351234793663025}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpj6fuqx4a/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpj6fuqx4a/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpj6fuqx4a/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpj6fuqx4a/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ electricity │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─EncoderDCNv2: 2-2 [128, 8] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 8] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 8] [128, 128] 1,152 -- 147,456 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 120,969 +Trainable params: 120,969 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 15.48 +========================================================================================================================================================================== +Input size (MB): 0.26 +Forward/backward pass size (MB): 1.06 +Params size (MB): 0.48 +Estimated Total Size (MB): 1.80 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 120 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 120 K +Non-trainable params: 0 +Total params: 120 K +Total estimated model params size (MB): 0.484 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:02 • 0:00:00 110.32it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.8325067758560181 │ +│ electricity_val_f1-score │ 0.831684410572052 │ +│ electricity_val_loss │ 0.37160301208496094 │ +│ electricity_val_precision │ 0.8320543169975281 │ +│ electricity_val_recall │ 0.8325067758560181 │ +│ electricity_val_roc_auc │ 0.9119824767112732 │ +│ electricity_val_score │ 0.8325067758560181 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 93.06it/s +Validation metrics: [{'electricity_val_loss': 0.37160301208496094, 'lr': 0.0, 'electricity_val_accuracy': 0.8325067758560181, 'electricity_val_precision': 0.8320543169975281, 'electricity_val_recall': 0.8325067758560181, 'electricity_val_f1-score': 0.831684410572052, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.9119824767112732, 'electricity_val_score': 0.8325067758560181}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3651 samples (40.43%) + Class 1: 5380 samples (59.57%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.834126889705658 │ +│ electricity_test_f1-score │ 0.8335598111152649 │ +│ electricity_test_loss │ 0.37350720167160034 │ +│ electricity_test_precision │ 0.8336272835731506 │ +│ electricity_test_recall │ 0.834126889705658 │ +│ electricity_test_roc_auc │ 0.9122480750083923 │ +│ electricity_test_score │ 0.834126889705658 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 107.30it/s +Test metrics: [{'electricity_test_loss': 0.37350720167160034, 'lr': 0.0, 'electricity_test_accuracy': 0.834126889705658, 'electricity_test_precision': 0.8336272835731506, 'electricity_test_recall': 0.834126889705658, 'electricity_test_f1-score': 0.8335598111152649, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.9122480750083923, 'electricity_test_score': 0.834126889705658}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpj6fuqx4a/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpj6fuqx4a/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpj6fuqx4a/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpj6fuqx4a/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ electricity │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─EncoderDCNv2: 2-2 [128, 8] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 8] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 8] [128, 128] 1,152 -- 147,456 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 120,969 +Trainable params: 120,969 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 15.48 +========================================================================================================================================================================== +Input size (MB): 0.26 +Forward/backward pass size (MB): 1.06 +Params size (MB): 0.48 +Estimated Total Size (MB): 1.80 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 120 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 120 K +Non-trainable params: 0 +Total params: 120 K +Total estimated model params size (MB): 0.484 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:02 • 0:00:00 113.94it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.8345355987548828 │ +│ electricity_val_f1-score │ 0.8340221047401428 │ +│ electricity_val_loss │ 0.3726440668106079 │ +│ electricity_val_precision │ 0.8340369462966919 │ +│ electricity_val_recall │ 0.8345355987548828 │ +│ electricity_val_roc_auc │ 0.9129883646965027 │ +│ electricity_val_score │ 0.8345355987548828 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 101.13it/s +Validation metrics: [{'electricity_val_loss': 0.3726440668106079, 'lr': 0.0, 'electricity_val_accuracy': 0.8345355987548828, 'electricity_val_precision': 0.8340369462966919, 'electricity_val_recall': 0.8345355987548828, 'electricity_val_f1-score': 0.8340221047401428, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.9129883646965027, 'electricity_val_score': 0.8345355987548828}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3721 samples (41.20%) + Class 1: 5310 samples (58.80%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.8334625363349915 │ +│ electricity_test_f1-score │ 0.8331384062767029 │ +│ electricity_test_loss │ 0.37364310026168823 │ +│ electricity_test_precision │ 0.8330397605895996 │ +│ electricity_test_recall │ 0.8334625363349915 │ +│ electricity_test_roc_auc │ 0.913081705570221 │ +│ electricity_test_score │ 0.8334625363349915 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 102.21it/s +Test metrics: [{'electricity_test_loss': 0.37364310026168823, 'lr': 0.0, 'electricity_test_accuracy': 0.8334625363349915, 'electricity_test_precision': 0.8330397605895996, 'electricity_test_recall': 0.8334625363349915, 'electricity_test_f1-score': 0.8331384062767029, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.913081705570221, 'electricity_test_score': 0.8334625363349915}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpj6fuqx4a/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpj6fuqx4a/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpj6fuqx4a/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpj6fuqx4a/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ electricity │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─EncoderDCNv2: 2-2 [128, 8] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 8] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 8] [128, 128] 1,152 -- 147,456 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 120,969 +Trainable params: 120,969 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 15.48 +========================================================================================================================================================================== +Input size (MB): 0.26 +Forward/backward pass size (MB): 1.06 +Params size (MB): 0.48 +Estimated Total Size (MB): 1.80 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 120 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 120 K +Non-trainable params: 0 +Total params: 120 K +Total estimated model params size (MB): 0.484 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 155.66it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.8279982209205627 │ +│ electricity_val_f1-score │ 0.8272846937179565 │ +│ electricity_val_loss │ 0.3699244558811188 │ +│ electricity_val_precision │ 0.8274524807929993 │ +│ electricity_val_recall │ 0.8279982209205627 │ +│ electricity_val_roc_auc │ 0.912844717502594 │ +│ electricity_val_score │ 0.8279982209205627 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 362.20it/s +Validation metrics: [{'electricity_val_loss': 0.3699244558811188, 'lr': 0.0, 'electricity_val_accuracy': 0.8279982209205627, 'electricity_val_precision': 0.8274524807929993, 'electricity_val_recall': 0.8279982209205627, 'electricity_val_f1-score': 0.8272846937179565, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.912844717502594, 'electricity_val_score': 0.8279982209205627}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3735 samples (41.36%) + Class 1: 5296 samples (58.64%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.8370058536529541 │ +│ electricity_test_f1-score │ 0.8367340564727783 │ +│ electricity_test_loss │ 0.3732665479183197 │ +│ electricity_test_precision │ 0.8366317749023438 │ +│ electricity_test_recall │ 0.8370058536529541 │ +│ electricity_test_roc_auc │ 0.9129190444946289 │ +│ electricity_test_score │ 0.8370058536529541 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 388.83it/s +Test metrics: [{'electricity_test_loss': 0.3732665479183197, 'lr': 0.0, 'electricity_test_accuracy': 0.8370058536529541, 'electricity_test_precision': 0.8366317749023438, 'electricity_test_recall': 0.8370058536529541, 'electricity_test_f1-score': 0.8367340564727783, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.9129190444946289, 'electricity_test_score': 0.8370058536529541}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpj6fuqx4a/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpj6fuqx4a/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpj6fuqx4a/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpj6fuqx4a/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ electricity │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─EncoderDCNv2: 2-2 [128, 8] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 8] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 8] [128, 128] 1,152 -- 147,456 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 120,969 +Trainable params: 120,969 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 15.48 +========================================================================================================================================================================== +Input size (MB): 0.26 +Forward/backward pass size (MB): 1.06 +Params size (MB): 0.48 +Estimated Total Size (MB): 1.80 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 120 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 120 K +Non-trainable params: 0 +Total params: 120 K +Total estimated model params size (MB): 0.484 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 139.10it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.829801619052887 │ +│ electricity_val_f1-score │ 0.8292050361633301 │ +│ electricity_val_loss │ 0.3687976896762848 │ +│ electricity_val_precision │ 0.8292626142501831 │ +│ electricity_val_recall │ 0.829801619052887 │ +│ electricity_val_roc_auc │ 0.9137072563171387 │ +│ electricity_val_score │ 0.829801619052887 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 226.00it/s +Validation metrics: [{'electricity_val_loss': 0.3687976896762848, 'lr': 0.0, 'electricity_val_accuracy': 0.829801619052887, 'electricity_val_precision': 0.8292626142501831, 'electricity_val_recall': 0.829801619052887, 'electricity_val_f1-score': 0.8292050361633301, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.9137072563171387, 'electricity_val_score': 0.829801619052887}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3702 samples (40.99%) + Class 1: 5329 samples (59.01%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.8304728269577026 │ +│ electricity_test_f1-score │ 0.8300773501396179 │ +│ electricity_test_loss │ 0.37557747960090637 │ +│ electricity_test_precision │ 0.8299964070320129 │ +│ electricity_test_recall │ 0.8304728269577026 │ +│ electricity_test_roc_auc │ 0.9117783904075623 │ +│ electricity_test_score │ 0.8304728269577026 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 323.58it/s +Test metrics: [{'electricity_test_loss': 0.37557747960090637, 'lr': 0.0, 'electricity_test_accuracy': 0.8304728269577026, 'electricity_test_precision': 0.8299964070320129, 'electricity_test_recall': 0.8304728269577026, 'electricity_test_f1-score': 0.8300773501396179, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.9117783904075623, 'electricity_test_score': 0.8304728269577026}] +Wrote results for dcnv2 on electricity to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results/dcnv2_electricity_downstream_results.json +Running downstream for dataset 'electricity', model 'resnet'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpcftj6il5/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpcftj6il5/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpcftj6il5/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpcftj6il5/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─ResNet: 2-2 [128, 8] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 8] [128, 128] 1,152 -- 147,456 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 220,681 +Trainable params: 220,681 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 28.25 +========================================================================================================================================================================== +Input size (MB): 0.26 +Forward/backward pass size (MB): 1.98 +Params size (MB): 0.88 +Estimated Total Size (MB): 3.12 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 220 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 220 K +Non-trainable params: 0 +Total params: 220 K +Total estimated model params size (MB): 0.883 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 155.58it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.8151487708091736 │ +│ electricity_val_f1-score │ 0.8148254752159119 │ +│ electricity_val_loss │ 0.4003864824771881 │ +│ electricity_val_precision │ 0.8146806955337524 │ +│ electricity_val_recall │ 0.8151487708091736 │ +│ electricity_val_roc_auc │ 0.8960868716239929 │ +│ electricity_val_score │ 0.8151487708091736 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 324.98it/s +Validation metrics: [{'electricity_val_loss': 0.4003864824771881, 'lr': 0.0, 'electricity_val_accuracy': 0.8151487708091736, 'electricity_val_precision': 0.8146806955337524, 'electricity_val_recall': 0.8151487708091736, 'electricity_val_f1-score': 0.8148254752159119, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.8960868716239929, 'electricity_val_score': 0.8151487708091736}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3791 samples (41.98%) + Class 1: 5240 samples (58.02%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.8135311603546143 │ +│ electricity_test_f1-score │ 0.8134183883666992 │ +│ electricity_test_loss │ 0.40950292348861694 │ +│ electricity_test_precision │ 0.8133290410041809 │ +│ electricity_test_recall │ 0.8135311603546143 │ +│ electricity_test_roc_auc │ 0.8915355801582336 │ +│ electricity_test_score │ 0.8135311603546143 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 311.19it/s +Test metrics: [{'electricity_test_loss': 0.40950292348861694, 'lr': 0.0, 'electricity_test_accuracy': 0.8135311603546143, 'electricity_test_precision': 0.8133290410041809, 'electricity_test_recall': 0.8135311603546143, 'electricity_test_f1-score': 0.8134183883666992, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.8915355801582336, 'electricity_test_score': 0.8135311603546143}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpcftj6il5/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpcftj6il5/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpcftj6il5/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpcftj6il5/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─ResNet: 2-2 [128, 8] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 8] [128, 128] 1,152 -- 147,456 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 220,681 +Trainable params: 220,681 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 28.25 +========================================================================================================================================================================== +Input size (MB): 0.26 +Forward/backward pass size (MB): 1.98 +Params size (MB): 0.88 +Estimated Total Size (MB): 3.12 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 220 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 220 K +Non-trainable params: 0 +Total params: 220 K +Total estimated model params size (MB): 0.883 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 161.06it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.8095130920410156 │ +│ electricity_val_f1-score │ 0.8105146288871765 │ +│ electricity_val_loss │ 0.40759947896003723 │ +│ electricity_val_precision │ 0.8156034350395203 │ +│ electricity_val_recall │ 0.8095130920410156 │ +│ electricity_val_roc_auc │ 0.9005982279777527 │ +│ electricity_val_score │ 0.8095130920410156 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 314.80it/s +Validation metrics: [{'electricity_val_loss': 0.40759947896003723, 'lr': 0.0, 'electricity_val_accuracy': 0.8095130920410156, 'electricity_val_precision': 0.8156034350395203, 'electricity_val_recall': 0.8095130920410156, 'electricity_val_f1-score': 0.8105146288871765, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.9005982279777527, 'electricity_val_score': 0.8095130920410156}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 4297 samples (47.58%) + Class 1: 4734 samples (52.42%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.8124238848686218 │ +│ electricity_test_f1-score │ 0.8134207129478455 │ +│ electricity_test_loss │ 0.41420847177505493 │ +│ electricity_test_precision │ 0.8188508152961731 │ +│ electricity_test_recall │ 0.8124238848686218 │ +│ electricity_test_roc_auc │ 0.8979076147079468 │ +│ electricity_test_score │ 0.8124238848686218 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 331.15it/s +Test metrics: [{'electricity_test_loss': 0.41420847177505493, 'lr': 0.0, 'electricity_test_accuracy': 0.8124238848686218, 'electricity_test_precision': 0.8188508152961731, 'electricity_test_recall': 0.8124238848686218, 'electricity_test_f1-score': 0.8134207129478455, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.8979076147079468, 'electricity_test_score': 0.8124238848686218}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpcftj6il5/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpcftj6il5/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpcftj6il5/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpcftj6il5/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─ResNet: 2-2 [128, 8] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 8] [128, 128] 1,152 -- 147,456 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 220,681 +Trainable params: 220,681 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 28.25 +========================================================================================================================================================================== +Input size (MB): 0.26 +Forward/backward pass size (MB): 1.98 +Params size (MB): 0.88 +Estimated Total Size (MB): 3.12 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 220 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 220 K +Non-trainable params: 0 +Total params: 220 K +Total estimated model params size (MB): 0.883 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 130.92it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.8196573257446289 │ +│ electricity_val_f1-score │ 0.8198521733283997 │ +│ electricity_val_loss │ 0.39245733618736267 │ +│ electricity_val_precision │ 0.8201331496238708 │ +│ electricity_val_recall │ 0.8196573257446289 │ +│ electricity_val_roc_auc │ 0.9018487930297852 │ +│ electricity_val_score │ 0.8196573257446289 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 171.17it/s +Validation metrics: [{'electricity_val_loss': 0.39245733618736267, 'lr': 0.0, 'electricity_val_accuracy': 0.8196573257446289, 'electricity_val_precision': 0.8201331496238708, 'electricity_val_recall': 0.8196573257446289, 'electricity_val_f1-score': 0.8198521733283997, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.9018487930297852, 'electricity_val_score': 0.8196573257446289}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3950 samples (43.74%) + Class 1: 5081 samples (56.26%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.8216144442558289 │ +│ electricity_test_f1-score │ 0.8219646215438843 │ +│ electricity_test_loss │ 0.3950406312942505 │ +│ electricity_test_precision │ 0.8226320147514343 │ +│ electricity_test_recall │ 0.8216144442558289 │ +│ electricity_test_roc_auc │ 0.9012686610221863 │ +│ electricity_test_score │ 0.8216144442558289 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 163.32it/s +Test metrics: [{'electricity_test_loss': 0.3950406312942505, 'lr': 0.0, 'electricity_test_accuracy': 0.8216144442558289, 'electricity_test_precision': 0.8226320147514343, 'electricity_test_recall': 0.8216144442558289, 'electricity_test_f1-score': 0.8219646215438843, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.9012686610221863, 'electricity_test_score': 0.8216144442558289}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpcftj6il5/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpcftj6il5/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpcftj6il5/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpcftj6il5/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─ResNet: 2-2 [128, 8] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 8] [128, 128] 1,152 -- 147,456 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 220,681 +Trainable params: 220,681 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 28.25 +========================================================================================================================================================================== +Input size (MB): 0.26 +Forward/backward pass size (MB): 1.98 +Params size (MB): 0.88 +Estimated Total Size (MB): 3.12 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 220 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 220 K +Non-trainable params: 0 +Total params: 220 K +Total estimated model params size (MB): 0.883 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 155.56it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.808836817741394 │ +│ electricity_val_f1-score │ 0.8085333108901978 │ +│ electricity_val_loss │ 0.39851632714271545 │ +│ electricity_val_precision │ 0.8083773255348206 │ +│ electricity_val_recall │ 0.808836817741394 │ +│ electricity_val_roc_auc │ 0.8958708643913269 │ +│ electricity_val_score │ 0.808836817741394 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 156.48it/s +Validation metrics: [{'electricity_val_loss': 0.39851632714271545, 'lr': 0.0, 'electricity_val_accuracy': 0.808836817741394, 'electricity_val_precision': 0.8083773255348206, 'electricity_val_recall': 0.808836817741394, 'electricity_val_f1-score': 0.8085333108901978, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.8958708643913269, 'electricity_val_score': 0.808836817741394}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3798 samples (42.06%) + Class 1: 5233 samples (57.94%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.8145277500152588 │ +│ electricity_test_f1-score │ 0.8144391179084778 │ +│ electricity_test_loss │ 0.4052407741546631 │ +│ electricity_test_precision │ 0.8143652677536011 │ +│ electricity_test_recall │ 0.8145277500152588 │ +│ electricity_test_roc_auc │ 0.8931828737258911 │ +│ electricity_test_score │ 0.8145277500152588 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 162.19it/s +Test metrics: [{'electricity_test_loss': 0.4052407741546631, 'lr': 0.0, 'electricity_test_accuracy': 0.8145277500152588, 'electricity_test_precision': 0.8143652677536011, 'electricity_test_recall': 0.8145277500152588, 'electricity_test_f1-score': 0.8144391179084778, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.8931828737258911, 'electricity_test_score': 0.8145277500152588}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpcftj6il5/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpcftj6il5/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpcftj6il5/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpcftj6il5/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─ResNet: 2-2 [128, 8] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 8] [128, 128] 1,152 -- 147,456 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================== +Total params: 220,681 +Trainable params: 220,681 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 28.25 +========================================================================================================================================================================== +Input size (MB): 0.26 +Forward/backward pass size (MB): 1.98 +Params size (MB): 0.88 +Estimated Total Size (MB): 3.12 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 220 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 220 K +Non-trainable params: 0 +Total params: 220 K +Total estimated model params size (MB): 0.883 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:01 • 0:00:00 178.46it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.8198827505111694 │ +│ electricity_val_f1-score │ 0.8194862008094788 │ +│ electricity_val_loss │ 0.3846282958984375 │ +│ electricity_val_precision │ 0.8193657398223877 │ +│ electricity_val_recall │ 0.8198827505111694 │ +│ electricity_val_roc_auc │ 0.9040712714195251 │ +│ electricity_val_score │ 0.8198827505111694 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 401.19it/s +Validation metrics: [{'electricity_val_loss': 0.3846282958984375, 'lr': 0.0, 'electricity_val_accuracy': 0.8198827505111694, 'electricity_val_precision': 0.8193657398223877, 'electricity_val_recall': 0.8198827505111694, 'electricity_val_f1-score': 0.8194862008094788, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.9040712714195251, 'electricity_val_score': 0.8198827505111694}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3773 samples (41.78%) + Class 1: 5258 samples (58.22%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.8197320103645325 │ +│ electricity_test_f1-score │ 0.819563090801239 │ +│ electricity_test_loss │ 0.3941192328929901 │ +│ electricity_test_precision │ 0.8194493651390076 │ +│ electricity_test_recall │ 0.8197320103645325 │ +│ electricity_test_roc_auc │ 0.8995412588119507 │ +│ electricity_test_score │ 0.8197320103645325 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 391.50it/s +Test metrics: [{'electricity_test_loss': 0.3941192328929901, 'lr': 0.0, 'electricity_test_accuracy': 0.8197320103645325, 'electricity_test_precision': 0.8194493651390076, 'electricity_test_recall': 0.8197320103645325, 'electricity_test_f1-score': 0.819563090801239, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.8995412588119507, 'electricity_test_score': 0.8197320103645325}] +Wrote results for resnet on electricity to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results/resnet_electricity_downstream_results.json +Running downstream for dataset 'electricity', model 'autoint'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpp3gogwpv/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpp3gogwpv/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpp3gogwpv/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpp3gogwpv/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ electricity │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─EncoderAutoIntBase: 2-2 [128, 8] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 8] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 8] [128, 8, 128] 1,024 -- 131,072 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 8, 8] [256, 8, 8] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 8, 128] [128, 8, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 8, 128] [128, 8, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 8, 8] [256, 8, 8] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 8, 128] [128, 8, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 8, 128] [128, 8, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 8, 8] [256, 8, 8] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 8, 128] [128, 8, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 8, 128] [128, 8, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 1024] [128, 128] 131,200 -- 16,793,600 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +==================================================================================================================================================================================== +Total params: 334,985 +Trainable params: 334,985 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 42.88 +==================================================================================================================================================================================== +Input size (MB): 0.26 +Forward/backward pass size (MB): 16.92 +Params size (MB): 1.34 +Estimated Total Size (MB): 18.52 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 334 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 334 K +Non-trainable params: 0 +Total params: 334 K +Total estimated model params size (MB): 1.340 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:02 • 0:00:00 103.83it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.8225879073143005 │ +│ electricity_val_f1-score │ 0.8221973180770874 │ +│ electricity_val_loss │ 0.38083615899086 │ +│ electricity_val_precision │ 0.8220837712287903 │ +│ electricity_val_recall │ 0.8225879073143005 │ +│ electricity_val_roc_auc │ 0.9075164198875427 │ +│ electricity_val_score │ 0.8225879073143005 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 148.58it/s +Validation metrics: [{'electricity_val_loss': 0.38083615899086, 'lr': 0.0, 'electricity_val_accuracy': 0.8225879073143005, 'electricity_val_precision': 0.8220837712287903, 'electricity_val_recall': 0.8225879073143005, 'electricity_val_f1-score': 0.8221973180770874, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.9075164198875427, 'electricity_val_score': 0.8225879073143005}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3741 samples (41.42%) + Class 1: 5290 samples (58.58%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.826154351234436 │ +│ electricity_test_f1-score │ 0.8258849382400513 │ +│ electricity_test_loss │ 0.3831469416618347 │ +│ electricity_test_precision │ 0.8257611393928528 │ +│ electricity_test_recall │ 0.826154351234436 │ +│ electricity_test_roc_auc │ 0.9055925607681274 │ +│ electricity_test_score │ 0.826154351234436 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 159.81it/s +Test metrics: [{'electricity_test_loss': 0.3831469416618347, 'lr': 0.0, 'electricity_test_accuracy': 0.826154351234436, 'electricity_test_precision': 0.8257611393928528, 'electricity_test_recall': 0.826154351234436, 'electricity_test_f1-score': 0.8258849382400513, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.9055925607681274, 'electricity_test_score': 0.826154351234436}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpp3gogwpv/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpp3gogwpv/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpp3gogwpv/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpp3gogwpv/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ electricity │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─EncoderAutoIntBase: 2-2 [128, 8] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 8] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 8] [128, 8, 128] 1,024 -- 131,072 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 8, 8] [256, 8, 8] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 8, 128] [128, 8, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 8, 128] [128, 8, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 8, 8] [256, 8, 8] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 8, 128] [128, 8, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 8, 128] [128, 8, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 8, 8] [256, 8, 8] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 8, 128] [128, 8, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 8, 128] [128, 8, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 1024] [128, 128] 131,200 -- 16,793,600 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +==================================================================================================================================================================================== +Total params: 334,985 +Trainable params: 334,985 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 42.88 +==================================================================================================================================================================================== +Input size (MB): 0.26 +Forward/backward pass size (MB): 16.92 +Params size (MB): 1.34 +Estimated Total Size (MB): 18.52 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 334 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 334 K +Non-trainable params: 0 +Total params: 334 K +Total estimated model params size (MB): 1.340 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:02 • 0:00:00 94.02it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.8221370577812195 │ +│ electricity_val_f1-score │ 0.8215613961219788 │ +│ electricity_val_loss │ 0.39049676060676575 │ +│ electricity_val_precision │ 0.8215466737747192 │ +│ electricity_val_recall │ 0.8221370577812195 │ +│ electricity_val_roc_auc │ 0.9021830558776855 │ +│ electricity_val_score │ 0.8221370577812195 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 150.29it/s +Validation metrics: [{'electricity_val_loss': 0.39049676060676575, 'lr': 0.0, 'electricity_val_accuracy': 0.8221370577812195, 'electricity_val_precision': 0.8215466737747192, 'electricity_val_recall': 0.8221370577812195, 'electricity_val_f1-score': 0.8215613961219788, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.9021830558776855, 'electricity_val_score': 0.8221370577812195}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3702 samples (40.99%) + Class 1: 5329 samples (59.01%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.8236075639724731 │ +│ electricity_test_f1-score │ 0.8231960535049438 │ +│ electricity_test_loss │ 0.39296814799308777 │ +│ electricity_test_precision │ 0.8230963349342346 │ +│ electricity_test_recall │ 0.8236075639724731 │ +│ electricity_test_roc_auc │ 0.9012018442153931 │ +│ electricity_test_score │ 0.8236075639724731 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 157.51it/s +Test metrics: [{'electricity_test_loss': 0.39296814799308777, 'lr': 0.0, 'electricity_test_accuracy': 0.8236075639724731, 'electricity_test_precision': 0.8230963349342346, 'electricity_test_recall': 0.8236075639724731, 'electricity_test_f1-score': 0.8231960535049438, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.9012018442153931, 'electricity_test_score': 0.8236075639724731}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpp3gogwpv/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpp3gogwpv/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpp3gogwpv/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpp3gogwpv/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ electricity │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─EncoderAutoIntBase: 2-2 [128, 8] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 8] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 8] [128, 8, 128] 1,024 -- 131,072 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 8, 8] [256, 8, 8] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 8, 128] [128, 8, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 8, 128] [128, 8, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 8, 8] [256, 8, 8] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 8, 128] [128, 8, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 8, 128] [128, 8, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 8, 8] [256, 8, 8] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 8, 128] [128, 8, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 8, 128] [128, 8, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 1024] [128, 128] 131,200 -- 16,793,600 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +==================================================================================================================================================================================== +Total params: 334,985 +Trainable params: 334,985 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 42.88 +==================================================================================================================================================================================== +Input size (MB): 0.26 +Forward/backward pass size (MB): 16.92 +Params size (MB): 1.34 +Estimated Total Size (MB): 18.52 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 334 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 334 K +Non-trainable params: 0 +Total params: 334 K +Total estimated model params size (MB): 1.340 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:02 • 0:00:00 105.84it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.8264201879501343 │ +│ electricity_val_f1-score │ 0.8260307908058167 │ +│ electricity_val_loss │ 0.37689918279647827 │ +│ electricity_val_precision │ 0.8259297013282776 │ +│ electricity_val_recall │ 0.8264201879501343 │ +│ electricity_val_roc_auc │ 0.9090912342071533 │ +│ electricity_val_score │ 0.8264201879501343 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 142.92it/s +Validation metrics: [{'electricity_val_loss': 0.37689918279647827, 'lr': 0.0, 'electricity_val_accuracy': 0.8264201879501343, 'electricity_val_precision': 0.8259297013282776, 'electricity_val_recall': 0.8264201879501343, 'electricity_val_f1-score': 0.8260307908058167, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.9090912342071533, 'electricity_val_score': 0.8264201879501343}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3696 samples (40.93%) + Class 1: 5335 samples (59.07%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.8295869827270508 │ +│ electricity_test_f1-score │ 0.8291682600975037 │ +│ electricity_test_loss │ 0.3787509799003601 │ +│ electricity_test_precision │ 0.8290953040122986 │ +│ electricity_test_recall │ 0.8295869827270508 │ +│ electricity_test_roc_auc │ 0.9082685112953186 │ +│ electricity_test_score │ 0.8295869827270508 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 151.52it/s +Test metrics: [{'electricity_test_loss': 0.3787509799003601, 'lr': 0.0, 'electricity_test_accuracy': 0.8295869827270508, 'electricity_test_precision': 0.8290953040122986, 'electricity_test_recall': 0.8295869827270508, 'electricity_test_f1-score': 0.8291682600975037, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.9082685112953186, 'electricity_test_score': 0.8295869827270508}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpp3gogwpv/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpp3gogwpv/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpp3gogwpv/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpp3gogwpv/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ electricity │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─EncoderAutoIntBase: 2-2 [128, 8] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 8] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 8] [128, 8, 128] 1,024 -- 131,072 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 8, 8] [256, 8, 8] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 8, 128] [128, 8, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 8, 128] [128, 8, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 8, 8] [256, 8, 8] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 8, 128] [128, 8, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 8, 128] [128, 8, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 8, 8] [256, 8, 8] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 8, 128] [128, 8, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 8, 128] [128, 8, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 1024] [128, 128] 131,200 -- 16,793,600 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +==================================================================================================================================================================================== +Total params: 334,985 +Trainable params: 334,985 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 42.88 +==================================================================================================================================================================================== +Input size (MB): 0.26 +Forward/backward pass size (MB): 16.92 +Params size (MB): 1.34 +Estimated Total Size (MB): 18.52 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 334 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 334 K +Non-trainable params: 0 +Total params: 334 K +Total estimated model params size (MB): 1.340 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:02 • 0:00:00 97.75it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.8155996203422546 │ +│ electricity_val_f1-score │ 0.8152470588684082 │ +│ electricity_val_loss │ 0.39570626616477966 │ +│ electricity_val_precision │ 0.8151055574417114 │ +│ electricity_val_recall │ 0.8155996203422546 │ +│ electricity_val_roc_auc │ 0.8984375596046448 │ +│ electricity_val_score │ 0.8155996203422546 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 144.81it/s +Validation metrics: [{'electricity_val_loss': 0.39570626616477966, 'lr': 0.0, 'electricity_val_accuracy': 0.8155996203422546, 'electricity_val_precision': 0.8151055574417114, 'electricity_val_recall': 0.8155996203422546, 'electricity_val_f1-score': 0.8152470588684082, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.8984375596046448, 'electricity_val_score': 0.8155996203422546}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3669 samples (40.63%) + Class 1: 5362 samples (59.37%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.8150813579559326 │ +│ electricity_test_f1-score │ 0.8145215511322021 │ +│ electricity_test_loss │ 0.3995775878429413 │ +│ electricity_test_precision │ 0.8144586682319641 │ +│ electricity_test_recall │ 0.8150813579559326 │ +│ electricity_test_roc_auc │ 0.8969169855117798 │ +│ electricity_test_score │ 0.8150813579559326 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 203.72it/s +Test metrics: [{'electricity_test_loss': 0.3995775878429413, 'lr': 0.0, 'electricity_test_accuracy': 0.8150813579559326, 'electricity_test_precision': 0.8144586682319641, 'electricity_test_recall': 0.8150813579559326, 'electricity_test_f1-score': 0.8145215511322021, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.8969169855117798, 'electricity_test_score': 0.8150813579559326}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpp3gogwpv/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpp3gogwpv/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpp3gogwpv/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpp3gogwpv/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ electricity │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─EncoderAutoIntBase: 2-2 [128, 8] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 8] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 8] [128, 8, 128] 1,024 -- 131,072 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 8, 8] [256, 8, 8] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 8, 128] [128, 8, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 8, 128] [128, 8, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 8, 8] [256, 8, 8] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 8, 128] [128, 8, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 8, 128] [128, 8, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 8, 128] [128, 8, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 8, 8] [256, 8, 8] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 8, 128] [128, 8, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 8, 128] [128, 8, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 8, 128] [128, 8, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 1024] [128, 128] 131,200 -- 16,793,600 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +==================================================================================================================================================================================== +Total params: 334,985 +Trainable params: 334,985 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 42.88 +==================================================================================================================================================================================== +Input size (MB): 0.26 +Forward/backward pass size (MB): 16.92 +Params size (MB): 1.34 +Estimated Total Size (MB): 18.52 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 334 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 334 K +Non-trainable params: 0 +Total params: 334 K +Total estimated model params size (MB): 1.340 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:02 • 0:00:00 110.99it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.8162758946418762 │ +│ electricity_val_f1-score │ 0.8146142959594727 │ +│ electricity_val_loss │ 0.39763420820236206 │ +│ electricity_val_precision │ 0.816234827041626 │ +│ electricity_val_recall │ 0.8162758946418762 │ +│ electricity_val_roc_auc │ 0.8983796834945679 │ +│ electricity_val_score │ 0.8162758946418762 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 298.63it/s +Validation metrics: [{'electricity_val_loss': 0.39763420820236206, 'lr': 0.0, 'electricity_val_accuracy': 0.8162758946418762, 'electricity_val_precision': 0.816234827041626, 'electricity_val_recall': 0.8162758946418762, 'electricity_val_f1-score': 0.8146142959594727, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.8983796834945679, 'electricity_val_score': 0.8162758946418762}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3471 samples (38.43%) + Class 1: 5560 samples (61.57%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.8195105791091919 │ +│ electricity_test_f1-score │ 0.8180992603302002 │ +│ electricity_test_loss │ 0.399333119392395 │ +│ electricity_test_precision │ 0.8192992210388184 │ +│ electricity_test_recall │ 0.8195105791091919 │ +│ electricity_test_roc_auc │ 0.8965386748313904 │ +│ electricity_test_score │ 0.8195105791091919 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 310.87it/s +Test metrics: [{'electricity_test_loss': 0.399333119392395, 'lr': 0.0, 'electricity_test_accuracy': 0.8195105791091919, 'electricity_test_precision': 0.8192992210388184, 'electricity_test_recall': 0.8195105791091919, 'electricity_test_f1-score': 0.8180992603302002, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.8965386748313904, 'electricity_test_score': 0.8195105791091919}] +Wrote results for autoint on electricity to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results/autoint_electricity_downstream_results.json +Running downstream for dataset 'electricity', model 'fttransformer'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpv0altbwp/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpv0altbwp/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpv0altbwp/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpv0altbwp/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ electricity │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─EncoderFTTransformer: 2-2 [128, 8] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 8] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 8] [128, 8, 128] 2,048 -- 262,144 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 9, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 9, 9] [512, 9, 9] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 9, 128] [128, 9, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 9, 512] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 9, 256] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 9, 256] [128, 9, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 9, 9] [512, 9, 9] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 9, 128] [128, 9, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 9, 512] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 9, 256] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 9, 256] [128, 9, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 9] [512, 1, 9] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 9, 128] [128, 9, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 9, 512] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 9, 256] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 9, 256] [128, 9, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================================= +Total params: 519,433 +Trainable params: 519,433 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 66.49 +========================================================================================================================================================================================= +Input size (MB): 0.26 +Forward/backward pass size (MB): 37.10 +Params size (MB): 2.08 +Estimated Total Size (MB): 39.44 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 519 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 519 K +Non-trainable params: 0 +Total params: 519 K +Total estimated model params size (MB): 2.078 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:03 • 0:00:00 67.81it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.8032010793685913 │ +│ electricity_val_f1-score │ 0.8018348813056946 │ +│ electricity_val_loss │ 0.4253382086753845 │ +│ electricity_val_precision │ 0.8025455474853516 │ +│ electricity_val_recall │ 0.8032010793685913 │ +│ electricity_val_roc_auc │ 0.8801766633987427 │ +│ electricity_val_score │ 0.8032010793685913 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 103.72it/s +Validation metrics: [{'electricity_val_loss': 0.4253382086753845, 'lr': 0.0, 'electricity_val_accuracy': 0.8032010793685913, 'electricity_val_precision': 0.8025455474853516, 'electricity_val_recall': 0.8032010793685913, 'electricity_val_f1-score': 0.8018348813056946, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.8801766633987427, 'electricity_val_score': 0.8032010793685913}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3496 samples (38.71%) + Class 1: 5535 samples (61.29%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.8034547567367554 │ +│ electricity_test_f1-score │ 0.8020485639572144 │ +│ electricity_test_loss │ 0.4293154776096344 │ +│ electricity_test_precision │ 0.802844762802124 │ +│ electricity_test_recall │ 0.8034547567367554 │ +│ electricity_test_roc_auc │ 0.8774914741516113 │ +│ electricity_test_score │ 0.8034547567367554 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 103.58it/s +Test metrics: [{'electricity_test_loss': 0.4293154776096344, 'lr': 0.0, 'electricity_test_accuracy': 0.8034547567367554, 'electricity_test_precision': 0.802844762802124, 'electricity_test_recall': 0.8034547567367554, 'electricity_test_f1-score': 0.8020485639572144, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.8774914741516113, 'electricity_test_score': 0.8034547567367554}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpv0altbwp/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpv0altbwp/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpv0altbwp/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpv0altbwp/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ electricity │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─EncoderFTTransformer: 2-2 [128, 8] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 8] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 8] [128, 8, 128] 2,048 -- 262,144 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 9, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 9, 9] [512, 9, 9] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 9, 128] [128, 9, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 9, 512] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 9, 256] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 9, 256] [128, 9, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 9, 9] [512, 9, 9] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 9, 128] [128, 9, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 9, 512] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 9, 256] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 9, 256] [128, 9, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 9] [512, 1, 9] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 9, 128] [128, 9, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 9, 512] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 9, 256] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 9, 256] [128, 9, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================================= +Total params: 519,433 +Trainable params: 519,433 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 66.49 +========================================================================================================================================================================================= +Input size (MB): 0.26 +Forward/backward pass size (MB): 37.10 +Params size (MB): 2.08 +Estimated Total Size (MB): 39.44 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 519 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 519 K +Non-trainable params: 0 +Total params: 519 K +Total estimated model params size (MB): 2.078 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:02 • 0:00:00 87.49it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.8011722564697266 │ +│ electricity_val_f1-score │ 0.7997601628303528 │ +│ electricity_val_loss │ 0.4262220859527588 │ +│ electricity_val_precision │ 0.8005000352859497 │ +│ electricity_val_recall │ 0.8011722564697266 │ +│ electricity_val_roc_auc │ 0.8806599974632263 │ +│ electricity_val_score │ 0.8011722564697266 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 237.11it/s +Validation metrics: [{'electricity_val_loss': 0.4262220859527588, 'lr': 0.0, 'electricity_val_accuracy': 0.8011722564697266, 'electricity_val_precision': 0.8005000352859497, 'electricity_val_recall': 0.8011722564697266, 'electricity_val_f1-score': 0.7997601628303528, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.8806599974632263, 'electricity_val_score': 0.8011722564697266}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3510 samples (38.87%) + Class 1: 5521 samples (61.13%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.7970324158668518 │ +│ electricity_test_f1-score │ 0.7956543564796448 │ +│ electricity_test_loss │ 0.4318920969963074 │ +│ electricity_test_precision │ 0.7962751984596252 │ +│ electricity_test_recall │ 0.7970324158668518 │ +│ electricity_test_roc_auc │ 0.8763056397438049 │ +│ electricity_test_score │ 0.7970324158668518 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 250.66it/s +Test metrics: [{'electricity_test_loss': 0.4318920969963074, 'lr': 0.0, 'electricity_test_accuracy': 0.7970324158668518, 'electricity_test_precision': 0.7962751984596252, 'electricity_test_recall': 0.7970324158668518, 'electricity_test_f1-score': 0.7956543564796448, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.8763056397438049, 'electricity_test_score': 0.7970324158668518}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpv0altbwp/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpv0altbwp/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpv0altbwp/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpv0altbwp/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ electricity │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─EncoderFTTransformer: 2-2 [128, 8] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 8] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 8] [128, 8, 128] 2,048 -- 262,144 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 9, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 9, 9] [512, 9, 9] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 9, 128] [128, 9, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 9, 512] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 9, 256] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 9, 256] [128, 9, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 9, 9] [512, 9, 9] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 9, 128] [128, 9, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 9, 512] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 9, 256] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 9, 256] [128, 9, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 9] [512, 1, 9] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 9, 128] [128, 9, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 9, 512] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 9, 256] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 9, 256] [128, 9, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================================= +Total params: 519,433 +Trainable params: 519,433 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 66.49 +========================================================================================================================================================================================= +Input size (MB): 0.26 +Forward/backward pass size (MB): 37.10 +Params size (MB): 2.08 +Estimated Total Size (MB): 39.44 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 519 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 519 K +Non-trainable params: 0 +Total params: 519 K +Total estimated model params size (MB): 2.078 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:03 • 0:00:00 67.25it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.798692524433136 │ +│ electricity_val_f1-score │ 0.7970987558364868 │ +│ electricity_val_loss │ 0.42772355675697327 │ +│ electricity_val_precision │ 0.7980766892433167 │ +│ electricity_val_recall │ 0.798692524433136 │ +│ electricity_val_roc_auc │ 0.8773455023765564 │ +│ electricity_val_score │ 0.798692524433136 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 111.83it/s +Validation metrics: [{'electricity_val_loss': 0.42772355675697327, 'lr': 0.0, 'electricity_val_accuracy': 0.798692524433136, 'electricity_val_precision': 0.7980766892433167, 'electricity_val_recall': 0.798692524433136, 'electricity_val_f1-score': 0.7970987558364868, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.8773455023765564, 'electricity_val_score': 0.798692524433136}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3423 samples (37.90%) + Class 1: 5608 samples (62.10%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.7938212752342224 │ +│ electricity_test_f1-score │ 0.7919358015060425 │ +│ electricity_test_loss │ 0.43564078211784363 │ +│ electricity_test_precision │ 0.7933134436607361 │ +│ electricity_test_recall │ 0.7938212752342224 │ +│ electricity_test_roc_auc │ 0.8736765384674072 │ +│ electricity_test_score │ 0.7938212752342224 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 104.41it/s +Test metrics: [{'electricity_test_loss': 0.43564078211784363, 'lr': 0.0, 'electricity_test_accuracy': 0.7938212752342224, 'electricity_test_precision': 0.7933134436607361, 'electricity_test_recall': 0.7938212752342224, 'electricity_test_f1-score': 0.7919358015060425, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.8736765384674072, 'electricity_test_score': 0.7938212752342224}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpv0altbwp/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpv0altbwp/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpv0altbwp/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpv0altbwp/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ electricity │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─EncoderFTTransformer: 2-2 [128, 8] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 8] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 8] [128, 8, 128] 2,048 -- 262,144 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 9, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 9, 9] [512, 9, 9] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 9, 128] [128, 9, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 9, 512] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 9, 256] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 9, 256] [128, 9, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 9, 9] [512, 9, 9] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 9, 128] [128, 9, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 9, 512] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 9, 256] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 9, 256] [128, 9, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 9] [512, 1, 9] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 9, 128] [128, 9, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 9, 512] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 9, 256] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 9, 256] [128, 9, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================================= +Total params: 519,433 +Trainable params: 519,433 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 66.49 +========================================================================================================================================================================================= +Input size (MB): 0.26 +Forward/backward pass size (MB): 37.10 +Params size (MB): 2.08 +Estimated Total Size (MB): 39.44 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 519 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 519 K +Non-trainable params: 0 +Total params: 519 K +Total estimated model params size (MB): 2.078 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:02 • 0:00:00 87.97it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.7984670996665955 │ +│ electricity_val_f1-score │ 0.7958927154541016 │ +│ electricity_val_loss │ 0.43353602290153503 │ +│ electricity_val_precision │ 0.7988982200622559 │ +│ electricity_val_recall │ 0.7984670996665955 │ +│ electricity_val_roc_auc │ 0.8743888139724731 │ +│ electricity_val_score │ 0.7984670996665955 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 238.35it/s +Validation metrics: [{'electricity_val_loss': 0.43353602290153503, 'lr': 0.0, 'electricity_val_accuracy': 0.7984670996665955, 'electricity_val_precision': 0.7988982200622559, 'electricity_val_recall': 0.7984670996665955, 'electricity_val_f1-score': 0.7958927154541016, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.8743888139724731, 'electricity_val_score': 0.7984670996665955}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3297 samples (36.51%) + Class 1: 5734 samples (63.49%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.7916066646575928 │ +│ electricity_test_f1-score │ 0.7889108061790466 │ +│ electricity_test_loss │ 0.4379521310329437 │ +│ electricity_test_precision │ 0.7918879985809326 │ +│ electricity_test_recall │ 0.7916066646575928 │ +│ electricity_test_roc_auc │ 0.8720296621322632 │ +│ electricity_test_score │ 0.7916066646575928 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 214.72it/s +Test metrics: [{'electricity_test_loss': 0.4379521310329437, 'lr': 0.0, 'electricity_test_accuracy': 0.7916066646575928, 'electricity_test_precision': 0.7918879985809326, 'electricity_test_recall': 0.7916066646575928, 'electricity_test_f1-score': 0.7889108061790466, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.8720296621322632, 'electricity_test_score': 0.7916066646575928}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=electricity --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpv0altbwp/split_data/split_config.json --sigma_aggregation=mean --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpv0altbwp/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ electricity │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpv0altbwp/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpv0altbwp/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 8 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 249 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 1 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 8, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 8 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 12450 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ electricity │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ binary_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 8, 64] [128, 1] -- -- -- +├─Sequential: 1-1 [128, 8, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 8, 64] [128, 8] -- -- -- +│ │ └─Linear: 3-1 [128, 512] [128, 8] 4,104 -- 525,312 +│ └─EncoderFTTransformer: 2-2 [128, 8] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 8] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 8] [128, 8, 128] 2,048 -- 262,144 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 9, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 9, 9] [512, 9, 9] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 9, 128] [128, 9, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 9, 512] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 9, 256] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 9, 256] [128, 9, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 9, 9] [512, 9, 9] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 9, 128] [128, 9, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 9, 512] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 9, 256] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 9, 256] [128, 9, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 9, 128] [128, 9, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 9] [512, 1, 9] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 9, 128] [128, 9, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 9, 128] [128, 9, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 9, 512] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 9, 256] [128, 9, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 9, 256] [128, 9, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 9, 128] [128, 9, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 1] 129 -- 16,512 +========================================================================================================================================================================================= +Total params: 519,433 +Trainable params: 519,433 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 66.49 +========================================================================================================================================================================================= +Input size (MB): 0.26 +Forward/backward pass size (MB): 37.10 +Params size (MB): 2.08 +Estimated Total Size (MB): 39.44 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 129 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 519 K │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 519 K +Non-trainable params: 0 +Total params: 519 K +Total estimated model params size (MB): 2.078 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249/249 0:00:03 • 0:00:00 72.05it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_val_accuracy │ 0.8237150311470032 │ +│ electricity_val_f1-score │ 0.822908878326416 │ +│ electricity_val_loss │ 0.3902754783630371 │ +│ electricity_val_precision │ 0.823143482208252 │ +│ electricity_val_recall │ 0.8237150311470032 │ +│ electricity_val_roc_auc │ 0.9009010195732117 │ +│ electricity_val_score │ 0.8237150311470032 │ +│ electricity_val_support │ 4436.0 │ +│ lr │ 0.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 35/35 0:00:00 • 0:00:00 106.12it/s +Validation metrics: [{'electricity_val_loss': 0.3902754783630371, 'lr': 0.0, 'electricity_val_accuracy': 0.8237150311470032, 'electricity_val_precision': 0.823143482208252, 'electricity_val_recall': 0.8237150311470032, 'electricity_val_f1-score': 0.822908878326416, 'electricity_val_support': 4436.0, 'electricity_val_roc_auc': 0.9009010195732117, 'electricity_val_score': 0.8237150311470032}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [3] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 3825 samples (42.35%) + Class 1.0: 5206 samples (57.65%) +Predicted class frequencies: + Class 0: 3595 samples (39.81%) + Class 1: 5436 samples (60.19%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ electricity_test_accuracy │ 0.8150813579559326 │ +│ electricity_test_f1-score │ 0.8142136931419373 │ +│ electricity_test_loss │ 0.3970939815044403 │ +│ electricity_test_precision │ 0.8144314885139465 │ +│ electricity_test_recall │ 0.8150813579559326 │ +│ electricity_test_roc_auc │ 0.8975225687026978 │ +│ electricity_test_score │ 0.8150813579559326 │ +│ electricity_test_support │ 9031.0 │ +│ lr │ 0.0 │ +└────────────────────────────┴────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71/71 0:00:00 • 0:00:00 107.12it/s +Test metrics: [{'electricity_test_loss': 0.3970939815044403, 'lr': 0.0, 'electricity_test_accuracy': 0.8150813579559326, 'electricity_test_precision': 0.8144314885139465, 'electricity_test_recall': 0.8150813579559326, 'electricity_test_f1-score': 0.8142136931419373, 'electricity_test_support': 9031.0, 'electricity_test_roc_auc': 0.8975225687026978, 'electricity_test_score': 0.8150813579559326}] +Wrote results for fttransformer on electricity to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results/fttransformer_electricity_downstream_results.json + +Downstream pipeline complete. Results under /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_182946/downstream_results diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/sim-direct-gpu2.log b/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/sim-direct-gpu2.log new file mode 100644 index 0000000000000000000000000000000000000000..a6f6626219c098cfa22c525485ea82876449d2e7 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/logs/sim-direct-gpu2.log @@ -0,0 +1,17495 @@ ++ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py --dataset sim --results-root /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42 --epochs 500 --objective w_jepa_joint --latent-architecture deterministic_no_z --sw-full-end-weight 0.00025 --torch-seed 42 --np-seed 42 --mask-seed 0 --sw-seed 0 --sw-full-num-projections 2048 --sw-marginal-num-projections 1024 --sw-p 2 --sw-s-x-end-weight 5e-05 --sw-z-end-weight 0.0 --sw-s-y-end-weight 1e-05 --model-selection unfiltered --deterministic-mode warn --sw-include-marginals --batch-size 512 +{ + "dataset": "sim", + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "sim", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "64", + "--model_num_layers", + "16", + "--model_num_heads", + "2", + "--model_dropout_prob", + "0.002", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.15", + "--mask_max_ctx_share", + "0.5", + "--mask_min_trgt_share", + "0.15", + "--mask_max_trgt_share", + "0.8", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.0005", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "1e-06", + "--vtjepa_kl_x_annealing_epochs", + "50", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "50", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "50", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.001", + "--recon_y_weight", + "0.1", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "p90", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.00025", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "50", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "5e-05", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "1e-05", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} +Not SLURM job. +╒══════════════════════════════════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ mock │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ activation_checkpointing │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 42 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 42 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_seed │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_seed │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ deterministic_mode │ warn │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pre_evaluation │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 512 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 500 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ improvement │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_val_cache_cadence │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ init_type │ trunc_normal │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_start_lr │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_lr │ 0.0005 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_final_lr │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 10 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_final_weight_decay │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_ipe_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_gradient_clipping │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_feature_type_embedding │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_feature_index_embedding │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dropout_prob │ 0.002 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_layer_norm_eps │ 1e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dim_hidden │ 64 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dim_feedforward │ 64 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_num_heads │ 2 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_num_layers │ 16 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_act_func │ relu │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ probe_cadence │ 2 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ probe_model │ linear_probe │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ n_cls_tokens │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_type │ transformer │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_num_layers │ 16 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_embed_dim │ 16 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_num_heads │ 4 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_p_dropout │ 0.002 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_layer_norm_eps │ 1e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_activation │ relu │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pred_dim_feedforward │ 256 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_allow_overlap │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_min_ctx_share │ 0.15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_max_ctx_share │ 0.5 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_min_trgt_share │ 0.15 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_max_trgt_share │ 0.8 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_num_preds │ 4 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mask_num_encs │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mu_init_std │ 0.03 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ logvar_init_std │ 0.005 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ logvar_init_bias │ -4.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_sigma_x_init │ -2.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_sigma_y_init │ -2.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ downstream_evaluation_epochs │ 50 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ downstream_probe_patience │ 20 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ probe_representation_type │ mean │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ probe_encoder_type │ target_posterior │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pooled_context_representation_size_for_target_posterior │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pooled_context_representation_size_for_auxiliary_encoder │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_end_weight │ 1e-06 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_end_weight │ 1e-06 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_end_weight │ 1e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_annealing_epochs │ 50 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_annealing_epochs │ 50 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_annealing_epochs │ 50 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_x_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_z_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ vtjepa_kl_y_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ recon_x_weight │ 0.001 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ recon_y_weight │ 0.1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ latent_architecture │ deterministic_no_z │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ jepa_objective │ w_jepa_joint │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_full_num_projections │ 2048 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_marginal_num_projections │ 1024 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_p │ 2 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_include_marginals │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_s_x_weight │ 5e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_z_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sw_s_y_weight │ 1e-05 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_start_weight │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_end_weight │ 0.00025 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_start_epoch │ 1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_joint_annealing_epochs │ 50 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_marginal_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_marginal_end_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_end_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_start_epoch │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_x_annealing_epochs │ 30 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_end_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_start_epoch │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_z_annealing_epochs │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_start_scale │ 0.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_end_scale │ 1.0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_start_epoch │ 0 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ wjepa_s_y_annealing_epochs │ 30 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ safe_exit │ False │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_selection │ unfiltered │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ sigma_aggregation │ p90 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ subset_size │ -1 │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ is_batchlearning │ True │ +├──────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ iteration │ 0 │ +╘══════════════════════════════════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ + +CUDA Device Information: +CUDA_VISIBLE_DEVICES: 2 +Number of CUDA devices: 1 +Current CUDA device: 0 +CUDA device name: NVIDIA A100-SXM4-80GB + +Using device: cuda:0 +Device name: NVIDIA A100-SXM4-80GB +Encoder Module parameters: ++-------------------+--------------------------------------------------------------------------------------------------------+ +| Parameter | Value | ++===================+========================================================================================================+ +| hidden_dim | 64 | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| idx_num_features | [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27] | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| idx_cat_features | [28, 29, 30, 31] | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| cardinalities | [(28, 4), (29, 4), (30, 4), (31, 4)] | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| n_input_features | 32 | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| device | cuda:0 | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| num_layers | 16 | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| num_heads | 2 | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| p_dropout | 0.002 | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| layer_norm_eps | 1e-05 | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| gradient_clipping | 1.0 | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| dim_feedforward | 64 | ++-------------------+--------------------------------------------------------------------------------------------------------+ +Encoder Module parameters: ++-------------------+--------------------------------------------------------------------------------------------------------+ +| Parameter | Value | ++===================+========================================================================================================+ +| hidden_dim | 64 | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| idx_num_features | [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27] | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| idx_cat_features | [28, 29, 30, 31] | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| cardinalities | [(28, 4), (29, 4), (30, 4), (31, 4)] | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| n_input_features | 32 | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| device | cuda:0 | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| num_layers | 16 | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| num_heads | 2 | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| p_dropout | 0.002 | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| layer_norm_eps | 1e-05 | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| gradient_clipping | 1.0 | ++-------------------+--------------------------------------------------------------------------------------------------------+ +| dim_feedforward | 64 | ++-------------------+--------------------------------------------------------------------------------------------------------+ +DJEPA parameters: ++-------------------------+---------+ +| Parameter | Value | ++=========================+=========+ +| hidden_dim | 64 | ++-------------------------+---------+ +| num_layers | 16 | ++-------------------------+---------+ +| num_heads | 2 | ++-------------------------+---------+ +| p_dropout | 0.002 | ++-------------------------+---------+ +| layer_norm_eps | 1e-05 | ++-------------------------+---------+ +| feature_type_embedding | True | ++-------------------------+---------+ +| feature_index_embedding | True | ++-------------------------+---------+ +| dim_feedforward | 64 | ++-------------------------+---------+ +| use_reg_token | True | ++-------------------------+---------+ +| gradient_clipping | 1.0 | ++-------------------------+---------+ +| activation | gelu | ++-------------------------+---------+ +| num_features | 32 | ++-------------------------+---------+ +| n_cls_tokens | 1 | ++-------------------------+---------+ +MaskCollator params: +----------------- ------------------------------------ +Allow overlap False +Min context share 0.15 +Max context share 0.5 +Min target share 0.15 +Max target share 0.8 +Min context 4 +Max context 16 +Min target 4 +Max target 25 +Num preds 4 +Num encs 1 +Num features 32 +Cardinalities [(28, 4), (29, 4), (30, 4), (31, 4)] +Base mask seed 0 +----------------- ------------------------------------ +KL-x annealing: start=0.0 for 20 steps, then anneal to 1e-06 over 1000 steps. +KL-z annealing: start=0.0 for 20 steps, then anneal to 1e-06 over 1000 steps. +KL-y annealing: start=0.0 for 20 steps, then anneal to 1e-05 over 1000 steps. +W-JEPA marginal-scale annealing (s_x): start=0.0, end=1.0, start_epoch=0, anneal_epochs=30. +W-JEPA marginal-scale annealing (z): start=0.0, end=0.0, start_epoch=0, anneal_epochs=0. +W-JEPA marginal-scale annealing (s_y): start=0.0, end=1.0, start_epoch=0, anneal_epochs=30. +Configuration saved to: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/vtjepa_config.yaml +All results will be saved to: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549 +[VTJEPA] Running downstream probe before training begins +Using full dataset for embedding generation: sim +Generating embedding using: JEPAEmbeddingEncoder on full dataset + 0%| | 0/20 [00:00= 15) +[VTJEPA] Early stopping triggered at epoch 84 + Best downstream validation score: 0.868108 at epoch 53 + No improvement for 15 probe evaluations +Restored best downstream-selected checkpoint from epoch 53. +Training completed. ++ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -W ignore -m src.downstream_pipeline --exp_path /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549 --num_runs 5 --epochs 50 --models mlp,dcnv2,resnet,autoint,fttransformer --no-filter_uncertainty +[Config] Loaded sigma_aggregation=p90 from /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/vtjepa_config.yaml +[Config] Loaded training data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets from /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/vtjepa_config.yaml +Split statistics: + val: 925 samples + test: 1962 samples + train: 7113 samples +Running downstream for dataset 'sim', model 'mlp'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpd85dycll/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpd85dycll/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpd85dycll/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpd85dycll/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ sim │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─Linear: 2-2 [128, 32] [128, 128] 4,224 -- 540,672 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +===================================================================================================================================================================== +Total params: 86,947 +Trainable params: 86,947 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 11.13 +===================================================================================================================================================================== +Input size (MB): 1.05 +Forward/backward pass size (MB): 0.43 +Params size (MB): 0.35 +Estimated Total Size (MB): 1.83 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 86.6 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 86.9 K +Non-trainable params: 0 +Total params: 86.9 K +Total estimated model params size (MB): 0.348 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=50` reached. +Epoch 49/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 197.39it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0010000000474974513 │ +│ sim_val_accuracy │ 0.8183783888816833 │ +│ sim_val_f1-score │ 0.8181406259536743 │ +│ sim_val_loss │ 0.42959123849868774 │ +│ sim_val_precision │ 0.8403798341751099 │ +│ sim_val_recall │ 0.8183783888816833 │ +│ sim_val_score │ 0.8183783888816833 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 233.85it/s +Validation metrics: [{'sim_val_loss': 0.42959123849868774, 'lr': 0.0010000000474974513, 'sim_val_accuracy': 0.8183783888816833, 'sim_val_precision': 0.8403798341751099, 'sim_val_recall': 0.8183783888816833, 'sim_val_f1-score': 0.8181406259536743, 'sim_val_support': 925.0, 'sim_val_score': 0.8183783888816833}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 898 samples (45.77%) + Class 1: 576 samples (29.36%) + Class 2: 488 samples (24.87%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0010000001639127731 │ +│ sim_test_accuracy │ 0.8216106295585632 │ +│ sim_test_f1-score │ 0.8210008144378662 │ +│ sim_test_loss │ 0.4512270987033844 │ +│ sim_test_precision │ 0.8440932631492615 │ +│ sim_test_recall │ 0.8216106295585632 │ +│ sim_test_score │ 0.8216106295585632 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 442.75it/s +Test metrics: [{'sim_test_loss': 0.4512270987033844, 'lr': 0.0010000001639127731, 'sim_test_accuracy': 0.8216106295585632, 'sim_test_precision': 0.8440932631492615, 'sim_test_recall': 0.8216106295585632, 'sim_test_f1-score': 0.8210008144378662, 'sim_test_support': 1962.0, 'sim_test_score': 0.8216106295585632}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpd85dycll/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpd85dycll/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpd85dycll/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpd85dycll/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ sim │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─Linear: 2-2 [128, 32] [128, 128] 4,224 -- 540,672 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +===================================================================================================================================================================== +Total params: 86,947 +Trainable params: 86,947 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 11.13 +===================================================================================================================================================================== +Input size (MB): 1.05 +Forward/backward pass size (MB): 0.43 +Params size (MB): 0.35 +Estimated Total Size (MB): 1.83 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 86.6 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 86.9 K +Non-trainable params: 0 +Total params: 86.9 K +Total estimated model params size (MB): 0.348 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 39/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 189.33it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0010000000474974513 │ +│ sim_val_accuracy │ 0.8291891813278198 │ +│ sim_val_f1-score │ 0.8309029936790466 │ +│ sim_val_loss │ 0.4242367744445801 │ +│ sim_val_precision │ 0.8540714383125305 │ +│ sim_val_recall │ 0.8291891813278198 │ +│ sim_val_score │ 0.8291891813278198 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 320.02it/s +Validation metrics: [{'sim_val_loss': 0.4242367744445801, 'lr': 0.0010000000474974513, 'sim_val_accuracy': 0.8291891813278198, 'sim_val_precision': 0.8540714383125305, 'sim_val_recall': 0.8291891813278198, 'sim_val_f1-score': 0.8309029936790466, 'sim_val_support': 925.0, 'sim_val_score': 0.8291891813278198}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 617 samples (31.45%) + Class 1: 851 samples (43.37%) + Class 2: 494 samples (25.18%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0010000001639127731 │ +│ sim_test_accuracy │ 0.8287461996078491 │ +│ sim_test_f1-score │ 0.829196572303772 │ +│ sim_test_loss │ 0.42744868993759155 │ +│ sim_test_precision │ 0.8508343696594238 │ +│ sim_test_recall │ 0.8287461996078491 │ +│ sim_test_score │ 0.8287461996078491 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 559.94it/s +Test metrics: [{'sim_test_loss': 0.42744868993759155, 'lr': 0.0010000001639127731, 'sim_test_accuracy': 0.8287461996078491, 'sim_test_precision': 0.8508343696594238, 'sim_test_recall': 0.8287461996078491, 'sim_test_f1-score': 0.829196572303772, 'sim_test_support': 1962.0, 'sim_test_score': 0.8287461996078491}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpd85dycll/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpd85dycll/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpd85dycll/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpd85dycll/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ sim │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─Linear: 2-2 [128, 32] [128, 128] 4,224 -- 540,672 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +===================================================================================================================================================================== +Total params: 86,947 +Trainable params: 86,947 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 11.13 +===================================================================================================================================================================== +Input size (MB): 1.05 +Forward/backward pass size (MB): 0.43 +Params size (MB): 0.35 +Estimated Total Size (MB): 1.83 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 86.6 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 86.9 K +Non-trainable params: 0 +Total params: 86.9 K +Total estimated model params size (MB): 0.348 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 44/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 182.19it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0010000000474974513 │ +│ sim_val_accuracy │ 0.8475675582885742 │ +│ sim_val_f1-score │ 0.8471525311470032 │ +│ sim_val_loss │ 0.3884407877922058 │ +│ sim_val_precision │ 0.8557260632514954 │ +│ sim_val_recall │ 0.8475675582885742 │ +│ sim_val_score │ 0.8475675582885742 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 118.68it/s +Validation metrics: [{'sim_val_loss': 0.3884407877922058, 'lr': 0.0010000000474974513, 'sim_val_accuracy': 0.8475675582885742, 'sim_val_precision': 0.8557260632514954, 'sim_val_recall': 0.8475675582885742, 'sim_val_f1-score': 0.8471525311470032, 'sim_val_support': 925.0, 'sim_val_score': 0.8475675582885742}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 841 samples (42.86%) + Class 1: 562 samples (28.64%) + Class 2: 559 samples (28.49%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0010000001639127731 │ +│ sim_test_accuracy │ 0.8394495248794556 │ +│ sim_test_f1-score │ 0.8390764594078064 │ +│ sim_test_loss │ 0.40132883191108704 │ +│ sim_test_precision │ 0.8511507511138916 │ +│ sim_test_recall │ 0.8394495248794556 │ +│ sim_test_score │ 0.8394495248794556 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 143.14it/s +Test metrics: [{'sim_test_loss': 0.40132883191108704, 'lr': 0.0010000001639127731, 'sim_test_accuracy': 0.8394495248794556, 'sim_test_precision': 0.8511507511138916, 'sim_test_recall': 0.8394495248794556, 'sim_test_f1-score': 0.8390764594078064, 'sim_test_support': 1962.0, 'sim_test_score': 0.8394495248794556}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpd85dycll/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpd85dycll/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpd85dycll/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpd85dycll/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ sim │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─Linear: 2-2 [128, 32] [128, 128] 4,224 -- 540,672 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +===================================================================================================================================================================== +Total params: 86,947 +Trainable params: 86,947 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 11.13 +===================================================================================================================================================================== +Input size (MB): 1.05 +Forward/backward pass size (MB): 0.43 +Params size (MB): 0.35 +Estimated Total Size (MB): 1.83 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 86.6 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 86.9 K +Non-trainable params: 0 +Total params: 86.9 K +Total estimated model params size (MB): 0.348 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 28/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 168.10it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0010000000474974513 │ +│ sim_val_accuracy │ 0.8572973012924194 │ +│ sim_val_f1-score │ 0.8573291897773743 │ +│ sim_val_loss │ 0.3681979179382324 │ +│ sim_val_precision │ 0.8630771040916443 │ +│ sim_val_recall │ 0.8572973012924194 │ +│ sim_val_score │ 0.8572973012924194 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 125.71it/s +Validation metrics: [{'sim_val_loss': 0.3681979179382324, 'lr': 0.0010000000474974513, 'sim_val_accuracy': 0.8572973012924194, 'sim_val_precision': 0.8630771040916443, 'sim_val_recall': 0.8572973012924194, 'sim_val_f1-score': 0.8573291897773743, 'sim_val_support': 925.0, 'sim_val_score': 0.8572973012924194}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 619 samples (31.55%) + Class 1: 632 samples (32.21%) + Class 2: 711 samples (36.24%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0010000001639127731 │ +│ sim_test_accuracy │ 0.860856294631958 │ +│ sim_test_f1-score │ 0.8608131408691406 │ +│ sim_test_loss │ 0.36493369936943054 │ +│ sim_test_precision │ 0.8632360100746155 │ +│ sim_test_recall │ 0.860856294631958 │ +│ sim_test_score │ 0.860856294631958 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 253.91it/s +Test metrics: [{'sim_test_loss': 0.36493369936943054, 'lr': 0.0010000001639127731, 'sim_test_accuracy': 0.860856294631958, 'sim_test_precision': 0.8632360100746155, 'sim_test_recall': 0.860856294631958, 'sim_test_f1-score': 0.8608131408691406, 'sim_test_support': 1962.0, 'sim_test_score': 0.860856294631958}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=mlp --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpd85dycll/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_hidden=2 --hidden_dim=128 --dropout=0.1 --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpd85dycll/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ mlp │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpd85dycll/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpd85dycll/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: mlp +╒══════════════════════╤════════════════╕ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├──────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +╘══════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ emb_dim │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ sim │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +MLP [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─Linear: 2-2 [128, 32] [128, 128] 4,224 -- 540,672 +│ └─Sequential: 2-3 [128, 128] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ └─ReLU: 3-3 [128, 128] [128, 128] -- -- -- +│ │ └─Dropout: 3-4 [128, 128] [128, 128] -- -- -- +│ │ └─BatchNorm1d: 3-5 [128, 128] [128, 128] 256 -- 32,768 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +===================================================================================================================================================================== +Total params: 86,947 +Trainable params: 86,947 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 11.13 +===================================================================================================================================================================== +Input size (MB): 1.05 +Forward/backward pass size (MB): 0.43 +Params size (MB): 0.35 +Estimated Total Size (MB): 1.83 +===================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 86.6 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 86.9 K +Non-trainable params: 0 +Total params: 86.9 K +Total estimated model params size (MB): 0.348 +Modules in train mode: 11 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 27/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 124.98it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0010000000474974513 │ +│ sim_val_accuracy │ 0.8129729628562927 │ +│ sim_val_f1-score │ 0.8110881447792053 │ +│ sim_val_loss │ 0.4813116490840912 │ +│ sim_val_precision │ 0.8407442569732666 │ +│ sim_val_recall │ 0.8129729628562927 │ +│ sim_val_score │ 0.8129729628562927 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 243.14it/s +Validation metrics: [{'sim_val_loss': 0.4813116490840912, 'lr': 0.0010000000474974513, 'sim_val_accuracy': 0.8129729628562927, 'sim_val_precision': 0.8407442569732666, 'sim_val_recall': 0.8129729628562927, 'sim_val_f1-score': 0.8110881447792053, 'sim_val_support': 925.0, 'sim_val_score': 0.8129729628562927}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 715 samples (36.44%) + Class 1: 825 samples (42.05%) + Class 2: 422 samples (21.51%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0010000001639127731 │ +│ sim_test_accuracy │ 0.8047910332679749 │ +│ sim_test_f1-score │ 0.800818681716919 │ +│ sim_test_loss │ 0.4823322296142578 │ +│ sim_test_precision │ 0.827860414981842 │ +│ sim_test_recall │ 0.8047910332679749 │ +│ sim_test_score │ 0.8047910332679749 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 407.23it/s +Test metrics: [{'sim_test_loss': 0.4823322296142578, 'lr': 0.0010000001639127731, 'sim_test_accuracy': 0.8047910332679749, 'sim_test_precision': 0.827860414981842, 'sim_test_recall': 0.8047910332679749, 'sim_test_f1-score': 0.800818681716919, 'sim_test_support': 1962.0, 'sim_test_score': 0.8047910332679749}] +Wrote results for mlp on sim to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results/mlp_sim_downstream_results.json +Running downstream for dataset 'sim', model 'dcnv2'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpsyiazfhq/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpsyiazfhq/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpsyiazfhq/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpsyiazfhq/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2800 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ sim │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─EncoderDCNv2: 2-2 [128, 32] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 32] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 32] [128, 128] 4,224 -- 540,672 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +========================================================================================================================================================================== +Total params: 185,763 +Trainable params: 185,763 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 23.78 +========================================================================================================================================================================== +Input size (MB): 1.05 +Forward/backward pass size (MB): 1.08 +Params size (MB): 0.74 +Estimated Total Size (MB): 2.88 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 185 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 185 K +Non-trainable params: 0 +Total params: 185 K +Total estimated model params size (MB): 0.743 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 29/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 144.69it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0003454915131442249 │ +│ sim_val_accuracy │ 0.8508108258247375 │ +│ sim_val_f1-score │ 0.8509405851364136 │ +│ sim_val_loss │ 0.3876427710056305 │ +│ sim_val_precision │ 0.8522735238075256 │ +│ sim_val_recall │ 0.8508108258247375 │ +│ sim_val_score │ 0.8508108258247375 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 161.40it/s +Validation metrics: [{'sim_val_loss': 0.3876427710056305, 'lr': 0.0003454915131442249, 'sim_val_accuracy': 0.8508108258247375, 'sim_val_precision': 0.8522735238075256, 'sim_val_recall': 0.8508108258247375, 'sim_val_f1-score': 0.8509405851364136, 'sim_val_support': 925.0, 'sim_val_score': 0.8508108258247375}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 659 samples (33.59%) + Class 1: 655 samples (33.38%) + Class 2: 648 samples (33.03%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0003454914840403944 │ +│ sim_test_accuracy │ 0.8516819477081299 │ +│ sim_test_f1-score │ 0.8517601490020752 │ +│ sim_test_loss │ 0.41305387020111084 │ +│ sim_test_precision │ 0.851963460445404 │ +│ sim_test_recall │ 0.8516819477081299 │ +│ sim_test_score │ 0.8516819477081299 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 269.52it/s +Test metrics: [{'sim_test_loss': 0.41305387020111084, 'lr': 0.0003454914840403944, 'sim_test_accuracy': 0.8516819477081299, 'sim_test_precision': 0.851963460445404, 'sim_test_recall': 0.8516819477081299, 'sim_test_f1-score': 0.8517601490020752, 'sim_test_support': 1962.0, 'sim_test_score': 0.8516819477081299}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpsyiazfhq/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpsyiazfhq/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpsyiazfhq/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpsyiazfhq/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2800 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ sim │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─EncoderDCNv2: 2-2 [128, 32] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 32] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 32] [128, 128] 4,224 -- 540,672 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +========================================================================================================================================================================== +Total params: 185,763 +Trainable params: 185,763 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 23.78 +========================================================================================================================================================================== +Input size (MB): 1.05 +Forward/backward pass size (MB): 1.08 +Params size (MB): 0.74 +Estimated Total Size (MB): 2.88 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 185 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 185 K +Non-trainable params: 0 +Total params: 185 K +Total estimated model params size (MB): 0.743 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 32/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 166.70it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0002591231605038047 │ +│ sim_val_accuracy │ 0.8335134983062744 │ +│ sim_val_f1-score │ 0.832685649394989 │ +│ sim_val_loss │ 0.4672321081161499 │ +│ sim_val_precision │ 0.8362681865692139 │ +│ sim_val_recall │ 0.8335134983062744 │ +│ sim_val_score │ 0.8335134983062744 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 290.20it/s +Validation metrics: [{'sim_val_loss': 0.4672321081161499, 'lr': 0.0002591231605038047, 'sim_val_accuracy': 0.8335134983062744, 'sim_val_precision': 0.8362681865692139, 'sim_val_recall': 0.8335134983062744, 'sim_val_f1-score': 0.832685649394989, 'sim_val_support': 925.0, 'sim_val_score': 0.8335134983062744}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 781 samples (39.81%) + Class 1: 568 samples (28.95%) + Class 2: 613 samples (31.24%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.00025912318960763514 │ +│ sim_test_accuracy │ 0.8414883017539978 │ +│ sim_test_f1-score │ 0.840985894203186 │ +│ sim_test_loss │ 0.5182449817657471 │ +│ sim_test_precision │ 0.8461096882820129 │ +│ sim_test_recall │ 0.8414883017539978 │ +│ sim_test_score │ 0.8414883017539978 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 311.98it/s +Test metrics: [{'sim_test_loss': 0.5182449817657471, 'lr': 0.00025912318960763514, 'sim_test_accuracy': 0.8414883017539978, 'sim_test_precision': 0.8461096882820129, 'sim_test_recall': 0.8414883017539978, 'sim_test_f1-score': 0.840985894203186, 'sim_test_support': 1962.0, 'sim_test_score': 0.8414883017539978}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpsyiazfhq/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpsyiazfhq/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpsyiazfhq/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpsyiazfhq/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2800 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ sim │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─EncoderDCNv2: 2-2 [128, 32] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 32] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 32] [128, 128] 4,224 -- 540,672 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +========================================================================================================================================================================== +Total params: 185,763 +Trainable params: 185,763 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 23.78 +========================================================================================================================================================================== +Input size (MB): 1.05 +Forward/backward pass size (MB): 1.08 +Params size (MB): 0.74 +Estimated Total Size (MB): 2.88 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 185 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 185 K +Non-trainable params: 0 +Total params: 185 K +Total estimated model params size (MB): 0.743 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 27/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 101.67it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.000406309321988374 │ +│ sim_val_accuracy │ 0.8540540337562561 │ +│ sim_val_f1-score │ 0.8539830446243286 │ +│ sim_val_loss │ 0.40959325432777405 │ +│ sim_val_precision │ 0.8541745543479919 │ +│ sim_val_recall │ 0.8540540337562561 │ +│ sim_val_score │ 0.8540540337562561 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 158.15it/s +Validation metrics: [{'sim_val_loss': 0.40959325432777405, 'lr': 0.000406309321988374, 'sim_val_accuracy': 0.8540540337562561, 'sim_val_precision': 0.8541745543479919, 'sim_val_recall': 0.8540540337562561, 'sim_val_f1-score': 0.8539830446243286, 'sim_val_support': 925.0, 'sim_val_score': 0.8540540337562561}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 732 samples (37.31%) + Class 1: 617 samples (31.45%) + Class 2: 613 samples (31.24%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0004063093801960349 │ +│ sim_test_accuracy │ 0.8465850949287415 │ +│ sim_test_f1-score │ 0.8464018106460571 │ +│ sim_test_loss │ 0.4582127034664154 │ +│ sim_test_precision │ 0.8478139638900757 │ +│ sim_test_recall │ 0.8465850949287415 │ +│ sim_test_score │ 0.8465850949287415 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 103.72it/s +Test metrics: [{'sim_test_loss': 0.4582127034664154, 'lr': 0.0004063093801960349, 'sim_test_accuracy': 0.8465850949287415, 'sim_test_precision': 0.8478139638900757, 'sim_test_recall': 0.8465850949287415, 'sim_test_f1-score': 0.8464018106460571, 'sim_test_support': 1962.0, 'sim_test_score': 0.8465850949287415}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpsyiazfhq/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpsyiazfhq/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpsyiazfhq/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpsyiazfhq/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2800 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ sim │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─EncoderDCNv2: 2-2 [128, 32] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 32] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 32] [128, 128] 4,224 -- 540,672 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +========================================================================================================================================================================== +Total params: 185,763 +Trainable params: 185,763 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 23.78 +========================================================================================================================================================================== +Input size (MB): 1.05 +Forward/backward pass size (MB): 1.08 +Params size (MB): 0.74 +Estimated Total Size (MB): 2.88 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 185 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 185 K +Non-trainable params: 0 +Total params: 185 K +Total estimated model params size (MB): 0.743 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 31/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 90.01it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0002871103642974049 │ +│ sim_val_accuracy │ 0.8399999737739563 │ +│ sim_val_f1-score │ 0.8400229811668396 │ +│ sim_val_loss │ 0.4338511526584625 │ +│ sim_val_precision │ 0.8406462073326111 │ +│ sim_val_recall │ 0.8399999737739563 │ +│ sim_val_score │ 0.8399999737739563 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 79.64it/s +Validation metrics: [{'sim_val_loss': 0.4338511526584625, 'lr': 0.0002871103642974049, 'sim_val_accuracy': 0.8399999737739563, 'sim_val_precision': 0.8406462073326111, 'sim_val_recall': 0.8399999737739563, 'sim_val_f1-score': 0.8400229811668396, 'sim_val_support': 925.0, 'sim_val_score': 0.8399999737739563}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 713 samples (36.34%) + Class 1: 651 samples (33.18%) + Class 2: 598 samples (30.48%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.00028711039340123534 │ +│ sim_test_accuracy │ 0.8567788004875183 │ +│ sim_test_f1-score │ 0.8567251563072205 │ +│ sim_test_loss │ 0.4340159595012665 │ +│ sim_test_precision │ 0.8579704165458679 │ +│ sim_test_recall │ 0.8567788004875183 │ +│ sim_test_score │ 0.8567788004875183 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 128.75it/s +Test metrics: [{'sim_test_loss': 0.4340159595012665, 'lr': 0.00028711039340123534, 'sim_test_accuracy': 0.8567788004875183, 'sim_test_precision': 0.8579704165458679, 'sim_test_recall': 0.8567788004875183, 'sim_test_f1-score': 0.8567251563072205, 'sim_test_support': 1962.0, 'sim_test_score': 0.8567788004875183}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=dcnv2 --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpsyiazfhq/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --hidden_dim=128 --n_hidden_layers=4 --n_cross_layers=2 --hidden_dropout=0.1 --cross_dropout=0.1 --d_embedding=128 --stacked=True --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpsyiazfhq/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ dcnv2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpsyiazfhq/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpsyiazfhq/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: dcnv2 +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ categories │ [] │ +├──────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ hidden_dim │ 128 │ +├──────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├──────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├──────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├──────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├──────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├──────────────────────┼────────────────┤ +│ stacked │ True │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +├──────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘══════════════════════╧════════════════╛ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/src/models/dcnv2.py:40: UserWarning: Initializing zero-element tensors is a no-op + nn.init.kaiming_uniform_(self.category_embeddings.weight, a=math.sqrt(5)) +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_hidden_layers │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cross_layers │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ hidden_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cross_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ stacked │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_embedding │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2800 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ sim │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +DCNv2 [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─EncoderDCNv2: 2-2 [128, 32] [128, 128] -- -- -- +│ │ └─DCNv2Base: 3-2 [128, 32] [128, 128] -- -- -- +│ │ │ └─Embedding: 4-1 [128, 0] [128, 0, 128] -- -- -- +│ │ │ └─Linear: 4-2 [128, 32] [128, 128] 4,224 -- 540,672 +│ │ │ └─ModuleList: 4-3 -- -- -- -- -- +│ │ │ │ └─CrossLayer: 5-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-1 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─CrossLayer: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ │ └─Dropout: 6-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-3 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-4 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-7 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-8 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-9 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-10 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-11 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Linear: 5-12 [128, 128] [128, 128] 16,512 -- 2,113,536 +│ │ │ │ └─ReLU: 5-13 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─Dropout: 5-14 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-5 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +========================================================================================================================================================================== +Total params: 185,763 +Trainable params: 185,763 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 23.78 +========================================================================================================================================================================== +Input size (MB): 1.05 +Forward/backward pass size (MB): 1.08 +Params size (MB): 0.74 +Estimated Total Size (MB): 2.88 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 185 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 185 K +Non-trainable params: 0 +Total params: 185 K +Total estimated model params size (MB): 0.743 +Modules in train mode: 30 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 19/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 127.14it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0006545084179379046 │ +│ sim_val_accuracy │ 0.837837815284729 │ +│ sim_val_f1-score │ 0.8372050523757935 │ +│ sim_val_loss │ 0.41764041781425476 │ +│ sim_val_precision │ 0.8485930562019348 │ +│ sim_val_recall │ 0.837837815284729 │ +│ sim_val_score │ 0.837837815284729 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 416.79it/s +Validation metrics: [{'sim_val_loss': 0.41764041781425476, 'lr': 0.0006545084179379046, 'sim_val_accuracy': 0.837837815284729, 'sim_val_precision': 0.8485930562019348, 'sim_val_recall': 0.837837815284729, 'sim_val_f1-score': 0.8372050523757935, 'sim_val_support': 925.0, 'sim_val_score': 0.837837815284729}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 756 samples (38.53%) + Class 1: 712 samples (36.29%) + Class 2: 494 samples (25.18%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0006545084179379046 │ +│ sim_test_accuracy │ 0.8363914489746094 │ +│ sim_test_f1-score │ 0.8348156213760376 │ +│ sim_test_loss │ 0.4272391200065613 │ +│ sim_test_precision │ 0.8461126089096069 │ +│ sim_test_recall │ 0.8363914489746094 │ +│ sim_test_score │ 0.8363914489746094 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 138.25it/s +Test metrics: [{'sim_test_loss': 0.4272391200065613, 'lr': 0.0006545084179379046, 'sim_test_accuracy': 0.8363914489746094, 'sim_test_precision': 0.8461126089096069, 'sim_test_recall': 0.8363914489746094, 'sim_test_f1-score': 0.8348156213760376, 'sim_test_support': 1962.0, 'sim_test_score': 0.8363914489746094}] +Wrote results for dcnv2 on sim to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results/dcnv2_sim_downstream_results.json +Running downstream for dataset 'sim', model 'resnet'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpyj2ieumd/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpyj2ieumd/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpyj2ieumd/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpyj2ieumd/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─ResNet: 2-2 [128, 32] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 32] [128, 128] 4,224 -- 540,672 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +========================================================================================================================================================================== +Total params: 285,475 +Trainable params: 285,475 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 36.54 +========================================================================================================================================================================== +Input size (MB): 1.05 +Forward/backward pass size (MB): 2.00 +Params size (MB): 1.14 +Estimated Total Size (MB): 4.19 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 285 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 285 K +Non-trainable params: 0 +Total params: 285 K +Total estimated model params size (MB): 1.142 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 25/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 171.29it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0006545084179379046 │ +│ sim_val_accuracy │ 0.8529729843139648 │ +│ sim_val_f1-score │ 0.8530251979827881 │ +│ sim_val_loss │ 0.3943820595741272 │ +│ sim_val_precision │ 0.853134274482727 │ +│ sim_val_recall │ 0.8529729843139648 │ +│ sim_val_score │ 0.8529729843139648 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 365.30it/s +Validation metrics: [{'sim_val_loss': 0.3943820595741272, 'lr': 0.0006545084179379046, 'sim_val_accuracy': 0.8529729843139648, 'sim_val_precision': 0.853134274482727, 'sim_val_recall': 0.8529729843139648, 'sim_val_f1-score': 0.8530251979827881, 'sim_val_support': 925.0, 'sim_val_score': 0.8529729843139648}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 690 samples (35.17%) + Class 1: 646 samples (32.93%) + Class 2: 626 samples (31.91%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0006545084179379046 │ +│ sim_test_accuracy │ 0.8491335511207581 │ +│ sim_test_f1-score │ 0.8490909934043884 │ +│ sim_test_loss │ 0.38298365473747253 │ +│ sim_test_precision │ 0.8492569923400879 │ +│ sim_test_recall │ 0.8491335511207581 │ +│ sim_test_score │ 0.8491335511207581 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 413.94it/s +Test metrics: [{'sim_test_loss': 0.38298365473747253, 'lr': 0.0006545084179379046, 'sim_test_accuracy': 0.8491335511207581, 'sim_test_precision': 0.8492569923400879, 'sim_test_recall': 0.8491335511207581, 'sim_test_f1-score': 0.8490909934043884, 'sim_test_support': 1962.0, 'sim_test_score': 0.8491335511207581}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpyj2ieumd/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpyj2ieumd/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpyj2ieumd/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpyj2ieumd/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─ResNet: 2-2 [128, 32] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 32] [128, 128] 4,224 -- 540,672 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +========================================================================================================================================================================== +Total params: 285,475 +Trainable params: 285,475 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 36.54 +========================================================================================================================================================================== +Input size (MB): 1.05 +Forward/backward pass size (MB): 2.00 +Params size (MB): 1.14 +Estimated Total Size (MB): 4.19 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 285 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 285 K +Non-trainable params: 0 +Total params: 285 K +Total estimated model params size (MB): 1.142 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 21/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 180.31it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 9.549150126986206e-05 │ +│ sim_val_accuracy │ 0.8421621918678284 │ +│ sim_val_f1-score │ 0.8418724536895752 │ +│ sim_val_loss │ 0.374656617641449 │ +│ sim_val_precision │ 0.8424528241157532 │ +│ sim_val_recall │ 0.8421621918678284 │ +│ sim_val_score │ 0.8421621918678284 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 300.40it/s +Validation metrics: [{'sim_val_loss': 0.374656617641449, 'lr': 9.549150126986206e-05, 'sim_val_accuracy': 0.8421621918678284, 'sim_val_precision': 0.8424528241157532, 'sim_val_recall': 0.8421621918678284, 'sim_val_f1-score': 0.8418724536895752, 'sim_val_support': 925.0, 'sim_val_score': 0.8421621918678284}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 731 samples (37.26%) + Class 1: 585 samples (29.82%) + Class 2: 646 samples (32.93%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 9.549151582177728e-05 │ +│ sim_test_accuracy │ 0.8430173397064209 │ +│ sim_test_f1-score │ 0.842693567276001 │ +│ sim_test_loss │ 0.37531402707099915 │ +│ sim_test_precision │ 0.8445531129837036 │ +│ sim_test_recall │ 0.8430173397064209 │ +│ sim_test_score │ 0.8430173397064209 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 407.91it/s +Test metrics: [{'sim_test_loss': 0.37531402707099915, 'lr': 9.549151582177728e-05, 'sim_test_accuracy': 0.8430173397064209, 'sim_test_precision': 0.8445531129837036, 'sim_test_recall': 0.8430173397064209, 'sim_test_f1-score': 0.842693567276001, 'sim_test_support': 1962.0, 'sim_test_score': 0.8430173397064209}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpyj2ieumd/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpyj2ieumd/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpyj2ieumd/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpyj2ieumd/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─ResNet: 2-2 [128, 32] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 32] [128, 128] 4,224 -- 540,672 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +========================================================================================================================================================================== +Total params: 285,475 +Trainable params: 285,475 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 36.54 +========================================================================================================================================================================== +Input size (MB): 1.05 +Forward/backward pass size (MB): 2.00 +Params size (MB): 1.14 +Estimated Total Size (MB): 4.19 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 285 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 285 K +Non-trainable params: 0 +Total params: 285 K +Total estimated model params size (MB): 1.142 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 25/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 148.64it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0006545084179379046 │ +│ sim_val_accuracy │ 0.8399999737739563 │ +│ sim_val_f1-score │ 0.8400732278823853 │ +│ sim_val_loss │ 0.3847091794013977 │ +│ sim_val_precision │ 0.8406162858009338 │ +│ sim_val_recall │ 0.8399999737739563 │ +│ sim_val_score │ 0.8399999737739563 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 155.92it/s +Validation metrics: [{'sim_val_loss': 0.3847091794013977, 'lr': 0.0006545084179379046, 'sim_val_accuracy': 0.8399999737739563, 'sim_val_precision': 0.8406162858009338, 'sim_val_recall': 0.8399999737739563, 'sim_val_f1-score': 0.8400732278823853, 'sim_val_support': 925.0, 'sim_val_score': 0.8399999737739563}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 680 samples (34.66%) + Class 1: 673 samples (34.30%) + Class 2: 609 samples (31.04%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0006545084179379046 │ +│ sim_test_accuracy │ 0.8338429927825928 │ +│ sim_test_f1-score │ 0.833795964717865 │ +│ sim_test_loss │ 0.38975510001182556 │ +│ sim_test_precision │ 0.8344905972480774 │ +│ sim_test_recall │ 0.8338429927825928 │ +│ sim_test_score │ 0.8338429927825928 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 130.67it/s +Test metrics: [{'sim_test_loss': 0.38975510001182556, 'lr': 0.0006545084179379046, 'sim_test_accuracy': 0.8338429927825928, 'sim_test_precision': 0.8344905972480774, 'sim_test_recall': 0.8338429927825928, 'sim_test_f1-score': 0.833795964717865, 'sim_test_support': 1962.0, 'sim_test_score': 0.8338429927825928}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpyj2ieumd/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpyj2ieumd/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpyj2ieumd/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpyj2ieumd/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─ResNet: 2-2 [128, 32] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 32] [128, 128] 4,224 -- 540,672 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +========================================================================================================================================================================== +Total params: 285,475 +Trainable params: 285,475 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 36.54 +========================================================================================================================================================================== +Input size (MB): 1.05 +Forward/backward pass size (MB): 2.00 +Params size (MB): 1.14 +Estimated Total Size (MB): 4.19 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 285 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 285 K +Non-trainable params: 0 +Total params: 285 K +Total estimated model params size (MB): 1.142 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 27/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 162.23it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 9.549150126986206e-05 │ +│ sim_val_accuracy │ 0.8454053997993469 │ +│ sim_val_f1-score │ 0.8451192378997803 │ +│ sim_val_loss │ 0.4110054075717926 │ +│ sim_val_precision │ 0.8454695343971252 │ +│ sim_val_recall │ 0.8454053997993469 │ +│ sim_val_score │ 0.8454053997993469 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 183.96it/s +Validation metrics: [{'sim_val_loss': 0.4110054075717926, 'lr': 9.549150126986206e-05, 'sim_val_accuracy': 0.8454053997993469, 'sim_val_precision': 0.8454695343971252, 'sim_val_recall': 0.8454053997993469, 'sim_val_f1-score': 0.8451192378997803, 'sim_val_support': 925.0, 'sim_val_score': 0.8454053997993469}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 718 samples (36.60%) + Class 1: 606 samples (30.89%) + Class 2: 638 samples (32.52%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 9.549151582177728e-05 │ +│ sim_test_accuracy │ 0.8389398455619812 │ +│ sim_test_f1-score │ 0.8385645151138306 │ +│ sim_test_loss │ 0.40992864966392517 │ +│ sim_test_precision │ 0.8392660021781921 │ +│ sim_test_recall │ 0.8389398455619812 │ +│ sim_test_score │ 0.8389398455619812 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 129.93it/s +Test metrics: [{'sim_test_loss': 0.40992864966392517, 'lr': 9.549151582177728e-05, 'sim_test_accuracy': 0.8389398455619812, 'sim_test_precision': 0.8392660021781921, 'sim_test_recall': 0.8389398455619812, 'sim_test_f1-score': 0.8385645151138306, 'sim_test_support': 1962.0, 'sim_test_score': 0.8389398455619812}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=resnet --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpyj2ieumd/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_out=128 --d_block=128 --d_hidden=256 --dropout1=0.1 --dropout2=0.0 --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpyj2ieumd/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ resnet │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpyj2ieumd/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpyj2ieumd/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: resnet +╒══════════════════════╤════════════════╕ +│ cardinalities │ [] │ +├──────────────────────┼────────────────┤ +│ d_block │ 128 │ +├──────────────────────┼────────────────┤ +│ d_hidden │ 256 │ +├──────────────────────┼────────────────┤ +│ d_hidden_multiplier │ │ +├──────────────────────┼────────────────┤ +│ d_out │ 128 │ +├──────────────────────┼────────────────┤ +│ dataset_name │ │ +├──────────────────────┼────────────────┤ +│ dropout1 │ 0.1 │ +├──────────────────────┼────────────────┤ +│ dropout2 │ 0.0 │ +├──────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├──────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├──────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├──────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├──────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├──────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├──────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├──────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├──────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +╘══════════════════════╧════════════════╛ +using embedding: True +========================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================== +ResNet [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─ResNet: 2-2 [128, 32] [128, 128] -- -- -- +│ │ └─Linear: 3-2 [128, 32] [128, 128] 4,224 -- 540,672 +│ │ └─ModuleList: 3-3 -- -- -- -- -- +│ │ │ └─Sequential: 4-1 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-1 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-2 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-3 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-4 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-5 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-6 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-7 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-8 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-9 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-10 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-11 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-12 [128, 128] [128, 128] -- -- -- +│ │ │ └─Sequential: 4-3 [128, 128] [128, 128] -- -- -- +│ │ │ │ └─BatchNorm1d: 5-13 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ └─Linear: 5-14 [128, 128] [128, 256] 33,024 -- 4,227,072 +│ │ │ │ └─ReLU: 5-15 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Dropout: 5-16 [128, 256] [128, 256] -- -- -- +│ │ │ │ └─Linear: 5-17 [128, 256] [128, 128] 32,896 -- 4,210,688 +│ │ │ │ └─Dropout: 5-18 [128, 128] [128, 128] -- -- -- +│ │ └─Sequential: 3-4 [128, 128] [128, 128] -- -- -- +│ │ │ └─BatchNorm1d: 4-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 128] [128, 128] -- -- -- +│ │ │ └─Linear: 4-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +========================================================================================================================================================================== +Total params: 285,475 +Trainable params: 285,475 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 36.54 +========================================================================================================================================================================== +Input size (MB): 1.05 +Forward/backward pass size (MB): 2.00 +Params size (MB): 1.14 +Estimated Total Size (MB): 4.19 +========================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 285 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 285 K +Non-trainable params: 0 +Total params: 285 K +Total estimated model params size (MB): 1.142 +Modules in train mode: 33 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 31/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 151.18it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 9.549150126986206e-05 │ +│ sim_val_accuracy │ 0.8302702903747559 │ +│ sim_val_f1-score │ 0.8307524919509888 │ +│ sim_val_loss │ 0.4437926113605499 │ +│ sim_val_precision │ 0.8343738913536072 │ +│ sim_val_recall │ 0.8302702903747559 │ +│ sim_val_score │ 0.8302702903747559 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 169.09it/s +Validation metrics: [{'sim_val_loss': 0.4437926113605499, 'lr': 9.549150126986206e-05, 'sim_val_accuracy': 0.8302702903747559, 'sim_val_precision': 0.8343738913536072, 'sim_val_recall': 0.8302702903747559, 'sim_val_f1-score': 0.8307524919509888, 'sim_val_support': 925.0, 'sim_val_score': 0.8302702903747559}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 629 samples (32.06%) + Class 1: 744 samples (37.92%) + Class 2: 589 samples (30.02%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 9.549151582177728e-05 │ +│ sim_test_accuracy │ 0.831294596195221 │ +│ sim_test_f1-score │ 0.8315088748931885 │ +│ sim_test_loss │ 0.4607800245285034 │ +│ sim_test_precision │ 0.8364284038543701 │ +│ sim_test_recall │ 0.831294596195221 │ +│ sim_test_score │ 0.831294596195221 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 408.35it/s +Test metrics: [{'sim_test_loss': 0.4607800245285034, 'lr': 9.549151582177728e-05, 'sim_test_accuracy': 0.831294596195221, 'sim_test_precision': 0.8364284038543701, 'sim_test_recall': 0.831294596195221, 'sim_test_f1-score': 0.8315088748931885, 'sim_test_support': 1962.0, 'sim_test_score': 0.831294596195221}] +Wrote results for resnet on sim to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results/resnet_sim_downstream_results.json +Running downstream for dataset 'sim', model 'autoint'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpazbhnyrm/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpazbhnyrm/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpazbhnyrm/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpazbhnyrm/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2800 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ sim │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─EncoderAutoIntBase: 2-2 [128, 32] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 32] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 32] [128, 32, 128] 4,096 -- 524,288 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 32, 32] [256, 32, 32] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 32, 128] [128, 32, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 32, 128] [128, 32, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 32, 32] [256, 32, 32] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 32, 128] [128, 32, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 32, 128] [128, 32, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 32, 32] [256, 32, 32] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 32, 128] [128, 32, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 32, 128] [128, 32, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 4096] [128, 128] 524,416 -- 67,125,248 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +==================================================================================================================================================================================== +Total params: 792,995 +Trainable params: 792,995 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 101.50 +==================================================================================================================================================================================== +Input size (MB): 1.05 +Forward/backward pass size (MB): 67.28 +Params size (MB): 3.17 +Estimated Total Size (MB): 71.50 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 792 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 792 K +Non-trainable params: 0 +Total params: 792 K +Total estimated model params size (MB): 3.172 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 30/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 119.88it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0003159377083647996 │ +│ sim_val_accuracy │ 0.8605405688285828 │ +│ sim_val_f1-score │ 0.8606148958206177 │ +│ sim_val_loss │ 0.38667839765548706 │ +│ sim_val_precision │ 0.8618879318237305 │ +│ sim_val_recall │ 0.8605405688285828 │ +│ sim_val_score │ 0.8605405688285828 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 277.08it/s +Validation metrics: [{'sim_val_loss': 0.38667839765548706, 'lr': 0.0003159377083647996, 'sim_val_accuracy': 0.8605405688285828, 'sim_val_precision': 0.8618879318237305, 'sim_val_recall': 0.8605405688285828, 'sim_val_f1-score': 0.8606148958206177, 'sim_val_support': 925.0, 'sim_val_score': 0.8605405688285828}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 666 samples (33.94%) + Class 1: 686 samples (34.96%) + Class 2: 610 samples (31.09%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0003159377083647996 │ +│ sim_test_accuracy │ 0.8583078384399414 │ +│ sim_test_f1-score │ 0.8583861589431763 │ +│ sim_test_loss │ 0.3953714072704315 │ +│ sim_test_precision │ 0.8595095276832581 │ +│ sim_test_recall │ 0.8583078384399414 │ +│ sim_test_score │ 0.8583078384399414 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 291.17it/s +Test metrics: [{'sim_test_loss': 0.3953714072704315, 'lr': 0.0003159377083647996, 'sim_test_accuracy': 0.8583078384399414, 'sim_test_precision': 0.8595095276832581, 'sim_test_recall': 0.8583078384399414, 'sim_test_f1-score': 0.8583861589431763, 'sim_test_support': 1962.0, 'sim_test_score': 0.8583078384399414}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpazbhnyrm/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=1 --np_seed=1 --output_file=/tmp/tmpazbhnyrm/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpazbhnyrm/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpazbhnyrm/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2800 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ sim │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─EncoderAutoIntBase: 2-2 [128, 32] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 32] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 32] [128, 32, 128] 4,096 -- 524,288 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 32, 32] [256, 32, 32] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 32, 128] [128, 32, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 32, 128] [128, 32, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 32, 32] [256, 32, 32] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 32, 128] [128, 32, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 32, 128] [128, 32, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 32, 32] [256, 32, 32] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 32, 128] [128, 32, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 32, 128] [128, 32, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 4096] [128, 128] 524,416 -- 67,125,248 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +==================================================================================================================================================================================== +Total params: 792,995 +Trainable params: 792,995 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 101.50 +==================================================================================================================================================================================== +Input size (MB): 1.05 +Forward/backward pass size (MB): 67.28 +Params size (MB): 3.17 +Estimated Total Size (MB): 71.50 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 792 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 792 K +Non-trainable params: 0 +Total params: 792 K +Total estimated model params size (MB): 3.172 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 23/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 102.93it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.000531395198777318 │ +│ sim_val_accuracy │ 0.8735135197639465 │ +│ sim_val_f1-score │ 0.8737159371376038 │ +│ sim_val_loss │ 0.3551470637321472 │ +│ sim_val_precision │ 0.8760514259338379 │ +│ sim_val_recall │ 0.8735135197639465 │ +│ sim_val_score │ 0.8735135197639465 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 181.06it/s +Validation metrics: [{'sim_val_loss': 0.3551470637321472, 'lr': 0.000531395198777318, 'sim_val_accuracy': 0.8735135197639465, 'sim_val_precision': 0.8760514259338379, 'sim_val_recall': 0.8735135197639465, 'sim_val_f1-score': 0.8737159371376038, 'sim_val_support': 925.0, 'sim_val_score': 0.8735135197639465}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 649 samples (33.08%) + Class 1: 705 samples (35.93%) + Class 2: 608 samples (30.99%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0005313953151926398 │ +│ sim_test_accuracy │ 0.8542304039001465 │ +│ sim_test_f1-score │ 0.8543589115142822 │ +│ sim_test_loss │ 0.36710795760154724 │ +│ sim_test_precision │ 0.856368362903595 │ +│ sim_test_recall │ 0.8542304039001465 │ +│ sim_test_score │ 0.8542304039001465 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 232.66it/s +Test metrics: [{'sim_test_loss': 0.36710795760154724, 'lr': 0.0005313953151926398, 'sim_test_accuracy': 0.8542304039001465, 'sim_test_precision': 0.856368362903595, 'sim_test_recall': 0.8542304039001465, 'sim_test_f1-score': 0.8543589115142822, 'sim_test_support': 1962.0, 'sim_test_score': 0.8542304039001465}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpazbhnyrm/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=2 --np_seed=2 --output_file=/tmp/tmpazbhnyrm/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpazbhnyrm/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpazbhnyrm/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2800 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ sim │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─EncoderAutoIntBase: 2-2 [128, 32] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 32] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 32] [128, 32, 128] 4,096 -- 524,288 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 32, 32] [256, 32, 32] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 32, 128] [128, 32, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 32, 128] [128, 32, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 32, 32] [256, 32, 32] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 32, 128] [128, 32, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 32, 128] [128, 32, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 32, 32] [256, 32, 32] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 32, 128] [128, 32, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 32, 128] [128, 32, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 4096] [128, 128] 524,416 -- 67,125,248 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +==================================================================================================================================================================================== +Total params: 792,995 +Trainable params: 792,995 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 101.50 +==================================================================================================================================================================================== +Input size (MB): 1.05 +Forward/backward pass size (MB): 67.28 +Params size (MB): 3.17 +Estimated Total Size (MB): 71.50 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 792 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 792 K +Non-trainable params: 0 +Total params: 792 K +Total estimated model params size (MB): 3.172 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 20/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 89.23it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0006243449170142412 │ +│ sim_val_accuracy │ 0.8529729843139648 │ +│ sim_val_f1-score │ 0.8526897430419922 │ +│ sim_val_loss │ 0.38884830474853516 │ +│ sim_val_precision │ 0.8547060489654541 │ +│ sim_val_recall │ 0.8529729843139648 │ +│ sim_val_score │ 0.8529729843139648 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 106.45it/s +Validation metrics: [{'sim_val_loss': 0.38884830474853516, 'lr': 0.0006243449170142412, 'sim_val_accuracy': 0.8529729843139648, 'sim_val_precision': 0.8547060489654541, 'sim_val_recall': 0.8529729843139648, 'sim_val_f1-score': 0.8526897430419922, 'sim_val_support': 925.0, 'sim_val_score': 0.8529729843139648}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 704 samples (35.88%) + Class 1: 576 samples (29.36%) + Class 2: 682 samples (34.76%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0006243448005989194 │ +│ sim_test_accuracy │ 0.8598368763923645 │ +│ sim_test_f1-score │ 0.8593538999557495 │ +│ sim_test_loss │ 0.406281054019928 │ +│ sim_test_precision │ 0.8612660765647888 │ +│ sim_test_recall │ 0.8598368763923645 │ +│ sim_test_score │ 0.8598368763923645 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 135.68it/s +Test metrics: [{'sim_test_loss': 0.406281054019928, 'lr': 0.0006243448005989194, 'sim_test_accuracy': 0.8598368763923645, 'sim_test_precision': 0.8612660765647888, 'sim_test_recall': 0.8598368763923645, 'sim_test_f1-score': 0.8593538999557495, 'sim_test_support': 1962.0, 'sim_test_score': 0.8598368763923645}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpazbhnyrm/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=3 --np_seed=3 --output_file=/tmp/tmpazbhnyrm/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpazbhnyrm/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpazbhnyrm/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2800 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ sim │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─EncoderAutoIntBase: 2-2 [128, 32] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 32] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 32] [128, 32, 128] 4,096 -- 524,288 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 32, 32] [256, 32, 32] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 32, 128] [128, 32, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 32, 128] [128, 32, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 32, 32] [256, 32, 32] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 32, 128] [128, 32, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 32, 128] [128, 32, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 32, 32] [256, 32, 32] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 32, 128] [128, 32, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 32, 128] [128, 32, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 4096] [128, 128] 524,416 -- 67,125,248 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +==================================================================================================================================================================================== +Total params: 792,995 +Trainable params: 792,995 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 101.50 +==================================================================================================================================================================================== +Input size (MB): 1.05 +Forward/backward pass size (MB): 67.28 +Params size (MB): 3.17 +Estimated Total Size (MB): 71.50 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 792 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 792 K +Non-trainable params: 0 +Total params: 792 K +Total estimated model params size (MB): 3.172 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 24/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 95.03it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0005000000237487257 │ +│ sim_val_accuracy │ 0.8605405688285828 │ +│ sim_val_f1-score │ 0.8607043027877808 │ +│ sim_val_loss │ 0.36793437600135803 │ +│ sim_val_precision │ 0.861944854259491 │ +│ sim_val_recall │ 0.8605405688285828 │ +│ sim_val_score │ 0.8605405688285828 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 176.44it/s +Validation metrics: [{'sim_val_loss': 0.36793437600135803, 'lr': 0.0005000000237487257, 'sim_val_accuracy': 0.8605405688285828, 'sim_val_precision': 0.861944854259491, 'sim_val_recall': 0.8605405688285828, 'sim_val_f1-score': 0.8607043027877808, 'sim_val_support': 925.0, 'sim_val_score': 0.8605405688285828}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 661 samples (33.69%) + Class 1: 652 samples (33.23%) + Class 2: 649 samples (33.08%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0005000000819563866 │ +│ sim_test_accuracy │ 0.8552497625350952 │ +│ sim_test_f1-score │ 0.8552965521812439 │ +│ sim_test_loss │ 0.3741350769996643 │ +│ sim_test_precision │ 0.8554317355155945 │ +│ sim_test_recall │ 0.8552497625350952 │ +│ sim_test_score │ 0.8552497625350952 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 126.30it/s +Test metrics: [{'sim_test_loss': 0.3741350769996643, 'lr': 0.0005000000819563866, 'sim_test_accuracy': 0.8552497625350952, 'sim_test_precision': 0.8554317355155945, 'sim_test_recall': 0.8552497625350952, 'sim_test_f1-score': 0.8552965521812439, 'sim_test_support': 1962.0, 'sim_test_score': 0.8552497625350952}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=autoint --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmpazbhnyrm/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --activation=relu --initialization=kaiming --n_layers=3 --d_token=128 --n_heads=2 --attention_dropout=0.1 --residual_dropout=0.1 --prenormalization=False --torch_seed=4 --np_seed=4 --output_file=/tmp/tmpazbhnyrm/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ autoint │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpazbhnyrm/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpazbhnyrm/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: autoint +╒════════════════════════╤════════════════╕ +│ activation │ relu │ +├────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────┼────────────────┤ +│ d_out │ 32 │ +├────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────┼────────────────┤ +│ dataset_name │ │ +├────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├────────────────────────┼────────────────┤ +│ kv_compression │ │ +├────────────────────────┼────────────────┤ +│ kv_compression_sharing │ │ +├────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├────────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +├────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ categories │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_layers │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_token │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_heads │ 2 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ activation │ relu │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prenormalization │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ initialization │ kaiming │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2800 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ sim │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +==================================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +==================================================================================================================================================================================== +AutoInt [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─EncoderAutoIntBase: 2-2 [128, 32] [128, 128] -- -- -- +│ │ └─AutoIntBase: 3-2 [128, 32] [128, 128] -- -- -- +│ │ │ └─Tokenizer: 4-1 [128, 32] [128, 32, 128] 4,096 -- 524,288 +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-1 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-1 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-1 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-2 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-3 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-4 [256, 32, 32] [256, 32, 32] -- -- -- +│ │ │ │ │ └─Linear: 6-2 [128, 32, 128] [128, 32, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-3 [128, 32, 128] [128, 32, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-3 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-2 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-4 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-5 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-6 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-7 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-8 [256, 32, 32] [256, 32, 32] -- -- -- +│ │ │ │ │ └─Linear: 6-5 [128, 32, 128] [128, 32, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-6 [128, 32, 128] [128, 32, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-5 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ └─ModuleList: 4-6 -- -- (recursive) -- -- +│ │ │ │ └─ModuleDict: 5-3 -- -- -- -- -- +│ │ │ │ │ └─MultiheadAttention: 6-7 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ │ │ │ └─Linear: 7-9 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-10 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Linear: 7-11 [128, 32, 128] [128, 32, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-12 [256, 32, 32] [256, 32, 32] -- -- -- +│ │ │ │ │ └─Linear: 6-8 [128, 32, 128] [128, 32, 128] 16,384 -- 2,097,152 +│ │ │ │ │ └─LayerNorm: 6-9 [128, 32, 128] [128, 32, 128] 256 -- 32,768 +│ │ │ └─ReLU: 4-7 [128, 32, 128] [128, 32, 128] -- -- -- +│ │ │ └─Linear: 4-8 [128, 4096] [128, 128] 524,416 -- 67,125,248 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +==================================================================================================================================================================================== +Total params: 792,995 +Trainable params: 792,995 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 101.50 +==================================================================================================================================================================================== +Input size (MB): 1.05 +Forward/backward pass size (MB): 67.28 +Params size (MB): 3.17 +Estimated Total Size (MB): 71.50 +==================================================================================================================================================================================== +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 792 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 792 K +Non-trainable params: 0 +Total params: 792 K +Total estimated model params size (MB): 3.172 +Modules in train mode: 35 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 20/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 94.38it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0006243449170142412 │ +│ sim_val_accuracy │ 0.8443243503570557 │ +│ sim_val_f1-score │ 0.8440408110618591 │ +│ sim_val_loss │ 0.3748020827770233 │ +│ sim_val_precision │ 0.8513504266738892 │ +│ sim_val_recall │ 0.8443243503570557 │ +│ sim_val_score │ 0.8443243503570557 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 199.09it/s +Validation metrics: [{'sim_val_loss': 0.3748020827770233, 'lr': 0.0006243449170142412, 'sim_val_accuracy': 0.8443243503570557, 'sim_val_precision': 0.8513504266738892, 'sim_val_recall': 0.8443243503570557, 'sim_val_f1-score': 0.8440408110618591, 'sim_val_support': 925.0, 'sim_val_score': 0.8443243503570557}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 797 samples (40.62%) + Class 1: 581 samples (29.61%) + Class 2: 584 samples (29.77%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0006243448005989194 │ +│ sim_test_accuracy │ 0.8450560569763184 │ +│ sim_test_f1-score │ 0.8448898792266846 │ +│ sim_test_loss │ 0.3910144865512848 │ +│ sim_test_precision │ 0.8516185283660889 │ +│ sim_test_recall │ 0.8450560569763184 │ +│ sim_test_score │ 0.8450560569763184 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 214.98it/s +Test metrics: [{'sim_test_loss': 0.3910144865512848, 'lr': 0.0006243448005989194, 'sim_test_accuracy': 0.8450560569763184, 'sim_test_precision': 0.8516185283660889, 'sim_test_recall': 0.8450560569763184, 'sim_test_f1-score': 0.8448898792266846, 'sim_test_support': 1962.0, 'sim_test_score': 0.8450560569763184}] +Wrote results for autoint on sim to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results/autoint_sim_downstream_results.json +Running downstream for dataset 'sim', model 'fttransformer'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp1tybfn2x/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=0 --np_seed=0 --output_file=/tmp/tmp1tybfn2x/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp1tybfn2x/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp1tybfn2x/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2800 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ sim │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─EncoderFTTransformer: 2-2 [128, 32] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 32] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 32] [128, 32, 128] 8,192 -- 1,048,576 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 33, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 33, 33] [512, 33, 33] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 33, 128] [128, 33, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 33, 512] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 33, 256] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 33, 256] [128, 33, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 33, 33] [512, 33, 33] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 33, 128] [128, 33, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 33, 512] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 33, 256] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 33, 256] [128, 33, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 33] [512, 1, 33] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 33, 128] [128, 33, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 33, 512] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 33, 256] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 33, 256] [128, 33, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +========================================================================================================================================================================================= +Total params: 587,299 +Trainable params: 587,299 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 75.17 +========================================================================================================================================================================================= +Input size (MB): 1.05 +Forward/backward pass size (MB): 134.65 +Params size (MB): 2.35 +Estimated Total Size (MB): 138.04 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 586 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 587 K +Non-trainable params: 0 +Total params: 587 K +Total estimated model params size (MB): 2.349 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 34/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 82.24it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0002061073755612597 │ +│ sim_val_accuracy │ 0.8594594597816467 │ +│ sim_val_f1-score │ 0.8594022989273071 │ +│ sim_val_loss │ 0.35619404911994934 │ +│ sim_val_precision │ 0.86168372631073 │ +│ sim_val_recall │ 0.8594594597816467 │ +│ sim_val_score │ 0.8594594597816467 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 190.18it/s +Validation metrics: [{'sim_val_loss': 0.35619404911994934, 'lr': 0.0002061073755612597, 'sim_val_accuracy': 0.8594594597816467, 'sim_val_precision': 0.86168372631073, 'sim_val_recall': 0.8594594597816467, 'sim_val_f1-score': 0.8594022989273071, 'sim_val_support': 925.0, 'sim_val_score': 0.8594594597816467}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 661 samples (33.69%) + Class 1: 609 samples (31.04%) + Class 2: 692 samples (35.27%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.000206107331905514 │ +│ sim_test_accuracy │ 0.8552497625350952 │ +│ sim_test_f1-score │ 0.8551145792007446 │ +│ sim_test_loss │ 0.37115034461021423 │ +│ sim_test_precision │ 0.8561407923698425 │ +│ sim_test_recall │ 0.8552497625350952 │ +│ sim_test_score │ 0.8552497625350952 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 236.68it/s +Test metrics: [{'sim_test_loss': 0.37115034461021423, 'lr': 0.000206107331905514, 'sim_test_accuracy': 0.8552497625350952, 'sim_test_precision': 0.8561407923698425, 'sim_test_recall': 0.8552497625350952, 'sim_test_f1-score': 0.8551145792007446, 'sim_test_support': 1962.0, 'sim_test_score': 0.8552497625350952}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp1tybfn2x/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=1 --np_seed=1 --output_file=/tmp/tmp1tybfn2x/output_1.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp1tybfn2x/output_1.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp1tybfn2x/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2800 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ sim │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─EncoderFTTransformer: 2-2 [128, 32] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 32] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 32] [128, 32, 128] 8,192 -- 1,048,576 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 33, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 33, 33] [512, 33, 33] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 33, 128] [128, 33, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 33, 512] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 33, 256] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 33, 256] [128, 33, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 33, 33] [512, 33, 33] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 33, 128] [128, 33, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 33, 512] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 33, 256] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 33, 256] [128, 33, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 33] [512, 1, 33] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 33, 128] [128, 33, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 33, 512] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 33, 256] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 33, 256] [128, 33, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +========================================================================================================================================================================================= +Total params: 587,299 +Trainable params: 587,299 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 75.17 +========================================================================================================================================================================================= +Input size (MB): 1.05 +Forward/backward pass size (MB): 134.65 +Params size (MB): 2.35 +Estimated Total Size (MB): 138.04 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 586 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 587 K +Non-trainable params: 0 +Total params: 587 K +Total estimated model params size (MB): 2.349 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 30/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 81.68it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0003159377083647996 │ +│ sim_val_accuracy │ 0.8681080937385559 │ +│ sim_val_f1-score │ 0.8680978417396545 │ +│ sim_val_loss │ 0.34779229760169983 │ +│ sim_val_precision │ 0.8688489198684692 │ +│ sim_val_recall │ 0.8681080937385559 │ +│ sim_val_score │ 0.8681080937385559 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 76.17it/s +Validation metrics: [{'sim_val_loss': 0.34779229760169983, 'lr': 0.0003159377083647996, 'sim_val_accuracy': 0.8681080937385559, 'sim_val_precision': 0.8688489198684692, 'sim_val_recall': 0.8681080937385559, 'sim_val_f1-score': 0.8680978417396545, 'sim_val_support': 925.0, 'sim_val_score': 0.8681080937385559}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 664 samples (33.84%) + Class 1: 677 samples (34.51%) + Class 2: 621 samples (31.65%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0003159377083647996 │ +│ sim_test_accuracy │ 0.8598368763923645 │ +│ sim_test_f1-score │ 0.8599119782447815 │ +│ sim_test_loss │ 0.3603042960166931 │ +│ sim_test_precision │ 0.860596239566803 │ +│ sim_test_recall │ 0.8598368763923645 │ +│ sim_test_score │ 0.8598368763923645 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 97.56it/s +Test metrics: [{'sim_test_loss': 0.3603042960166931, 'lr': 0.0003159377083647996, 'sim_test_accuracy': 0.8598368763923645, 'sim_test_precision': 0.860596239566803, 'sim_test_recall': 0.8598368763923645, 'sim_test_f1-score': 0.8599119782447815, 'sim_test_support': 1962.0, 'sim_test_score': 0.8598368763923645}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp1tybfn2x/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=2 --np_seed=2 --output_file=/tmp/tmp1tybfn2x/output_2.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp1tybfn2x/output_2.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp1tybfn2x/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 2 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2800 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ sim │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─EncoderFTTransformer: 2-2 [128, 32] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 32] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 32] [128, 32, 128] 8,192 -- 1,048,576 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 33, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 33, 33] [512, 33, 33] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 33, 128] [128, 33, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 33, 512] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 33, 256] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 33, 256] [128, 33, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 33, 33] [512, 33, 33] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 33, 128] [128, 33, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 33, 512] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 33, 256] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 33, 256] [128, 33, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 33] [512, 1, 33] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 33, 128] [128, 33, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 33, 512] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 33, 256] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 33, 256] [128, 33, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +========================================================================================================================================================================================= +Total params: 587,299 +Trainable params: 587,299 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 75.17 +========================================================================================================================================================================================= +Input size (MB): 1.05 +Forward/backward pass size (MB): 134.65 +Params size (MB): 2.35 +Estimated Total Size (MB): 138.04 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 586 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 587 K +Non-trainable params: 0 +Total params: 587 K +Total estimated model params size (MB): 2.349 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 31/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 80.50it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0002871103642974049 │ +│ sim_val_accuracy │ 0.8659459352493286 │ +│ sim_val_f1-score │ 0.8660420179367065 │ +│ sim_val_loss │ 0.36381348967552185 │ +│ sim_val_precision │ 0.8692029714584351 │ +│ sim_val_recall │ 0.8659459352493286 │ +│ sim_val_score │ 0.8659459352493286 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 76.34it/s +Validation metrics: [{'sim_val_loss': 0.36381348967552185, 'lr': 0.0002871103642974049, 'sim_val_accuracy': 0.8659459352493286, 'sim_val_precision': 0.8692029714584351, 'sim_val_recall': 0.8659459352493286, 'sim_val_f1-score': 0.8660420179367065, 'sim_val_support': 925.0, 'sim_val_score': 0.8659459352493286}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 631 samples (32.16%) + Class 1: 713 samples (36.34%) + Class 2: 618 samples (31.50%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.00028711039340123534 │ +│ sim_test_accuracy │ 0.8572884798049927 │ +│ sim_test_f1-score │ 0.8574735522270203 │ +│ sim_test_loss │ 0.37445560097694397 │ +│ sim_test_precision │ 0.8600521087646484 │ +│ sim_test_recall │ 0.8572884798049927 │ +│ sim_test_score │ 0.8572884798049927 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 97.35it/s +Test metrics: [{'sim_test_loss': 0.37445560097694397, 'lr': 0.00028711039340123534, 'sim_test_accuracy': 0.8572884798049927, 'sim_test_precision': 0.8600521087646484, 'sim_test_recall': 0.8572884798049927, 'sim_test_f1-score': 0.8574735522270203, 'sim_test_support': 1962.0, 'sim_test_score': 0.8572884798049927}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp1tybfn2x/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=3 --np_seed=3 --output_file=/tmp/tmp1tybfn2x/output_3.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp1tybfn2x/output_3.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp1tybfn2x/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 3 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2800 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ sim │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─EncoderFTTransformer: 2-2 [128, 32] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 32] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 32] [128, 32, 128] 8,192 -- 1,048,576 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 33, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 33, 33] [512, 33, 33] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 33, 128] [128, 33, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 33, 512] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 33, 256] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 33, 256] [128, 33, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 33, 33] [512, 33, 33] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 33, 128] [128, 33, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 33, 512] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 33, 256] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 33, 256] [128, 33, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 33] [512, 1, 33] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 33, 128] [128, 33, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 33, 512] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 33, 256] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 33, 256] [128, 33, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +========================================================================================================================================================================================= +Total params: 587,299 +Trainable params: 587,299 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 75.17 +========================================================================================================================================================================================= +Input size (MB): 1.05 +Forward/backward pass size (MB): 134.65 +Params size (MB): 2.35 +Estimated Total Size (MB): 138.04 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 586 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 587 K +Non-trainable params: 0 +Total params: 587 K +Total estimated model params size (MB): 2.349 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 41/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 74.25it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 6.18466583546251e-05 │ +│ sim_val_accuracy │ 0.8594594597816467 │ +│ sim_val_f1-score │ 0.8595176935195923 │ +│ sim_val_loss │ 0.35820046067237854 │ +│ sim_val_precision │ 0.8600558638572693 │ +│ sim_val_recall │ 0.8594594597816467 │ +│ sim_val_score │ 0.8594594597816467 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 212.73it/s +Validation metrics: [{'sim_val_loss': 0.35820046067237854, 'lr': 6.18466583546251e-05, 'sim_val_accuracy': 0.8594594597816467, 'sim_val_precision': 0.8600558638572693, 'sim_val_recall': 0.8594594597816467, 'sim_val_f1-score': 0.8595176935195923, 'sim_val_support': 925.0, 'sim_val_score': 0.8594594597816467}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 694 samples (35.37%) + Class 1: 661 samples (33.69%) + Class 2: 607 samples (30.94%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 6.184667290654033e-05 │ +│ sim_test_accuracy │ 0.8623853325843811 │ +│ sim_test_f1-score │ 0.8623414635658264 │ +│ sim_test_loss │ 0.39219722151756287 │ +│ sim_test_precision │ 0.863053560256958 │ +│ sim_test_recall │ 0.8623853325843811 │ +│ sim_test_score │ 0.8623853325843811 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 224.34it/s +Test metrics: [{'sim_test_loss': 0.39219722151756287, 'lr': 6.184667290654033e-05, 'sim_test_accuracy': 0.8623853325843811, 'sim_test_precision': 0.863053560256958, 'sim_test_recall': 0.8623853325843811, 'sim_test_f1-score': 0.8623414635658264, 'sim_test_support': 1962.0, 'sim_test_score': 0.8623853325843811}] +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=sim --model_name=fttransformer --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy --use_split_assignments=True --exp_train_total_epochs=50 --split_config_file=/tmp/tmp1tybfn2x/split_data/split_config.json --sigma_aggregation=p90 --lr=0.001 --weight_decay=0.01 --batch_size=128 --exp_patience=16 --encoder_type=linear_flatten --n_blocks=3 --d_block=128 --attention_n_heads=4 --attention_dropout=0.1 --ffn_d_hidden_multiplier=2.0 --ffn_dropout=0.1 --residual_dropout=0.0 --torch_seed=4 --np_seed=4 --output_file=/tmp/tmp1tybfn2x/output_4.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 128 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ sim │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ 16 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 50 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ fttransformer │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmp1tybfn2x/output_4.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmp1tybfn2x/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 4 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 0.01 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: fttransformer +╒═════════════════════════╤════════════════╕ +│ attention_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├─────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├─────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├─────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├─────────────────────────┼────────────────┤ +│ dataset_name │ │ +├─────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden │ │ +├─────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├─────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├─────────────────────────┼────────────────┤ +│ input_dim │ 32 │ +├─────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├─────────────────────────┼────────────────┤ +│ iterations_per_epoch │ 56 │ +├─────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├─────────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├─────────────────────────┼────────────────┤ +│ num_epochs │ 50 │ +├─────────────────────────┼────────────────┤ +│ out_dim │ 3 │ +├─────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├─────────────────────────┼────────────────┤ +│ summary_input │ (128, 32, 64) │ +╘═════════════════════════╧════════════════╛ +Model Parameters +╒════════════════════════════════════════════════════╤════════════════╕ +│ Param │ Value │ +╞════════════════════════════════════════════════════╪════════════════╡ +│ input_dim │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_cont_features │ 32 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ cat_cardinalities │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_out │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ n_blocks │ 3 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ d_block │ 128 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_n_heads │ 4 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ attention_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_d_hidden_multiplier │ 2.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ ffn_dropout │ 0.1 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ residual_dropout │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ encoder_type │ linear_flatten │ +├────────────────────────────────────────────────────┼────────────────┤ +│ input_embed_dim │ 64 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ using_embedding │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ training │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ prepare_data_per_node │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ allow_zero_length_dataloader_with_multiple_devices │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _log_hyperparams │ False │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _automatic_optimization │ True │ +├────────────────────────────────────────────────────┼────────────────┤ +│ _fabric_optimizers │ [] │ +├────────────────────────────────────────────────────┼────────────────┤ +│ lr │ 0.001 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ weight_decay │ 0.01 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ T_max │ 2800 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────────────────────────────┼────────────────┤ +│ dataset_name │ sim │ +├────────────────────────────────────────────────────┼────────────────┤ +│ task_type │ multi_class │ +├────────────────────────────────────────────────────┼────────────────┤ +│ suppress_class_frequency_prints │ False │ +╘════════════════════════════════════════════════════╧════════════════╛ +========================================================================================================================================================================================= +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +========================================================================================================================================================================================= +FTTransformer [128, 32, 64] [128, 3] -- -- -- +├─Sequential: 1-1 [128, 32, 64] [128, 128] -- -- -- +│ └─EncodeEmbeddingFeatures: 2-1 [128, 32, 64] [128, 32] -- -- -- +│ │ └─Linear: 3-1 [128, 2048] [128, 32] 65,568 -- 8,392,704 +│ └─EncoderFTTransformer: 2-2 [128, 32] [128, 128] -- -- -- +│ │ └─FTTransformer: 3-2 [128, 32] [128, 128] -- -- -- +│ │ │ └─_CLSEmbedding: 4-1 -- [128, 1, 128] 128 -- 16,384 +│ │ │ └─LinearEmbeddings: 4-2 [128, 32] [128, 32, 128] 8,192 -- 1,048,576 +│ │ │ └─FTTransformerBackbone: 4-3 [128, 33, 128] [128, 128] -- -- -- +│ │ │ │ └─ModuleList: 5-1 -- -- -- -- -- +│ │ │ │ │ └─ModuleDict: 6-1 -- -- -- -- -- +│ │ │ │ │ │ └─MultiheadAttention: 7-1 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-1 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-2 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-3 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-4 [512, 33, 33] [512, 33, 33] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-5 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-2 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-3 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-4 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-6 [128, 33, 128] [128, 33, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-7 [128, 33, 512] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-8 [128, 33, 256] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-9 [128, 33, 256] [128, 33, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-5 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-6 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-2 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-7 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-8 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-10 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-11 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-12 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-13 [512, 33, 33] [512, 33, 33] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-14 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-9 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-10 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-11 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-15 [128, 33, 128] [128, 33, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-16 [128, 33, 512] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-17 [128, 33, 256] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-18 [128, 33, 256] [128, 33, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-12 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-13 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ └─ModuleDict: 6-3 -- -- -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-14 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─MultiheadAttention: 7-15 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-19 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-20 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Linear: 8-21 [128, 33, 128] [128, 33, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ │ └─Dropout: 8-22 [512, 1, 33] [512, 1, 33] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-23 [128, 1, 128] [128, 1, 128] 16,512 -- 2,113,536 +│ │ │ │ │ │ └─Dropout: 7-16 [128, 1, 128] [128, 1, 128] -- -- -- +│ │ │ │ │ │ └─LayerNorm: 7-17 [128, 33, 128] [128, 33, 128] 256 -- 32,768 +│ │ │ │ │ │ └─Sequential: 7-18 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-24 [128, 33, 128] [128, 33, 512] 66,048 -- 8,454,144 +│ │ │ │ │ │ │ └─_ReGLU: 8-25 [128, 33, 512] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Dropout: 8-26 [128, 33, 256] [128, 33, 256] -- -- -- +│ │ │ │ │ │ │ └─Linear: 8-27 [128, 33, 256] [128, 33, 128] 32,896 -- 4,210,688 +│ │ │ │ │ │ └─Dropout: 7-19 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ │ │ └─Identity: 7-20 [128, 33, 128] [128, 33, 128] -- -- -- +│ │ │ │ └─Sequential: 5-2 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─LayerNorm: 6-4 [128, 128] [128, 128] 256 -- 32,768 +│ │ │ │ │ └─ReLU: 6-5 [128, 128] [128, 128] -- -- -- +│ │ │ │ │ └─Linear: 6-6 [128, 128] [128, 128] 16,512 -- 2,113,536 +├─Linear: 1-2 [128, 128] [128, 3] 387 -- 49,536 +========================================================================================================================================================================================= +Total params: 587,299 +Trainable params: 587,299 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 75.17 +========================================================================================================================================================================================= +Input size (MB): 1.05 +Forward/backward pass size (MB): 134.65 +Params size (MB): 2.35 +Estimated Total Size (MB): 138.04 +========================================================================================================================================================================================= +GPU available: True (cuda), used: True +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1.0)` was configured so 100% of the batches per epoch will be used.. +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ CrossEntropyLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Linear │ 387 │ train │ 0 │ +│ 2 │ encoder │ Sequential │ 586 K │ train │ 0 │ +└───┴─────────┴──────────────────┴────────┴───────┴───────┘ +Trainable params: 587 K +Non-trainable params: 0 +Total params: 587 K +Total estimated model params size (MB): 2.349 +Modules in train mode: 65 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +Epoch 30/49 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 56/56 0:00:00 • 0:00:00 78.06it/s +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0003159377083647996 │ +│ sim_val_accuracy │ 0.861621618270874 │ +│ sim_val_f1-score │ 0.8617360591888428 │ +│ sim_val_loss │ 0.3629548251628876 │ +│ sim_val_precision │ 0.8630407452583313 │ +│ sim_val_recall │ 0.861621618270874 │ +│ sim_val_score │ 0.861621618270874 │ +│ sim_val_support │ 925.0 │ +└───────────────────────────┴───────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 8/8 0:00:00 • 0:00:00 75.72it/s +Validation metrics: [{'sim_val_loss': 0.3629548251628876, 'lr': 0.0003159377083647996, 'sim_val_accuracy': 0.861621618270874, 'sim_val_precision': 0.8630407452583313, 'sim_val_recall': 0.861621618270874, 'sim_val_f1-score': 0.8617360591888428, 'sim_val_support': 925.0, 'sim_val_score': 0.861621618270874}] +LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [2] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0: 674 samples (34.35%) + Class 1: 643 samples (32.77%) + Class 2: 645 samples (32.87%) +Predicted class frequencies: + Class 0: 670 samples (34.15%) + Class 1: 672 samples (34.25%) + Class 2: 620 samples (31.60%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ lr │ 0.0003159377083647996 │ +│ sim_test_accuracy │ 0.8552497625350952 │ +│ sim_test_f1-score │ 0.8553267121315002 │ +│ sim_test_loss │ 0.36703062057495117 │ +│ sim_test_precision │ 0.8558992743492126 │ +│ sim_test_recall │ 0.8552497625350952 │ +│ sim_test_score │ 0.8552497625350952 │ +│ sim_test_support │ 1962.0 │ +└───────────────────────────┴───────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 16/16 0:00:00 • 0:00:00 97.60it/s +Test metrics: [{'sim_test_loss': 0.36703062057495117, 'lr': 0.0003159377083647996, 'sim_test_accuracy': 0.8552497625350952, 'sim_test_precision': 0.8558992743492126, 'sim_test_recall': 0.8552497625350952, 'sim_test_f1-score': 0.8553267121315002, 'sim_test_support': 1962.0, 'sim_test_score': 0.8552497625350952}] +Wrote results for fttransformer on sim to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results/fttransformer_sim_downstream_results.json + +Downstream pipeline complete. Results under /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..3905d86e8bd66dec1aa6a69efaa0eabb0348de33 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-15T19:29:55.164444+00:00", + "spec": { + "dataset": "sim", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 5e-05, + "sw_z_weight": 0.0, + "sw_s_y_weight": 1e-05, + "lambda_joint": 0.00025, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..489ae1af2301197297c34c2080ed6ec7ea574b0c --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "sim", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 5e-05, + "sw_z_weight": 0.0, + "sw_s_y_weight": 1e-05, + "lambda_joint": 0.00025, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "sim", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.00025", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "5e-05", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "1e-05", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..79aad71794be4d9c38826c2ae59eee39b7d8b89c --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "sim", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.00025, + "s_x": 5e-05, + "z": 0.0, + "s_y": 1e-05 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 50 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "sim", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "64", + "--model_num_layers", + "16", + "--model_num_heads", + "2", + "--model_dropout_prob", + "0.002", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.15", + "--mask_max_ctx_share", + "0.5", + "--mask_min_trgt_share", + "0.15", + "--mask_max_trgt_share", + "0.8", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.0005", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "1e-06", + "--vtjepa_kl_x_annealing_epochs", + "50", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "50", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "50", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.001", + "--recon_y_weight", + "0.1", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "p90", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.00025", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "50", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "5e-05", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "1e-05", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results/autoint_sim_downstream_results.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results/autoint_sim_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..350c74b3b1bdf1277c7ed131eaf4f0e0f8dbcff2 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results/autoint_sim_downstream_results.json @@ -0,0 +1,210 @@ +{ + "dataset": "sim", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpazbhnyrm", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "sim", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpazbhnyrm/split_data/split_config.json", + "sigma_aggregation": "p90", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "sim_val_loss": 0.38667839765548706, + "lr": 0.0003159377083647996, + "sim_val_accuracy": 0.8605405688285828, + "sim_val_precision": 0.8618879318237305, + "sim_val_recall": 0.8605405688285828, + "sim_val_f1-score": 0.8606148958206177, + "sim_val_support": 925.0, + "sim_val_score": 0.8605405688285828 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.3953714072704315, + "lr": 0.0003159377083647996, + "sim_test_accuracy": 0.8583078384399414, + "sim_test_precision": 0.8595095276832581, + "sim_test_recall": 0.8583078384399414, + "sim_test_f1-score": 0.8583861589431763, + "sim_test_support": 1962.0, + "sim_test_score": 0.8583078384399414 + } + } + }, + { + "seed": 1, + "val_metrics": { + "sim_val_loss": 0.3551470637321472, + "lr": 0.000531395198777318, + "sim_val_accuracy": 0.8735135197639465, + "sim_val_precision": 0.8760514259338379, + "sim_val_recall": 0.8735135197639465, + "sim_val_f1-score": 0.8737159371376038, + "sim_val_support": 925.0, + "sim_val_score": 0.8735135197639465 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.36710795760154724, + "lr": 0.0005313953151926398, + "sim_test_accuracy": 0.8542304039001465, + "sim_test_precision": 0.856368362903595, + "sim_test_recall": 0.8542304039001465, + "sim_test_f1-score": 0.8543589115142822, + "sim_test_support": 1962.0, + "sim_test_score": 0.8542304039001465 + } + } + }, + { + "seed": 2, + "val_metrics": { + "sim_val_loss": 0.38884830474853516, + "lr": 0.0006243449170142412, + "sim_val_accuracy": 0.8529729843139648, + "sim_val_precision": 0.8547060489654541, + "sim_val_recall": 0.8529729843139648, + "sim_val_f1-score": 0.8526897430419922, + "sim_val_support": 925.0, + "sim_val_score": 0.8529729843139648 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.406281054019928, + "lr": 0.0006243448005989194, + "sim_test_accuracy": 0.8598368763923645, + "sim_test_precision": 0.8612660765647888, + "sim_test_recall": 0.8598368763923645, + "sim_test_f1-score": 0.8593538999557495, + "sim_test_support": 1962.0, + "sim_test_score": 0.8598368763923645 + } + } + }, + { + "seed": 3, + "val_metrics": { + "sim_val_loss": 0.36793437600135803, + "lr": 0.0005000000237487257, + "sim_val_accuracy": 0.8605405688285828, + "sim_val_precision": 0.861944854259491, + "sim_val_recall": 0.8605405688285828, + "sim_val_f1-score": 0.8607043027877808, + "sim_val_support": 925.0, + "sim_val_score": 0.8605405688285828 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.3741350769996643, + "lr": 0.0005000000819563866, + "sim_test_accuracy": 0.8552497625350952, + "sim_test_precision": 0.8554317355155945, + "sim_test_recall": 0.8552497625350952, + "sim_test_f1-score": 0.8552965521812439, + "sim_test_support": 1962.0, + "sim_test_score": 0.8552497625350952 + } + } + }, + { + "seed": 4, + "val_metrics": { + "sim_val_loss": 0.3748020827770233, + "lr": 0.0006243449170142412, + "sim_val_accuracy": 0.8443243503570557, + "sim_val_precision": 0.8513504266738892, + "sim_val_recall": 0.8443243503570557, + "sim_val_f1-score": 0.8440408110618591, + "sim_val_support": 925.0, + "sim_val_score": 0.8443243503570557 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.3910144865512848, + "lr": 0.0006243448005989194, + "sim_test_accuracy": 0.8450560569763184, + "sim_test_precision": 0.8516185283660889, + "sim_test_recall": 0.8450560569763184, + "sim_test_f1-score": 0.8448898792266846, + "sim_test_support": 1962.0, + "sim_test_score": 0.8450560569763184 + } + } + } + ], + "summary": { + "0": { + "sim_test_loss": { + "mean": 0.38678199648857114, + "std": 0.015960244147636752, + "n": 5, + "ci95": 0.019814083560813295 + }, + "lr": { + "mean": 0.0005192045413423329, + "std": 0.00012643553678572396, + "n": 5, + "ci95": 0.00015696528622963237 + }, + "sim_test_accuracy": { + "mean": 0.8545361876487731, + "std": 0.005761905527537918, + "n": 5, + "ci95": 0.007153203706414047 + }, + "sim_test_precision": { + "mean": 0.8568388462066651, + "std": 0.0037478641112419124, + "n": 5, + "ci95": 0.004652841898143289 + }, + "sim_test_recall": { + "mean": 0.8545361876487731, + "std": 0.005761905527537918, + "n": 5, + "ci95": 0.007153203706414047 + }, + "sim_test_f1-score": { + "mean": 0.8544570803642273, + "std": 0.0057372072529309716, + "n": 5, + "ci95": 0.0071225416643836675 + }, + "sim_test_support": { + "mean": 1962.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "sim_test_score": { + "mean": 0.8545361876487731, + "std": 0.005761905527537918, + "n": 5, + "ci95": 0.007153203706414047 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results/dcnv2_sim_downstream_results.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results/dcnv2_sim_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..2f8d23852b680f9bb684cc521c830fbd7fb4bd42 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results/dcnv2_sim_downstream_results.json @@ -0,0 +1,209 @@ +{ + "dataset": "sim", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpsyiazfhq", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "sim", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpsyiazfhq/split_data/split_config.json", + "sigma_aggregation": "p90", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "sim_val_loss": 0.3876427710056305, + "lr": 0.0003454915131442249, + "sim_val_accuracy": 0.8508108258247375, + "sim_val_precision": 0.8522735238075256, + "sim_val_recall": 0.8508108258247375, + "sim_val_f1-score": 0.8509405851364136, + "sim_val_support": 925.0, + "sim_val_score": 0.8508108258247375 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.41305387020111084, + "lr": 0.0003454914840403944, + "sim_test_accuracy": 0.8516819477081299, + "sim_test_precision": 0.851963460445404, + "sim_test_recall": 0.8516819477081299, + "sim_test_f1-score": 0.8517601490020752, + "sim_test_support": 1962.0, + "sim_test_score": 0.8516819477081299 + } + } + }, + { + "seed": 1, + "val_metrics": { + "sim_val_loss": 0.4672321081161499, + "lr": 0.0002591231605038047, + "sim_val_accuracy": 0.8335134983062744, + "sim_val_precision": 0.8362681865692139, + "sim_val_recall": 0.8335134983062744, + "sim_val_f1-score": 0.832685649394989, + "sim_val_support": 925.0, + "sim_val_score": 0.8335134983062744 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.5182449817657471, + "lr": 0.00025912318960763514, + "sim_test_accuracy": 0.8414883017539978, + "sim_test_precision": 0.8461096882820129, + "sim_test_recall": 0.8414883017539978, + "sim_test_f1-score": 0.840985894203186, + "sim_test_support": 1962.0, + "sim_test_score": 0.8414883017539978 + } + } + }, + { + "seed": 2, + "val_metrics": { + "sim_val_loss": 0.40959325432777405, + "lr": 0.000406309321988374, + "sim_val_accuracy": 0.8540540337562561, + "sim_val_precision": 0.8541745543479919, + "sim_val_recall": 0.8540540337562561, + "sim_val_f1-score": 0.8539830446243286, + "sim_val_support": 925.0, + "sim_val_score": 0.8540540337562561 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.4582127034664154, + "lr": 0.0004063093801960349, + "sim_test_accuracy": 0.8465850949287415, + "sim_test_precision": 0.8478139638900757, + "sim_test_recall": 0.8465850949287415, + "sim_test_f1-score": 0.8464018106460571, + "sim_test_support": 1962.0, + "sim_test_score": 0.8465850949287415 + } + } + }, + { + "seed": 3, + "val_metrics": { + "sim_val_loss": 0.4338511526584625, + "lr": 0.0002871103642974049, + "sim_val_accuracy": 0.8399999737739563, + "sim_val_precision": 0.8406462073326111, + "sim_val_recall": 0.8399999737739563, + "sim_val_f1-score": 0.8400229811668396, + "sim_val_support": 925.0, + "sim_val_score": 0.8399999737739563 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.4340159595012665, + "lr": 0.00028711039340123534, + "sim_test_accuracy": 0.8567788004875183, + "sim_test_precision": 0.8579704165458679, + "sim_test_recall": 0.8567788004875183, + "sim_test_f1-score": 0.8567251563072205, + "sim_test_support": 1962.0, + "sim_test_score": 0.8567788004875183 + } + } + }, + { + "seed": 4, + "val_metrics": { + "sim_val_loss": 0.41764041781425476, + "lr": 0.0006545084179379046, + "sim_val_accuracy": 0.837837815284729, + "sim_val_precision": 0.8485930562019348, + "sim_val_recall": 0.837837815284729, + "sim_val_f1-score": 0.8372050523757935, + "sim_val_support": 925.0, + "sim_val_score": 0.837837815284729 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.4272391200065613, + "lr": 0.0006545084179379046, + "sim_test_accuracy": 0.8363914489746094, + "sim_test_precision": 0.8461126089096069, + "sim_test_recall": 0.8363914489746094, + "sim_test_f1-score": 0.8348156213760376, + "sim_test_support": 1962.0, + "sim_test_score": 0.8363914489746094 + } + } + } + ], + "summary": { + "0": { + "sim_test_loss": { + "mean": 0.4501533269882202, + "std": 0.04142243415247671, + "n": 5, + "ci95": 0.051424499775649675 + }, + "lr": { + "mean": 0.0003905085730366409, + "std": 0.0001580549148612422, + "n": 5, + "ci95": 0.00019621963557002356 + }, + "sim_test_accuracy": { + "mean": 0.8465851187705994, + "std": 0.008058803567764608, + "n": 5, + "ci95": 0.010004722096654886 + }, + "sim_test_precision": { + "mean": 0.8499940276145935, + "std": 0.005059832682881966, + "n": 5, + "ci95": 0.006281604883669802 + }, + "sim_test_recall": { + "mean": 0.8465851187705994, + "std": 0.008058803567764608, + "n": 5, + "ci95": 0.010004722096654886 + }, + "sim_test_f1-score": { + "mean": 0.8461377263069153, + "std": 0.008638670117607968, + "n": 5, + "ci95": 0.010724606088806606 + }, + "sim_test_support": { + "mean": 1962.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "sim_test_score": { + "mean": 0.8465851187705994, + "std": 0.008058803567764608, + "n": 5, + "ci95": 0.010004722096654886 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results/fttransformer_sim_downstream_results.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results/fttransformer_sim_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..de9ff9a4e84050991747729945b4ccd9d529e0c9 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results/fttransformer_sim_downstream_results.json @@ -0,0 +1,209 @@ +{ + "dataset": "sim", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp1tybfn2x", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "sim", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp1tybfn2x/split_data/split_config.json", + "sigma_aggregation": "p90", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "sim_val_loss": 0.35619404911994934, + "lr": 0.0002061073755612597, + "sim_val_accuracy": 0.8594594597816467, + "sim_val_precision": 0.86168372631073, + "sim_val_recall": 0.8594594597816467, + "sim_val_f1-score": 0.8594022989273071, + "sim_val_support": 925.0, + "sim_val_score": 0.8594594597816467 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.37115034461021423, + "lr": 0.000206107331905514, + "sim_test_accuracy": 0.8552497625350952, + "sim_test_precision": 0.8561407923698425, + "sim_test_recall": 0.8552497625350952, + "sim_test_f1-score": 0.8551145792007446, + "sim_test_support": 1962.0, + "sim_test_score": 0.8552497625350952 + } + } + }, + { + "seed": 1, + "val_metrics": { + "sim_val_loss": 0.34779229760169983, + "lr": 0.0003159377083647996, + "sim_val_accuracy": 0.8681080937385559, + "sim_val_precision": 0.8688489198684692, + "sim_val_recall": 0.8681080937385559, + "sim_val_f1-score": 0.8680978417396545, + "sim_val_support": 925.0, + "sim_val_score": 0.8681080937385559 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.3603042960166931, + "lr": 0.0003159377083647996, + "sim_test_accuracy": 0.8598368763923645, + "sim_test_precision": 0.860596239566803, + "sim_test_recall": 0.8598368763923645, + "sim_test_f1-score": 0.8599119782447815, + "sim_test_support": 1962.0, + "sim_test_score": 0.8598368763923645 + } + } + }, + { + "seed": 2, + "val_metrics": { + "sim_val_loss": 0.36381348967552185, + "lr": 0.0002871103642974049, + "sim_val_accuracy": 0.8659459352493286, + "sim_val_precision": 0.8692029714584351, + "sim_val_recall": 0.8659459352493286, + "sim_val_f1-score": 0.8660420179367065, + "sim_val_support": 925.0, + "sim_val_score": 0.8659459352493286 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.37445560097694397, + "lr": 0.00028711039340123534, + "sim_test_accuracy": 0.8572884798049927, + "sim_test_precision": 0.8600521087646484, + "sim_test_recall": 0.8572884798049927, + "sim_test_f1-score": 0.8574735522270203, + "sim_test_support": 1962.0, + "sim_test_score": 0.8572884798049927 + } + } + }, + { + "seed": 3, + "val_metrics": { + "sim_val_loss": 0.35820046067237854, + "lr": 6.18466583546251e-05, + "sim_val_accuracy": 0.8594594597816467, + "sim_val_precision": 0.8600558638572693, + "sim_val_recall": 0.8594594597816467, + "sim_val_f1-score": 0.8595176935195923, + "sim_val_support": 925.0, + "sim_val_score": 0.8594594597816467 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.39219722151756287, + "lr": 6.184667290654033e-05, + "sim_test_accuracy": 0.8623853325843811, + "sim_test_precision": 0.863053560256958, + "sim_test_recall": 0.8623853325843811, + "sim_test_f1-score": 0.8623414635658264, + "sim_test_support": 1962.0, + "sim_test_score": 0.8623853325843811 + } + } + }, + { + "seed": 4, + "val_metrics": { + "sim_val_loss": 0.3629548251628876, + "lr": 0.0003159377083647996, + "sim_val_accuracy": 0.861621618270874, + "sim_val_precision": 0.8630407452583313, + "sim_val_recall": 0.861621618270874, + "sim_val_f1-score": 0.8617360591888428, + "sim_val_support": 925.0, + "sim_val_score": 0.861621618270874 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.36703062057495117, + "lr": 0.0003159377083647996, + "sim_test_accuracy": 0.8552497625350952, + "sim_test_precision": 0.8558992743492126, + "sim_test_recall": 0.8552497625350952, + "sim_test_f1-score": 0.8553267121315002, + "sim_test_support": 1962.0, + "sim_test_score": 0.8552497625350952 + } + } + } + ], + "summary": { + "0": { + "sim_test_loss": { + "mean": 0.3730276167392731, + "std": 0.011946588322507943, + "n": 5, + "ci95": 0.014831270568242447 + }, + "lr": { + "mean": 0.00023738796298857778, + "std": 0.00010794172905797472, + "n": 5, + "ci95": 0.00013400587230804164 + }, + "sim_test_accuracy": { + "mean": 0.8580020427703857, + "std": 0.0030918889495962455, + "n": 5, + "ci95": 0.003838471732722618 + }, + "sim_test_precision": { + "mean": 0.8591483950614929, + "std": 0.00307269127665353, + "n": 5, + "ci95": 0.003814638494813381 + }, + "sim_test_recall": { + "mean": 0.8580020427703857, + "std": 0.0030918889495962455, + "n": 5, + "ci95": 0.003838471732722618 + }, + "sim_test_f1-score": { + "mean": 0.8580336570739746, + "std": 0.003092231038550163, + "n": 5, + "ci95": 0.0038388964241656462 + }, + "sim_test_support": { + "mean": 1962.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "sim_test_score": { + "mean": 0.8580020427703857, + "std": 0.0030918889495962455, + "n": 5, + "ci95": 0.003838471732722618 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results/mlp_sim_downstream_results.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results/mlp_sim_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..26361a292a115067291bb79f12c7f88c62c8c88a --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results/mlp_sim_downstream_results.json @@ -0,0 +1,205 @@ +{ + "dataset": "sim", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpd85dycll", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "sim", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpd85dycll/split_data/split_config.json", + "sigma_aggregation": "p90", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "sim_val_loss": 0.42959123849868774, + "lr": 0.0010000000474974513, + "sim_val_accuracy": 0.8183783888816833, + "sim_val_precision": 0.8403798341751099, + "sim_val_recall": 0.8183783888816833, + "sim_val_f1-score": 0.8181406259536743, + "sim_val_support": 925.0, + "sim_val_score": 0.8183783888816833 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.4512270987033844, + "lr": 0.0010000001639127731, + "sim_test_accuracy": 0.8216106295585632, + "sim_test_precision": 0.8440932631492615, + "sim_test_recall": 0.8216106295585632, + "sim_test_f1-score": 0.8210008144378662, + "sim_test_support": 1962.0, + "sim_test_score": 0.8216106295585632 + } + } + }, + { + "seed": 1, + "val_metrics": { + "sim_val_loss": 0.4242367744445801, + "lr": 0.0010000000474974513, + "sim_val_accuracy": 0.8291891813278198, + "sim_val_precision": 0.8540714383125305, + "sim_val_recall": 0.8291891813278198, + "sim_val_f1-score": 0.8309029936790466, + "sim_val_support": 925.0, + "sim_val_score": 0.8291891813278198 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.42744868993759155, + "lr": 0.0010000001639127731, + "sim_test_accuracy": 0.8287461996078491, + "sim_test_precision": 0.8508343696594238, + "sim_test_recall": 0.8287461996078491, + "sim_test_f1-score": 0.829196572303772, + "sim_test_support": 1962.0, + "sim_test_score": 0.8287461996078491 + } + } + }, + { + "seed": 2, + "val_metrics": { + "sim_val_loss": 0.3884407877922058, + "lr": 0.0010000000474974513, + "sim_val_accuracy": 0.8475675582885742, + "sim_val_precision": 0.8557260632514954, + "sim_val_recall": 0.8475675582885742, + "sim_val_f1-score": 0.8471525311470032, + "sim_val_support": 925.0, + "sim_val_score": 0.8475675582885742 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.40132883191108704, + "lr": 0.0010000001639127731, + "sim_test_accuracy": 0.8394495248794556, + "sim_test_precision": 0.8511507511138916, + "sim_test_recall": 0.8394495248794556, + "sim_test_f1-score": 0.8390764594078064, + "sim_test_support": 1962.0, + "sim_test_score": 0.8394495248794556 + } + } + }, + { + "seed": 3, + "val_metrics": { + "sim_val_loss": 0.3681979179382324, + "lr": 0.0010000000474974513, + "sim_val_accuracy": 0.8572973012924194, + "sim_val_precision": 0.8630771040916443, + "sim_val_recall": 0.8572973012924194, + "sim_val_f1-score": 0.8573291897773743, + "sim_val_support": 925.0, + "sim_val_score": 0.8572973012924194 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.36493369936943054, + "lr": 0.0010000001639127731, + "sim_test_accuracy": 0.860856294631958, + "sim_test_precision": 0.8632360100746155, + "sim_test_recall": 0.860856294631958, + "sim_test_f1-score": 0.8608131408691406, + "sim_test_support": 1962.0, + "sim_test_score": 0.860856294631958 + } + } + }, + { + "seed": 4, + "val_metrics": { + "sim_val_loss": 0.4813116490840912, + "lr": 0.0010000000474974513, + "sim_val_accuracy": 0.8129729628562927, + "sim_val_precision": 0.8407442569732666, + "sim_val_recall": 0.8129729628562927, + "sim_val_f1-score": 0.8110881447792053, + "sim_val_support": 925.0, + "sim_val_score": 0.8129729628562927 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.4823322296142578, + "lr": 0.0010000001639127731, + "sim_test_accuracy": 0.8047910332679749, + "sim_test_precision": 0.827860414981842, + "sim_test_recall": 0.8047910332679749, + "sim_test_f1-score": 0.800818681716919, + "sim_test_support": 1962.0, + "sim_test_score": 0.8047910332679749 + } + } + } + ], + "summary": { + "0": { + "sim_test_loss": { + "mean": 0.42545410990715027, + "std": 0.04513339930817911, + "n": 5, + "ci95": 0.05603153291412715 + }, + "lr": { + "mean": 0.0010000001639127731, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "sim_test_accuracy": { + "mean": 0.8310907363891602, + "std": 0.020874035696480813, + "n": 5, + "ci95": 0.0259143834966154 + }, + "sim_test_precision": { + "mean": 0.8474349617958069, + "std": 0.012936054495143389, + "n": 5, + "ci95": 0.016059658131981555 + }, + "sim_test_recall": { + "mean": 0.8310907363891602, + "std": 0.020874035696480813, + "n": 5, + "ci95": 0.0259143834966154 + }, + "sim_test_f1-score": { + "mean": 0.8301811337471008, + "std": 0.022163302393904613, + "n": 5, + "ci95": 0.027514962901204993 + }, + "sim_test_support": { + "mean": 1962.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "sim_test_score": { + "mean": 0.8310907363891602, + "std": 0.020874035696480813, + "n": 5, + "ci95": 0.0259143834966154 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results/resnet_sim_downstream_results.json b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results/resnet_sim_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..4c0450b4fa465aef80a493dfda94a0dc44ba53b8 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/downstream_results/resnet_sim_downstream_results.json @@ -0,0 +1,208 @@ +{ + "dataset": "sim", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpyj2ieumd", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "sim", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpyj2ieumd/split_data/split_config.json", + "sigma_aggregation": "p90", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "sim_val_loss": 0.3943820595741272, + "lr": 0.0006545084179379046, + "sim_val_accuracy": 0.8529729843139648, + "sim_val_precision": 0.853134274482727, + "sim_val_recall": 0.8529729843139648, + "sim_val_f1-score": 0.8530251979827881, + "sim_val_support": 925.0, + "sim_val_score": 0.8529729843139648 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.38298365473747253, + "lr": 0.0006545084179379046, + "sim_test_accuracy": 0.8491335511207581, + "sim_test_precision": 0.8492569923400879, + "sim_test_recall": 0.8491335511207581, + "sim_test_f1-score": 0.8490909934043884, + "sim_test_support": 1962.0, + "sim_test_score": 0.8491335511207581 + } + } + }, + { + "seed": 1, + "val_metrics": { + "sim_val_loss": 0.374656617641449, + "lr": 9.549150126986206e-05, + "sim_val_accuracy": 0.8421621918678284, + "sim_val_precision": 0.8424528241157532, + "sim_val_recall": 0.8421621918678284, + "sim_val_f1-score": 0.8418724536895752, + "sim_val_support": 925.0, + "sim_val_score": 0.8421621918678284 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.37531402707099915, + "lr": 9.549151582177728e-05, + "sim_test_accuracy": 0.8430173397064209, + "sim_test_precision": 0.8445531129837036, + "sim_test_recall": 0.8430173397064209, + "sim_test_f1-score": 0.842693567276001, + "sim_test_support": 1962.0, + "sim_test_score": 0.8430173397064209 + } + } + }, + { + "seed": 2, + "val_metrics": { + "sim_val_loss": 0.3847091794013977, + "lr": 0.0006545084179379046, + "sim_val_accuracy": 0.8399999737739563, + "sim_val_precision": 0.8406162858009338, + "sim_val_recall": 0.8399999737739563, + "sim_val_f1-score": 0.8400732278823853, + "sim_val_support": 925.0, + "sim_val_score": 0.8399999737739563 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.38975510001182556, + "lr": 0.0006545084179379046, + "sim_test_accuracy": 0.8338429927825928, + "sim_test_precision": 0.8344905972480774, + "sim_test_recall": 0.8338429927825928, + "sim_test_f1-score": 0.833795964717865, + "sim_test_support": 1962.0, + "sim_test_score": 0.8338429927825928 + } + } + }, + { + "seed": 3, + "val_metrics": { + "sim_val_loss": 0.4110054075717926, + "lr": 9.549150126986206e-05, + "sim_val_accuracy": 0.8454053997993469, + "sim_val_precision": 0.8454695343971252, + "sim_val_recall": 0.8454053997993469, + "sim_val_f1-score": 0.8451192378997803, + "sim_val_support": 925.0, + "sim_val_score": 0.8454053997993469 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.40992864966392517, + "lr": 9.549151582177728e-05, + "sim_test_accuracy": 0.8389398455619812, + "sim_test_precision": 0.8392660021781921, + "sim_test_recall": 0.8389398455619812, + "sim_test_f1-score": 0.8385645151138306, + "sim_test_support": 1962.0, + "sim_test_score": 0.8389398455619812 + } + } + }, + { + "seed": 4, + "val_metrics": { + "sim_val_loss": 0.4437926113605499, + "lr": 9.549150126986206e-05, + "sim_val_accuracy": 0.8302702903747559, + "sim_val_precision": 0.8343738913536072, + "sim_val_recall": 0.8302702903747559, + "sim_val_f1-score": 0.8307524919509888, + "sim_val_support": 925.0, + "sim_val_score": 0.8302702903747559 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.4607800245285034, + "lr": 9.549151582177728e-05, + "sim_test_accuracy": 0.831294596195221, + "sim_test_precision": 0.8364284038543701, + "sim_test_recall": 0.831294596195221, + "sim_test_f1-score": 0.8315088748931885, + "sim_test_support": 1962.0, + "sim_test_score": 0.831294596195221 + } + } + } + ], + "summary": { + "0": { + "sim_test_loss": { + "mean": 0.4037522912025452, + "std": 0.03437405366262303, + "n": 5, + "ci95": 0.04267418250590751 + }, + "lr": { + "mean": 0.00031909827666822823, + "std": 0.00030618616731566036, + "n": 5, + "ci95": 0.00038011939217458474 + }, + "sim_test_accuracy": { + "mean": 0.8392456650733948, + "std": 0.007150134785161039, + "n": 5, + "ci95": 0.008876641659973376 + }, + "sim_test_precision": { + "mean": 0.8407990217208863, + "std": 0.006059685160702775, + "n": 5, + "ci95": 0.007522886681164184 + }, + "sim_test_recall": { + "mean": 0.8392456650733948, + "std": 0.007150134785161039, + "n": 5, + "ci95": 0.008876641659973376 + }, + "sim_test_f1-score": { + "mean": 0.8391307830810547, + "std": 0.007049355497277663, + "n": 5, + "ci95": 0.008751527707276345 + }, + "sim_test_support": { + "mean": 1962.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "sim_test_score": { + "mean": 0.8392456650733948, + "std": 0.007150134785161039, + "n": 5, + "ci95": 0.008876641659973376 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..9e0db0892fe1ca318f7b162b200f763713675e06 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:48a48c0eb88ba1aace0c6cd7498b4a61dc0e50303086550a946f41e2e017a074 +size 81920128 diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/vtjepa_best_model.pth b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..722047f43b4eb4777ad801d8109fec42371d879e --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b3b08c08f09cb06bcdaf6306318582f6c5615de4cac5701ab650fc22da15f8db +size 15822934 diff --git a/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/vtjepa_config.yaml b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..2ded16b78c95258f910244ae952aae9ef16a2187 --- /dev/null +++ b/JEPA/artifacts/djepa_real_ebe7d90_v1/sim/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260915_185549/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: sim +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.0005 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.5 +mask_max_trgt_share: 0.8 +mask_min_ctx_share: 0.15 +mask_min_trgt_share: 0.15 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 64 +model_dim_hidden: 64 +model_dropout_prob: 0.002 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 2 +model_num_layers: 16 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.001 +recon_y_weight: 0.1 +res_dir: ./ +safe_exit: false +sigma_aggregation: p90 +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 5.0e-05 +sw_s_y_weight: 1.0e-05 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 50 +vtjepa_kl_x_end_weight: 1.0e-06 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 50 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 50 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 50 +wjepa_joint_end_weight: 0.00025 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..062e506c99e7fad4132e20a7048e4da1fb4129c4 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-15T21:16:59.318681+00:00", + "spec": { + "dataset": "adult", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.000125, + "sw_z_weight": 0.0, + "sw_s_y_weight": 2.5e-05, + "lambda_joint": 0.0225, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..c32f6acc893f8dc987735b01e48c7c2bb88914da --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "adult", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.000125, + "sw_z_weight": 0.0, + "sw_s_y_weight": 2.5e-05, + "lambda_joint": 0.0225, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "adult", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.0225", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.000125", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "2.5e-05", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..96a34db8547e68e4296130cf54ef5e51a09b078b --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "adult", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.0225, + "s_x": 0.000125, + "z": 0.0, + "s_y": 2.5e-05 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 15 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "adult", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.0225", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.000125", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "2.5e-05", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/autoint_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/autoint_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..5b23cd75c24719d59f28914a8ae1502565a7897b --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/autoint_adult_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "adult", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp4beb1bes", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp4beb1bes/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.3169384002685547, + "lr": 6.184666563058272e-05, + "adult_val_accuracy": 0.8558013439178467, + "adult_val_precision": 0.8503040075302124, + "adult_val_recall": 0.8558013439178467, + "adult_val_f1-score": 0.8516346216201782, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9102479219436646, + "adult_val_score": 0.8558013439178467 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.33180513978004456, + "lr": 6.18466583546251e-05, + "adult_test_accuracy": 0.8551185727119446, + "adult_test_precision": 0.8494860529899597, + "adult_test_recall": 0.8551185727119446, + "adult_test_f1-score": 0.8504323363304138, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9060373306274414, + "adult_test_score": 0.8551185727119446 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.3093198835849762, + "lr": 0.0002320865314686671, + "adult_val_accuracy": 0.8614357113838196, + "adult_val_precision": 0.8569654822349548, + "adult_val_recall": 0.8614357113838196, + "adult_val_f1-score": 0.8582793474197388, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9129817485809326, + "adult_val_score": 0.8614357113838196 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32346802949905396, + "lr": 0.00023208664788398892, + "adult_test_accuracy": 0.8532703518867493, + "adult_test_precision": 0.8479794859886169, + "adult_test_recall": 0.8532703518867493, + "adult_test_f1-score": 0.8492692112922668, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9079545140266418, + "adult_test_score": 0.8532703518867493 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.3187181055545807, + "lr": 3.5111756005790085e-05, + "adult_val_accuracy": 0.8568447232246399, + "adult_val_precision": 0.851814329624176, + "adult_val_recall": 0.8568447232246399, + "adult_val_f1-score": 0.8532171249389648, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9111811518669128, + "adult_val_score": 0.8568447232246399 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3409523367881775, + "lr": 3.511177419568412e-05, + "adult_test_accuracy": 0.8494712114334106, + "adult_test_precision": 0.8436961770057678, + "adult_test_recall": 0.8494712114334106, + "adult_test_f1-score": 0.8450287580490112, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9035384654998779, + "adult_test_score": 0.8494712114334106 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.3087535798549652, + "lr": 9.866357686405536e-07, + "adult_val_accuracy": 0.8566360473632812, + "adult_val_precision": 0.8512328267097473, + "adult_val_recall": 0.8566360473632812, + "adult_val_f1-score": 0.8525477051734924, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9130929112434387, + "adult_val_score": 0.8566360473632812 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.33292198181152344, + "lr": 9.866350865195272e-07, + "adult_test_accuracy": 0.8548105359077454, + "adult_test_precision": 0.8492507934570312, + "adult_test_recall": 0.8548105359077454, + "adult_test_f1-score": 0.8502878546714783, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9082933068275452, + "adult_test_score": 0.8548105359077454 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.3125564157962799, + "lr": 2.4471735741826706e-05, + "adult_val_accuracy": 0.856218695640564, + "adult_val_precision": 0.8498589992523193, + "adult_val_recall": 0.856218695640564, + "adult_val_f1-score": 0.8503601551055908, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.912286102771759, + "adult_val_score": 0.856218695640564 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3343268632888794, + "lr": 2.4471757569699548e-05, + "adult_test_accuracy": 0.8541944622993469, + "adult_test_precision": 0.8479709625244141, + "adult_test_recall": 0.8541944622993469, + "adult_test_f1-score": 0.8480719327926636, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9048032760620117, + "adult_test_score": 0.8541944622993469 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.3326948702335358, + "std": 0.006261528270357307, + "n": 5, + "ci95": 0.007773467825404478 + }, + "lr": { + "mean": 7.090069461810345e-05, + "std": 9.271914020784134e-05, + "n": 5, + "ci95": 0.0001151075619377013 + }, + "adult_test_accuracy": { + "mean": 0.8533730268478393, + "std": 0.002292532689947326, + "n": 5, + "ci95": 0.002846098960913354 + }, + "adult_test_precision": { + "mean": 0.847676694393158, + "std": 0.0023331503139197842, + "n": 5, + "ci95": 0.0028965243170662636 + }, + "adult_test_recall": { + "mean": 0.8533730268478393, + "std": 0.002292532689947326, + "n": 5, + "ci95": 0.002846098960913354 + }, + "adult_test_f1-score": { + "mean": 0.8486180186271668, + "std": 0.0022183977476336473, + "n": 5, + "ci95": 0.002754063029119867 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9061253786087036, + "std": 0.0020306080470386995, + "n": 5, + "ci95": 0.0025209286995300927 + }, + "adult_test_score": { + "mean": 0.8533730268478393, + "std": 0.002292532689947326, + "n": 5, + "ci95": 0.002846098960913354 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/dcnv2_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/dcnv2_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..f4c5194591b4dc9e84f3fa5feecc6abe59a48493 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/dcnv2_adult_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "adult", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpj0d7wk4l", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpj0d7wk4l/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.3531002104282379, + "lr": 6.184666563058272e-05, + "adult_val_accuracy": 0.8505843281745911, + "adult_val_precision": 0.8465133309364319, + "adult_val_recall": 0.8505843281745911, + "adult_val_f1-score": 0.8480285406112671, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9006636738777161, + "adult_val_score": 0.8505843281745911 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3771567940711975, + "lr": 6.18466583546251e-05, + "adult_test_accuracy": 0.8489577770233154, + "adult_test_precision": 0.8448347449302673, + "adult_test_recall": 0.8489577770233154, + "adult_test_f1-score": 0.8463144302368164, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.8958659172058105, + "adult_test_score": 0.8489577770233154 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.3206446170806885, + "lr": 0.00028711033519357443, + "adult_val_accuracy": 0.856218695640564, + "adult_val_precision": 0.8509974479675293, + "adult_val_recall": 0.856218695640564, + "adult_val_f1-score": 0.8523884415626526, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9107237458229065, + "adult_val_score": 0.856218695640564 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3439182937145233, + "lr": 0.0002871101605705917, + "adult_test_accuracy": 0.8526542782783508, + "adult_test_precision": 0.8471377491950989, + "adult_test_recall": 0.8526542782783508, + "adult_test_f1-score": 0.8483723402023315, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9033202528953552, + "adult_test_score": 0.8526542782783508 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.3322629928588867, + "lr": 0.0002320865314686671, + "adult_val_accuracy": 0.8541318774223328, + "adult_val_precision": 0.8484034538269043, + "adult_val_recall": 0.8541318774223328, + "adult_val_f1-score": 0.8497503399848938, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9058226943016052, + "adult_val_score": 0.8541318774223328 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3570747673511505, + "lr": 0.00023208664788398892, + "adult_test_accuracy": 0.851216733455658, + "adult_test_precision": 0.8454431891441345, + "adult_test_recall": 0.851216733455658, + "adult_test_f1-score": 0.8466501235961914, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.8996191024780273, + "adult_test_score": 0.851216733455658 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.3296643793582916, + "lr": 0.0001147433795267716, + "adult_val_accuracy": 0.85454922914505, + "adult_val_precision": 0.8503016233444214, + "adult_val_recall": 0.85454922914505, + "adult_val_f1-score": 0.8517816662788391, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9074008464813232, + "adult_val_score": 0.85454922914505 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.36518388986587524, + "lr": 0.00011474348866613582, + "adult_test_accuracy": 0.8474175930023193, + "adult_test_precision": 0.842684268951416, + "adult_test_recall": 0.8474175930023193, + "adult_test_f1-score": 0.8442500829696655, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.8978670835494995, + "adult_test_score": 0.8474175930023193 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.3197488486766815, + "lr": 0.0004063092346768826, + "adult_val_accuracy": 0.8539232015609741, + "adult_val_precision": 0.8485876321792603, + "adult_val_recall": 0.8539232015609741, + "adult_val_f1-score": 0.8500589728355408, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9095285534858704, + "adult_val_score": 0.8539232015609741 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3436557650566101, + "lr": 0.0004063095839228481, + "adult_test_accuracy": 0.853065013885498, + "adult_test_precision": 0.8474377989768982, + "adult_test_recall": 0.853065013885498, + "adult_test_f1-score": 0.8485819101333618, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9032925963401794, + "adult_test_score": 0.853065013885498 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.3573979020118713, + "std": 0.014332532977729795, + "n": 5, + "ci95": 0.01779333720912411 + }, + "lr": { + "mean": 0.00022041930787963793, + "std": 0.000137342323526537, + "n": 5, + "ci95": 0.0001705056795884921 + }, + "adult_test_accuracy": { + "mean": 0.8506622791290284, + "std": 0.0024229078609392576, + "n": 5, + "ci95": 0.0030079551648907825 + }, + "adult_test_precision": { + "mean": 0.845507550239563, + "std": 0.001924261896769542, + "n": 5, + "ci95": 0.002388903682349143 + }, + "adult_test_recall": { + "mean": 0.8506622791290284, + "std": 0.0024229078609392576, + "n": 5, + "ci95": 0.0030079551648907825 + }, + "adult_test_f1-score": { + "mean": 0.8468337774276733, + "std": 0.0017608286866608383, + "n": 5, + "ci95": 0.002186007081786469 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.8999929904937745, + "std": 0.003303410371830299, + "n": 5, + "ci95": 0.004101068162719473 + }, + "adult_test_score": { + "mean": 0.8506622791290284, + "std": 0.0024229078609392576, + "n": 5, + "ci95": 0.0030079551648907825 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/fttransformer_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/fttransformer_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..4c1b05b8dca4f99217b5fb354ae2763cdc3ea948 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/fttransformer_adult_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "adult", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp6p8eqt8e", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp6p8eqt8e/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.3129483759403229, + "lr": 3.942650891985977e-06, + "adult_val_accuracy": 0.859557569026947, + "adult_val_precision": 0.8543425798416138, + "adult_val_recall": 0.859557569026947, + "adult_val_f1-score": 0.8555525541305542, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9119886755943298, + "adult_val_score": 0.859557569026947 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32320699095726013, + "lr": 3.9426490729965735e-06, + "adult_test_accuracy": 0.8551185727119446, + "adult_test_precision": 0.8495733141899109, + "adult_test_recall": 0.8551185727119446, + "adult_test_f1-score": 0.8505922555923462, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9081900715827942, + "adult_test_score": 0.8551185727119446 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.3062211871147156, + "lr": 0.0001147433795267716, + "adult_val_accuracy": 0.8583055138587952, + "adult_val_precision": 0.8520856499671936, + "adult_val_recall": 0.8583055138587952, + "adult_val_f1-score": 0.8522975444793701, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9134030342102051, + "adult_val_score": 0.8583055138587952 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3214559555053711, + "lr": 0.00011474348866613582, + "adult_test_accuracy": 0.8547078967094421, + "adult_test_precision": 0.8484896421432495, + "adult_test_recall": 0.8547078967094421, + "adult_test_f1-score": 0.8483326435089111, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9084650874137878, + "adult_test_score": 0.8547078967094421 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.3081017732620239, + "lr": 2.4471735741826706e-05, + "adult_val_accuracy": 0.8620617985725403, + "adult_val_precision": 0.8562042117118835, + "adult_val_recall": 0.8620617985725403, + "adult_val_f1-score": 0.8560980558395386, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9130727648735046, + "adult_val_score": 0.8620617985725403 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3231293559074402, + "lr": 2.4471757569699548e-05, + "adult_test_accuracy": 0.8527569770812988, + "adult_test_precision": 0.8463513255119324, + "adult_test_recall": 0.8527569770812988, + "adult_test_f1-score": 0.8463989496231079, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9080302715301514, + "adult_test_score": 0.8527569770812988 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.3088344633579254, + "lr": 4.758648356073536e-05, + "adult_val_accuracy": 0.8580968379974365, + "adult_val_precision": 0.8522157669067383, + "adult_val_recall": 0.8580968379974365, + "adult_val_f1-score": 0.8530885577201843, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9125703573226929, + "adult_val_score": 0.8580968379974365 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.321559339761734, + "lr": 4.758644354296848e-05, + "adult_test_accuracy": 0.8524489402770996, + "adult_test_precision": 0.8461886048316956, + "adult_test_recall": 0.8524489402770996, + "adult_test_f1-score": 0.8467285633087158, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9080004692077637, + "adult_test_score": 0.8524489402770996 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.30886310338974, + "lr": 0.0, + "adult_val_accuracy": 0.8591402173042297, + "adult_val_precision": 0.8534672856330872, + "adult_val_recall": 0.8591402173042297, + "adult_val_f1-score": 0.8544185757637024, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9129846692085266, + "adult_val_score": 0.8591402173042297 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3232496976852417, + "lr": 0.0, + "adult_test_accuracy": 0.8543998599052429, + "adult_test_precision": 0.8486642837524414, + "adult_test_recall": 0.8543998599052429, + "adult_test_f1-score": 0.8495959639549255, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9083588123321533, + "adult_test_score": 0.8543998599052429 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.32252026796340943, + "std": 0.0009261193673131382, + "n": 5, + "ci95": 0.0011497447258002752 + }, + "lr": { + "mean": 3.814886777036009e-05, + "std": 4.681951687011684e-05, + "n": 5, + "ci95": 5.8124788753859144e-05 + }, + "adult_test_accuracy": { + "mean": 0.8538864493370056, + "std": 0.0012040176578911573, + "n": 5, + "ci95": 0.001494745710746697 + }, + "adult_test_precision": { + "mean": 0.8478534340858459, + "std": 0.0015040123054500946, + "n": 5, + "ci95": 0.0018671785482111331 + }, + "adult_test_recall": { + "mean": 0.8538864493370056, + "std": 0.0012040176578911573, + "n": 5, + "ci95": 0.001494745710746697 + }, + "adult_test_f1-score": { + "mean": 0.8483296751976013, + "std": 0.0018037479449201739, + "n": 5, + "ci95": 0.002239289836213789 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9082089424133301, + "std": 0.00020236830532947893, + "n": 5, + "ci95": 0.00025123315625796367 + }, + "adult_test_score": { + "mean": 0.8538864493370056, + "std": 0.0012040176578911573, + "n": 5, + "ci95": 0.001494745710746697 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/mlp_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/mlp_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..9f047037cf05010f459571d208947c24ac4cfdff --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/mlp_adult_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "adult", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpfo4f46z9", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpfo4f46z9/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.3160879611968994, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8568447232246399, + "adult_val_precision": 0.8522077798843384, + "adult_val_recall": 0.8568447232246399, + "adult_val_f1-score": 0.8451939225196838, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9109522104263306, + "adult_val_score": 0.8568447232246399 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3285183310508728, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8460827469825745, + "adult_test_precision": 0.8401943445205688, + "adult_test_recall": 0.8460827469825745, + "adult_test_f1-score": 0.8336168527603149, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9060031175613403, + "adult_test_score": 0.8460827469825745 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.30662330985069275, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8597663044929504, + "adult_val_precision": 0.8537729382514954, + "adult_val_recall": 0.8597663044929504, + "adult_val_f1-score": 0.8541386127471924, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9136431813240051, + "adult_val_score": 0.8597663044929504 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3239207863807678, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8521409034729004, + "adult_test_precision": 0.8461464047431946, + "adult_test_recall": 0.8521409034729004, + "adult_test_f1-score": 0.8470972180366516, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9071840047836304, + "adult_test_score": 0.8521409034729004 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.3131153881549835, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8574708104133606, + "adult_val_precision": 0.8515012860298157, + "adult_val_recall": 0.8574708104133606, + "adult_val_f1-score": 0.848116934299469, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9116476774215698, + "adult_val_score": 0.8574708104133606 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3252173662185669, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8501899838447571, + "adult_test_precision": 0.8436403274536133, + "adult_test_recall": 0.8501899838447571, + "adult_test_f1-score": 0.8408549427986145, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9070713520050049, + "adult_test_score": 0.8501899838447571 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.30879148840904236, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8580968379974365, + "adult_val_precision": 0.8519830703735352, + "adult_val_recall": 0.8580968379974365, + "adult_val_f1-score": 0.8525181412696838, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9122077822685242, + "adult_val_score": 0.8580968379974365 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32266324758529663, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8520382046699524, + "adult_test_precision": 0.8457303047180176, + "adult_test_recall": 0.8520382046699524, + "adult_test_f1-score": 0.8462733626365662, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9068222641944885, + "adult_test_score": 0.8520382046699524 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.3125172555446625, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.860601007938385, + "adult_val_precision": 0.854961633682251, + "adult_val_recall": 0.860601007938385, + "adult_val_f1-score": 0.8518827557563782, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9111517071723938, + "adult_val_score": 0.860601007938385 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3269507586956024, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8503953218460083, + "adult_test_precision": 0.843791127204895, + "adult_test_recall": 0.8503953218460083, + "adult_test_f1-score": 0.8413733243942261, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9050098657608032, + "adult_test_score": 0.8503953218460083 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.3254540979862213, + "std": 0.0023358858021682704, + "n": 5, + "ci95": 0.0028999203298235724 + }, + "lr": { + "mean": 0.0009999996982514858, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_accuracy": { + "mean": 0.8501694321632385, + "std": 0.002456188820760928, + "n": 5, + "ci95": 0.0030492723097158064 + }, + "adult_test_precision": { + "mean": 0.8439005017280579, + "std": 0.0023562424547087864, + "n": 5, + "ci95": 0.0029251924002709376 + }, + "adult_test_recall": { + "mean": 0.8501694321632385, + "std": 0.002456188820760928, + "n": 5, + "ci95": 0.0030492723097158064 + }, + "adult_test_f1-score": { + "mean": 0.8418431401252746, + "std": 0.005387514983432253, + "n": 5, + "ci95": 0.00668841097162456 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9064181208610534, + "std": 0.0009126925975381138, + "n": 5, + "ci95": 0.0011330758618522552 + }, + "adult_test_score": { + "mean": 0.8501694321632385, + "std": 0.002456188820760928, + "n": 5, + "ci95": 0.0030492723097158064 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/resnet_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/resnet_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..17126aa12a762abb6f09bebbbad89fcbc388df72 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/resnet_adult_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "adult", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmph_72b6sy", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmph_72b6sy/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.3075842261314392, + "lr": 0.0, + "adult_val_accuracy": 0.8580968379974365, + "adult_val_precision": 0.8517798185348511, + "adult_val_recall": 0.8580968379974365, + "adult_val_f1-score": 0.8516930341720581, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9137892723083496, + "adult_val_score": 0.8580968379974365 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3297344744205475, + "lr": 0.0, + "adult_test_accuracy": 0.8532703518867493, + "adult_test_precision": 0.8468907475471497, + "adult_test_recall": 0.8532703518867493, + "adult_test_f1-score": 0.8467732667922974, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9057698845863342, + "adult_test_score": 0.8532703518867493 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.30536195635795593, + "lr": 0.0009755284409038723, + "adult_val_accuracy": 0.8597663044929504, + "adult_val_precision": 0.853827178478241, + "adult_val_recall": 0.8597663044929504, + "adult_val_f1-score": 0.851453423500061, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9132334589958191, + "adult_val_score": 0.8597663044929504 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.318236380815506, + "lr": 0.0009755276259966195, + "adult_test_accuracy": 0.8550159335136414, + "adult_test_precision": 0.8491094708442688, + "adult_test_recall": 0.8550159335136414, + "adult_test_f1-score": 0.8463206887245178, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9095875024795532, + "adult_test_score": 0.8550159335136414 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.32872506976127625, + "lr": 2.4471735741826706e-05, + "adult_val_accuracy": 0.84620201587677, + "adult_val_precision": 0.8571587800979614, + "adult_val_recall": 0.84620201587677, + "adult_val_f1-score": 0.8501240015029907, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9124865531921387, + "adult_val_score": 0.84620201587677 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.33793482184410095, + "lr": 2.4471757569699548e-05, + "adult_test_accuracy": 0.8407434225082397, + "adult_test_precision": 0.8506613373756409, + "adult_test_recall": 0.8407434225082397, + "adult_test_f1-score": 0.844384491443634, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9072854518890381, + "adult_test_score": 0.8407434225082397 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.30318400263786316, + "lr": 0.0003454915713518858, + "adult_val_accuracy": 0.8603923320770264, + "adult_val_precision": 0.8552676439285278, + "adult_val_recall": 0.8603923320770264, + "adult_val_f1-score": 0.8564637899398804, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9147263765335083, + "adult_val_score": 0.8603923320770264 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3200436532497406, + "lr": 0.0003454913094174117, + "adult_test_accuracy": 0.8566588163375854, + "adult_test_precision": 0.8512600660324097, + "adult_test_recall": 0.8566588163375854, + "adult_test_f1-score": 0.852246105670929, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9085975885391235, + "adult_test_score": 0.8566588163375854 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.3075188994407654, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8568447232246399, + "adult_val_precision": 0.8505741953849792, + "adult_val_recall": 0.8568447232246399, + "adult_val_f1-score": 0.8510998487472534, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9117891788482666, + "adult_val_score": 0.8568447232246399 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.318407267332077, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8521409034729004, + "adult_test_precision": 0.845607340335846, + "adult_test_recall": 0.8521409034729004, + "adult_test_f1-score": 0.8453704118728638, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9075758457183838, + "adult_test_score": 0.8521409034729004 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.3248713195323944, + "std": 0.008709450506146969, + "n": 5, + "ci95": 0.01081247745969577 + }, + "lr": { + "mean": 0.0004690980782470433, + "std": 0.0004927858617878117, + "n": 5, + "ci95": 0.000611776370883203 + }, + "adult_test_accuracy": { + "mean": 0.8515658855438233, + "std": 0.006288925431931841, + "n": 5, + "ci95": 0.007807480440985131 + }, + "adult_test_precision": { + "mean": 0.848705792427063, + "std": 0.002418998720496966, + "n": 5, + "ci95": 0.0030031021040818106 + }, + "adult_test_recall": { + "mean": 0.8515658855438233, + "std": 0.006288925431931841, + "n": 5, + "ci95": 0.007807480440985131 + }, + "adult_test_f1-score": { + "mean": 0.8470189929008484, + "std": 0.0030630513906943786, + "n": 5, + "ci95": 0.003802670914358816 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9077632546424865, + "std": 0.001437092114211967, + "n": 5, + "ci95": 0.001784099476936763 + }, + "adult_test_score": { + "mean": 0.8515658855438233, + "std": 0.006288925431931841, + "n": 5, + "ci95": 0.007807480440985131 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..bf24706eba3ae400b2e32c474f731b8daafb407f --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6746e6620716629c1113cc6787d89eabbbde622e573e2c57ac00e8034a88bdbc +size 175049856 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..5e998c5400b657dd34baaf7e1b9cad19341e2d6a --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ea30be8ee9ce337012ee77c6821a14f9369282b7f0fe41ad99df4d4a3747ade2 +size 15770390 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..8f705e4292c45dca78f5dcc08b67d693cfaaa38c --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.0225_sx0.000125_sy2.5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: adult +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.001 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.3 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.1 +mask_min_trgt_share: 0.1 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 256 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.1 +recon_y_weight: 1.0 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.000125 +sw_s_y_weight: 2.5e-05 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 15 +vtjepa_kl_x_end_weight: 0.0001 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 15 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 15 +wjepa_joint_end_weight: 0.0225 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..f2ee46961fe95a4d2da25ff5847c6a94704d7cc7 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-15T21:12:28.601961+00:00", + "spec": { + "dataset": "adult", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.00025, + "sw_z_weight": 0.0, + "sw_s_y_weight": 5e-05, + "lambda_joint": 0.045, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..db1a35ef53c98aa2dd0c5ce1ed83a6904886d322 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "adult", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.00025, + "sw_z_weight": 0.0, + "sw_s_y_weight": 5e-05, + "lambda_joint": 0.045, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "adult", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.045", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.00025", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "5e-05", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..e8f36dec894f25db84248e0564053a06e27ba67c --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "adult", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.045, + "s_x": 0.00025, + "z": 0.0, + "s_y": 5e-05 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 15 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "adult", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.045", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.00025", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "5e-05", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/autoint_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/autoint_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..f9721cb1454065d25c49a1c46550a3df04b533f3 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/autoint_adult_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "adult", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpdyxhwb_d", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpdyxhwb_d/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.32278579473495483, + "lr": 4.758648356073536e-05, + "adult_val_accuracy": 0.8532971739768982, + "adult_val_precision": 0.8473170399665833, + "adult_val_recall": 0.8532971739768982, + "adult_val_f1-score": 0.8486082553863525, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9078726172447205, + "adult_val_score": 0.8532971739768982 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3365781605243683, + "lr": 4.758644354296848e-05, + "adult_test_accuracy": 0.8534756898880005, + "adult_test_precision": 0.8475935459136963, + "adult_test_recall": 0.8534756898880005, + "adult_test_f1-score": 0.8484913110733032, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.904283344745636, + "adult_test_score": 0.8534756898880005 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.31339266896247864, + "lr": 6.184666563058272e-05, + "adult_val_accuracy": 0.8541318774223328, + "adult_val_precision": 0.8495864272117615, + "adult_val_recall": 0.8541318774223328, + "adult_val_f1-score": 0.8510977625846863, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9099447727203369, + "adult_val_score": 0.8541318774223328 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3257199823856354, + "lr": 6.18466583546251e-05, + "adult_test_accuracy": 0.8542971611022949, + "adult_test_precision": 0.8498412370681763, + "adult_test_recall": 0.8542971611022949, + "adult_test_f1-score": 0.8512354493141174, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9070783853530884, + "adult_test_score": 0.8542971611022949 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.3089822232723236, + "lr": 0.0003454915713518858, + "adult_val_accuracy": 0.8580968379974365, + "adult_val_precision": 0.8518032431602478, + "adult_val_recall": 0.8580968379974365, + "adult_val_f1-score": 0.850139319896698, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9109589457511902, + "adult_val_score": 0.8580968379974365 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3235788941383362, + "lr": 0.0003454913094174117, + "adult_test_accuracy": 0.8520382046699524, + "adult_test_precision": 0.845557451248169, + "adult_test_recall": 0.8520382046699524, + "adult_test_f1-score": 0.8436344265937805, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9064478874206543, + "adult_test_score": 0.8520382046699524 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.3150617778301239, + "lr": 0.0003454915713518858, + "adult_val_accuracy": 0.8576794862747192, + "adult_val_precision": 0.8525338172912598, + "adult_val_recall": 0.8576794862747192, + "adult_val_f1-score": 0.8469738364219666, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9112944602966309, + "adult_val_score": 0.8576794862747192 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32834744453430176, + "lr": 0.0003454913094174117, + "adult_test_accuracy": 0.8507033586502075, + "adult_test_precision": 0.8453572392463684, + "adult_test_recall": 0.8507033586502075, + "adult_test_f1-score": 0.8392548561096191, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9075199365615845, + "adult_test_score": 0.8507033586502075 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.3150351941585541, + "lr": 3.5111756005790085e-05, + "adult_val_accuracy": 0.85454922914505, + "adult_val_precision": 0.8497071862220764, + "adult_val_recall": 0.85454922914505, + "adult_val_f1-score": 0.8512096405029297, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9099196195602417, + "adult_val_score": 0.85454922914505 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32739031314849854, + "lr": 3.511177419568412e-05, + "adult_test_accuracy": 0.8569668531417847, + "adult_test_precision": 0.8514852523803711, + "adult_test_recall": 0.8569668531417847, + "adult_test_f1-score": 0.8523667454719543, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.907568633556366, + "adult_test_score": 0.8569668531417847 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.32832295894622804, + "std": 0.004957328380797661, + "n": 5, + "ci95": 0.006154349386319406 + }, + "lr": { + "mean": 0.0001671054989856202, + "std": 0.00016311772302207106, + "n": 5, + "ci95": 0.00020250493440524743 + }, + "adult_test_accuracy": { + "mean": 0.853496253490448, + "std": 0.0023776619690311683, + "n": 5, + "ci95": 0.002951783976357733 + }, + "adult_test_precision": { + "mean": 0.8479669451713562, + "std": 0.00267614969302472, + "n": 5, + "ci95": 0.0033223460210468135 + }, + "adult_test_recall": { + "mean": 0.853496253490448, + "std": 0.0023776619690311683, + "n": 5, + "ci95": 0.002951783976357733 + }, + "adult_test_f1-score": { + "mean": 0.8469965577125549, + "std": 0.005483595434548363, + "n": 5, + "ci95": 0.00680769148321104 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9065796375274658, + "std": 0.0013602705101225826, + "n": 5, + "ci95": 0.0016887281487401222 + }, + "adult_test_score": { + "mean": 0.853496253490448, + "std": 0.0023776619690311683, + "n": 5, + "ci95": 0.002951783976357733 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/dcnv2_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/dcnv2_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..84acee83e015acac5493131e189d926bc2f88b16 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/dcnv2_adult_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "adult", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpxd_h1fyh", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpxd_h1fyh/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.3234885632991791, + "lr": 0.0003454915713518858, + "adult_val_accuracy": 0.8560100197792053, + "adult_val_precision": 0.8513171076774597, + "adult_val_recall": 0.8560100197792053, + "adult_val_f1-score": 0.8527821898460388, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9101245999336243, + "adult_val_score": 0.8560100197792053 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.34689992666244507, + "lr": 0.0003454913094174117, + "adult_test_accuracy": 0.8539891242980957, + "adult_test_precision": 0.8495069146156311, + "adult_test_recall": 0.8539891242980957, + "adult_test_f1-score": 0.8509085178375244, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9041141867637634, + "adult_test_score": 0.8539891242980957 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.3368978798389435, + "lr": 0.00020610730280168355, + "adult_val_accuracy": 0.8524624109268188, + "adult_val_precision": 0.8484298586845398, + "adult_val_recall": 0.8524624109268188, + "adult_val_f1-score": 0.8499131798744202, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9038345813751221, + "adult_val_score": 0.8524624109268188 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.360731840133667, + "lr": 0.00020610728824976832, + "adult_test_accuracy": 0.8496765494346619, + "adult_test_precision": 0.8453969359397888, + "adult_test_recall": 0.8496765494346619, + "adult_test_f1-score": 0.8468839526176453, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.8984493017196655, + "adult_test_score": 0.8496765494346619 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.34177181124687195, + "lr": 0.0002320865314686671, + "adult_val_accuracy": 0.8507930040359497, + "adult_val_precision": 0.8451995253562927, + "adult_val_recall": 0.8507930040359497, + "adult_val_f1-score": 0.846762478351593, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9026687145233154, + "adult_val_score": 0.8507930040359497 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3622872531414032, + "lr": 0.00023208664788398892, + "adult_test_accuracy": 0.8521409034729004, + "adult_test_precision": 0.8469328284263611, + "adult_test_recall": 0.8521409034729004, + "adult_test_f1-score": 0.8483052849769592, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9014325141906738, + "adult_test_score": 0.8521409034729004 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.3199891746044159, + "lr": 0.00043733330676332116, + "adult_val_accuracy": 0.8537145256996155, + "adult_val_precision": 0.8512647151947021, + "adult_val_recall": 0.8537145256996155, + "adult_val_f1-score": 0.8523194789886475, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9096705913543701, + "adult_val_score": 0.8537145256996155 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3429616391658783, + "lr": 0.000437333743320778, + "adult_test_accuracy": 0.8503953218460083, + "adult_test_precision": 0.8472055196762085, + "adult_test_recall": 0.8503953218460083, + "adult_test_f1-score": 0.8484789729118347, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9009769558906555, + "adult_test_score": 0.8503953218460083 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.35007110238075256, + "lr": 6.184666563058272e-05, + "adult_val_accuracy": 0.8510016798973083, + "adult_val_precision": 0.8461163640022278, + "adult_val_recall": 0.8510016798973083, + "adult_val_f1-score": 0.8477153182029724, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9038039445877075, + "adult_val_score": 0.8510016798973083 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3974786400794983, + "lr": 6.18466583546251e-05, + "adult_test_accuracy": 0.8502926230430603, + "adult_test_precision": 0.8457162976264954, + "adult_test_recall": 0.8502926230430603, + "adult_test_f1-score": 0.847210168838501, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.8954405784606934, + "adult_test_score": 0.8502926230430603 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.36207185983657836, + "std": 0.02151087065051992, + "n": 5, + "ci95": 0.026704991765327004 + }, + "lr": { + "mean": 0.0002565731294453144, + "std": 0.00014286849697225503, + "n": 5, + "ci95": 0.00017736623017983237 + }, + "adult_test_accuracy": { + "mean": 0.8512989044189453, + "std": 0.0017612030528615255, + "n": 5, + "ci95": 0.002186471844299758 + }, + "adult_test_precision": { + "mean": 0.846951699256897, + "std": 0.0016230690530319018, + "n": 5, + "ci95": 0.0020149833263362776 + }, + "adult_test_recall": { + "mean": 0.8512989044189453, + "std": 0.0017612030528615255, + "n": 5, + "ci95": 0.002186471844299758 + }, + "adult_test_f1-score": { + "mean": 0.8483573794364929, + "std": 0.001582157126664565, + "n": 5, + "ci95": 0.001964192604078042 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9000827074050903, + "std": 0.0032822195158663725, + "n": 5, + "ci95": 0.0040747604579681905 + }, + "adult_test_score": { + "mean": 0.8512989044189453, + "std": 0.0017612030528615255, + "n": 5, + "ci95": 0.002186471844299758 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/fttransformer_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/fttransformer_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..3389309d9db370f41efddd326eb9ca31d1cb1043 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/fttransformer_adult_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "adult", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp1evvomvb", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp1evvomvb/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.3094805181026459, + "lr": 3.942650891985977e-06, + "adult_val_accuracy": 0.8601836562156677, + "adult_val_precision": 0.8550272583961487, + "adult_val_recall": 0.8601836562156677, + "adult_val_f1-score": 0.8562228679656982, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9132583737373352, + "adult_val_score": 0.8601836562156677 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3231779634952545, + "lr": 3.9426490729965735e-06, + "adult_test_accuracy": 0.8525515794754028, + "adult_test_precision": 0.8467822670936584, + "adult_test_recall": 0.8525515794754028, + "adult_test_f1-score": 0.8478773832321167, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9077754616737366, + "adult_test_score": 0.8525515794754028 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.31014251708984375, + "lr": 2.4471735741826706e-05, + "adult_val_accuracy": 0.8576794862747192, + "adult_val_precision": 0.8524425029754639, + "adult_val_recall": 0.8576794862747192, + "adult_val_f1-score": 0.8537549376487732, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9121675491333008, + "adult_val_score": 0.8576794862747192 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32064536213874817, + "lr": 2.4471757569699548e-05, + "adult_test_accuracy": 0.8562480807304382, + "adult_test_precision": 0.8511675000190735, + "adult_test_recall": 0.8562480807304382, + "adult_test_f1-score": 0.8523664474487305, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9081677794456482, + "adult_test_score": 0.8562480807304382 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.30822569131851196, + "lr": 0.0, + "adult_val_accuracy": 0.8570533990859985, + "adult_val_precision": 0.851487934589386, + "adult_val_recall": 0.8570533990859985, + "adult_val_f1-score": 0.8527049422264099, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9131405353546143, + "adult_val_score": 0.8570533990859985 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32104435563087463, + "lr": 0.0, + "adult_test_accuracy": 0.8566588163375854, + "adult_test_precision": 0.8511611819267273, + "adult_test_recall": 0.8566588163375854, + "adult_test_f1-score": 0.8520619869232178, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9082602262496948, + "adult_test_score": 0.8566588163375854 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.30827492475509644, + "lr": 0.00028711033519357443, + "adult_val_accuracy": 0.8580968379974365, + "adult_val_precision": 0.8521385192871094, + "adult_val_recall": 0.8580968379974365, + "adult_val_f1-score": 0.8529190421104431, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9124599695205688, + "adult_val_score": 0.8580968379974365 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3206789791584015, + "lr": 0.0002871101605705917, + "adult_test_accuracy": 0.8548105359077454, + "adult_test_precision": 0.8490400910377502, + "adult_test_recall": 0.8548105359077454, + "adult_test_f1-score": 0.849885106086731, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9080359935760498, + "adult_test_score": 0.8548105359077454 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.3089883029460907, + "lr": 0.0002320865314686671, + "adult_val_accuracy": 0.8580968379974365, + "adult_val_precision": 0.8540231585502625, + "adult_val_recall": 0.8580968379974365, + "adult_val_f1-score": 0.8554217219352722, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.913561999797821, + "adult_val_score": 0.8580968379974365 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3222561776638031, + "lr": 0.00023208664788398892, + "adult_test_accuracy": 0.8540918231010437, + "adult_test_precision": 0.8498237729072571, + "adult_test_recall": 0.8540918231010437, + "adult_test_f1-score": 0.8512226939201355, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9086015820503235, + "adult_test_score": 0.8540918231010437 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.32156056761741636, + "std": 0.0011159207920573785, + "n": 5, + "ci95": 0.0013853765403925758 + }, + "lr": { + "mean": 0.00010952224301945535, + "std": 0.00013868602962185608, + "n": 5, + "ci95": 0.0001721738435969837 + }, + "adult_test_accuracy": { + "mean": 0.8548721671104431, + "std": 0.0016645657016613783, + "n": 5, + "ci95": 0.0020664999607832454 + }, + "adult_test_precision": { + "mean": 0.8495949625968933, + "std": 0.0018164119530337088, + "n": 5, + "ci95": 0.0022550117583006483 + }, + "adult_test_recall": { + "mean": 0.8548721671104431, + "std": 0.0016645657016613783, + "n": 5, + "ci95": 0.0020664999607832454 + }, + "adult_test_f1-score": { + "mean": 0.8506827235221863, + "std": 0.0018394965029413473, + "n": 5, + "ci95": 0.002283670417692237 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9081682085990905, + "std": 0.0003033195262926628, + "n": 5, + "ci95": 0.00037656055784579165 + }, + "adult_test_score": { + "mean": 0.8548721671104431, + "std": 0.0016645657016613783, + "n": 5, + "ci95": 0.0020664999607832454 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/mlp_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/mlp_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..4f9450a84408fae18f9a260d232a0a9c602b68fc --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/mlp_adult_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "adult", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpwbitwwuv", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpwbitwwuv/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.31118321418762207, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8583055138587952, + "adult_val_precision": 0.8531601428985596, + "adult_val_recall": 0.8583055138587952, + "adult_val_f1-score": 0.8478253483772278, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.913169264793396, + "adult_val_score": 0.8583055138587952 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3252963125705719, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8509086966514587, + "adult_test_precision": 0.8451564311981201, + "adult_test_recall": 0.8509086966514587, + "adult_test_f1-score": 0.840156614780426, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9070892333984375, + "adult_test_score": 0.8509086966514587 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.3133411407470703, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.855592668056488, + "adult_val_precision": 0.8490106463432312, + "adult_val_recall": 0.855592668056488, + "adult_val_f1-score": 0.8470990061759949, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.91008460521698, + "adult_val_score": 0.855592668056488 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3276556134223938, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8495738506317139, + "adult_test_precision": 0.8428592681884766, + "adult_test_recall": 0.8495738506317139, + "adult_test_f1-score": 0.8404355645179749, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9044444561004639, + "adult_test_score": 0.8495738506317139 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.31216827034950256, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8599749803543091, + "adult_val_precision": 0.8538382053375244, + "adult_val_recall": 0.8599749803543091, + "adult_val_f1-score": 0.8529024124145508, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9103741645812988, + "adult_val_score": 0.8599749803543091 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.327891081571579, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8499845862388611, + "adult_test_precision": 0.8431604504585266, + "adult_test_recall": 0.8499845862388611, + "adult_test_f1-score": 0.8427016735076904, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9035426378250122, + "adult_test_score": 0.8499845862388611 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.3108472228050232, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8585141897201538, + "adult_val_precision": 0.8522019386291504, + "adult_val_recall": 0.8585141897201538, + "adult_val_f1-score": 0.8512124419212341, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.912028431892395, + "adult_val_score": 0.8585141897201538 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32450437545776367, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8498819470405579, + "adult_test_precision": 0.8430442810058594, + "adult_test_recall": 0.8498819470405579, + "adult_test_f1-score": 0.8425785303115845, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9071388244628906, + "adult_test_score": 0.8498819470405579 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.31239211559295654, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8578881621360779, + "adult_val_precision": 0.8515350818634033, + "adult_val_recall": 0.8578881621360779, + "adult_val_f1-score": 0.8513845801353455, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9103643298149109, + "adult_val_score": 0.8578881621360779 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32648536562919617, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8515247702598572, + "adult_test_precision": 0.8450044393539429, + "adult_test_recall": 0.8515247702598572, + "adult_test_f1-score": 0.8451725840568542, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9051697254180908, + "adult_test_score": 0.8515247702598572 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.3263665497303009, + "std": 0.0014674042008371066, + "n": 5, + "ci95": 0.0018217308697736985 + }, + "lr": { + "mean": 0.0009999996982514858, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_accuracy": { + "mean": 0.8503747701644897, + "std": 0.0008130448846950365, + "n": 5, + "ci95": 0.001009366719895989 + }, + "adult_test_precision": { + "mean": 0.8438449740409851, + "std": 0.0011341932800522972, + "n": 5, + "ci95": 0.0014080611936250818 + }, + "adult_test_recall": { + "mean": 0.8503747701644897, + "std": 0.0008130448846950365, + "n": 5, + "ci95": 0.001009366719895989 + }, + "adult_test_f1-score": { + "mean": 0.842208993434906, + "std": 0.002032207469515377, + "n": 5, + "ci95": 0.0025229143264609076 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.905476975440979, + "std": 0.00160181761373319, + "n": 5, + "ci95": 0.0019886004094988446 + }, + "adult_test_score": { + "mean": 0.8503747701644897, + "std": 0.0008130448846950365, + "n": 5, + "ci95": 0.001009366719895989 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/resnet_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/resnet_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..18bb63abe99c8b8421c7b0234d458103e114d9ae --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/resnet_adult_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "adult", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp18vt_1cn", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp18vt_1cn/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.30639025568962097, + "lr": 0.000904508342500776, + "adult_val_accuracy": 0.8591402173042297, + "adult_val_precision": 0.8533318638801575, + "adult_val_recall": 0.8591402173042297, + "adult_val_f1-score": 0.8541408181190491, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9127894639968872, + "adult_val_score": 0.8591402173042297 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32352277636528015, + "lr": 0.0009045081678777933, + "adult_test_accuracy": 0.8550159335136414, + "adult_test_precision": 0.8489621877670288, + "adult_test_recall": 0.8550159335136414, + "adult_test_f1-score": 0.8492968678474426, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9065967798233032, + "adult_test_score": 0.8550159335136414 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.31435495615005493, + "lr": 0.0, + "adult_val_accuracy": 0.8499582409858704, + "adult_val_precision": 0.8426545262336731, + "adult_val_recall": 0.8499582409858704, + "adult_val_f1-score": 0.8403261303901672, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9087414145469666, + "adult_val_score": 0.8499582409858704 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3291483223438263, + "lr": 0.0, + "adult_test_accuracy": 0.8492658138275146, + "adult_test_precision": 0.8430268168449402, + "adult_test_recall": 0.8492658138275146, + "adult_test_f1-score": 0.8387151956558228, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9042423963546753, + "adult_test_score": 0.8492658138275146 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.31962329149246216, + "lr": 2.4471735741826706e-05, + "adult_val_accuracy": 0.8484975099563599, + "adult_val_precision": 0.8554238677024841, + "adult_val_recall": 0.8484975099563599, + "adult_val_f1-score": 0.8512298464775085, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9136029481887817, + "adult_val_score": 0.8484975099563599 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.33114585280418396, + "lr": 2.4471757569699548e-05, + "adult_test_accuracy": 0.845569372177124, + "adult_test_precision": 0.8525738716125488, + "adult_test_recall": 0.845569372177124, + "adult_test_f1-score": 0.8483086228370667, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9083555340766907, + "adult_test_score": 0.845569372177124 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.31106337904930115, + "lr": 9.54915230977349e-05, + "adult_val_accuracy": 0.8614357113838196, + "adult_val_precision": 0.8559917211532593, + "adult_val_recall": 0.8614357113838196, + "adult_val_f1-score": 0.8526389598846436, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9113643765449524, + "adult_val_score": 0.8614357113838196 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3260580003261566, + "lr": 9.54914212343283e-05, + "adult_test_accuracy": 0.8506006598472595, + "adult_test_precision": 0.844619870185852, + "adult_test_recall": 0.8506006598472595, + "adult_test_f1-score": 0.8401259779930115, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.906965970993042, + "adult_test_score": 0.8506006598472595 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.3103778660297394, + "lr": 0.0006545087671838701, + "adult_val_accuracy": 0.8566360473632812, + "adult_val_precision": 0.8509407639503479, + "adult_val_recall": 0.8566360473632812, + "adult_val_f1-score": 0.846389651298523, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9131426811218262, + "adult_val_score": 0.8566360473632812 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.327822744846344, + "lr": 0.0006545083015225828, + "adult_test_accuracy": 0.8508060574531555, + "adult_test_precision": 0.8453177213668823, + "adult_test_recall": 0.8508060574531555, + "adult_test_f1-score": 0.8396003842353821, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.906296968460083, + "adult_test_score": 0.8508060574531555 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.3275395393371582, + "std": 0.0029156143245806145, + "n": 5, + "ci95": 0.0036196329657587727 + }, + "lr": { + "mean": 0.0003357959296408808, + "std": 0.00041606462513370555, + "n": 5, + "ci95": 0.0005165296453386892 + }, + "adult_test_accuracy": { + "mean": 0.850251567363739, + "std": 0.003392018994322789, + "n": 5, + "ci95": 0.004211072661023762 + }, + "adult_test_precision": { + "mean": 0.8469000935554505, + "std": 0.0038454382979458204, + "n": 5, + "ci95": 0.004773976830093306 + }, + "adult_test_recall": { + "mean": 0.850251567363739, + "std": 0.003392018994322789, + "n": 5, + "ci95": 0.004211072661023762 + }, + "adult_test_f1-score": { + "mean": 0.8432094097137451, + "std": 0.005142705928907988, + "n": 5, + "ci95": 0.006384489113166916 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9064915299415588, + "std": 0.0014838412420513476, + "n": 5, + "ci95": 0.0018421368801767243 + }, + "adult_test_score": { + "mean": 0.850251567363739, + "std": 0.003392018994322789, + "n": 5, + "ci95": 0.004211072661023762 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..acf73f0000eb35394b2e3e2c50cc828ec560b55a --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0b2a9829c79945fb9a80ceb2267c0d5424e420a0bed3d8e6db38027a308dffbc +size 175049856 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..11171d58970ca13c123b16f5f86489e9c373285e --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ea42733c74e0de75e75b00f9d527e7212ee15fd4d268f9dea0a5f2993ebf25be +size 15771926 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..54f06f0981bf70e26823d59fa97bce148b6f7571 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.045_sx0.00025_sy5e-05/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: adult +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.001 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.3 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.1 +mask_min_trgt_share: 0.1 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 256 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.1 +recon_y_weight: 1.0 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.00025 +sw_s_y_weight: 5.0e-05 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 15 +vtjepa_kl_x_end_weight: 0.0001 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 15 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 15 +wjepa_joint_end_weight: 0.045 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..baa97124a3a5b21826c2d0dbff22e797f84af5a1 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-15T22:29:17.179546+00:00", + "spec": { + "dataset": "adult", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.00025, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0001, + "lambda_joint": 0.09, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..a8ef8b3fb92df380e37d99627f0b69b5598e057b --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "adult", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.00025, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0001, + "lambda_joint": 0.09, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "adult", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.09", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.00025", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "0.0001", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..31dd26c4b37588abe6c228119314a2972da0799d --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "adult", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.09, + "s_x": 0.00025, + "z": 0.0, + "s_y": 0.0001 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 15 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "adult", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.09", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.00025", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "0.0001", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/downstream_results/autoint_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/downstream_results/autoint_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..6067bda22534994eba9fe3b6fcf4d2f441e18a45 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/downstream_results/autoint_adult_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "adult", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmprd_1e095", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmprd_1e095/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.31842491030693054, + "lr": 1.570841959619429e-05, + "adult_val_accuracy": 0.8535058498382568, + "adult_val_precision": 0.8480098247528076, + "adult_val_recall": 0.8535058498382568, + "adult_val_f1-score": 0.8494662642478943, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9093296527862549, + "adult_val_score": 0.8535058498382568 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.33525848388671875, + "lr": 1.5708421415183693e-05, + "adult_test_accuracy": 0.8509086966514587, + "adult_test_precision": 0.8450875282287598, + "adult_test_recall": 0.8509086966514587, + "adult_test_f1-score": 0.8462917804718018, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9040221571922302, + "adult_test_score": 0.8509086966514587 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.31809237599372864, + "lr": 3.5111756005790085e-05, + "adult_val_accuracy": 0.8549666404724121, + "adult_val_precision": 0.8501964807510376, + "adult_val_recall": 0.8549666404724121, + "adult_val_f1-score": 0.8516891598701477, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9085416793823242, + "adult_val_score": 0.8549666404724121 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3326101303100586, + "lr": 3.511177419568412e-05, + "adult_test_accuracy": 0.8522435426712036, + "adult_test_precision": 0.8473129272460938, + "adult_test_recall": 0.8522435426712036, + "adult_test_f1-score": 0.8487333655357361, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.904654324054718, + "adult_test_score": 0.8522435426712036 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.31052538752555847, + "lr": 8.856373824528418e-06, + "adult_val_accuracy": 0.8572620749473572, + "adult_val_precision": 0.8520403504371643, + "adult_val_recall": 0.8572620749473572, + "adult_val_f1-score": 0.8533796072006226, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9130572080612183, + "adult_val_score": 0.8572620749473572 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.33057260513305664, + "lr": 8.856371096044313e-06, + "adult_test_accuracy": 0.8539891242980957, + "adult_test_precision": 0.8485177159309387, + "adult_test_recall": 0.8539891242980957, + "adult_test_f1-score": 0.8496800065040588, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9058446884155273, + "adult_test_score": 0.8539891242980957 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.30909404158592224, + "lr": 4.758648356073536e-05, + "adult_val_accuracy": 0.8612270355224609, + "adult_val_precision": 0.855715274810791, + "adult_val_recall": 0.8612270355224609, + "adult_val_f1-score": 0.8565753102302551, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9145788550376892, + "adult_val_score": 0.8612270355224609 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.33231499791145325, + "lr": 4.758644354296848e-05, + "adult_test_accuracy": 0.8541944622993469, + "adult_test_precision": 0.8483242988586426, + "adult_test_recall": 0.8541944622993469, + "adult_test_f1-score": 0.8491389155387878, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9056450128555298, + "adult_test_score": 0.8541944622993469 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.31253084540367126, + "lr": 0.0, + "adult_val_accuracy": 0.855592668056488, + "adult_val_precision": 0.8497631549835205, + "adult_val_recall": 0.855592668056488, + "adult_val_f1-score": 0.8509491086006165, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9105568528175354, + "adult_val_score": 0.855592668056488 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3291236162185669, + "lr": 0.0, + "adult_test_accuracy": 0.8534756898880005, + "adult_test_precision": 0.8475159406661987, + "adult_test_recall": 0.8534756898880005, + "adult_test_f1-score": 0.8483263254165649, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9053313136100769, + "adult_test_score": 0.8534756898880005 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.3319759666919708, + "std": 0.0023128635759741304, + "n": 5, + "ci95": 0.002871339043137292 + }, + "lr": { + "mean": 2.145260204997612e-05, + "std": 1.9502726956835346e-05, + "n": 5, + "ci95": 2.4211951772910718e-05 + }, + "adult_test_accuracy": { + "mean": 0.8529623031616211, + "std": 0.001375687106410938, + "n": 5, + "ci95": 0.0017078673125433292 + }, + "adult_test_precision": { + "mean": 0.8473516821861267, + "std": 0.0013656952278975976, + "n": 5, + "ci95": 0.001695462745673208 + }, + "adult_test_recall": { + "mean": 0.8529623031616211, + "std": 0.001375687106410938, + "n": 5, + "ci95": 0.0017078673125433292 + }, + "adult_test_f1-score": { + "mean": 0.8484340786933899, + "std": 0.0012980520156099867, + "n": 5, + "ci95": 0.001611486069114221 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9050994992256165, + "std": 0.0007525750374064488, + "n": 5, + "ci95": 0.0009342955244930599 + }, + "adult_test_score": { + "mean": 0.8529623031616211, + "std": 0.001375687106410938, + "n": 5, + "ci95": 0.0017078673125433292 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/downstream_results/dcnv2_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/downstream_results/dcnv2_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..eb6871a7fc3368f5744b1f44f0d41ff2980af4d4 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/downstream_results/dcnv2_adult_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "adult", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpf774dx96", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpf774dx96/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.35762733221054077, + "lr": 4.758648356073536e-05, + "adult_val_accuracy": 0.8507930040359497, + "adult_val_precision": 0.8470856547355652, + "adult_val_recall": 0.8507930040359497, + "adult_val_f1-score": 0.8485233187675476, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.8990230560302734, + "adult_val_score": 0.8507930040359497 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3927539885044098, + "lr": 4.758644354296848e-05, + "adult_test_accuracy": 0.8494712114334106, + "adult_test_precision": 0.8455278277397156, + "adult_test_recall": 0.8494712114334106, + "adult_test_f1-score": 0.8469728827476501, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.893765389919281, + "adult_test_score": 0.8494712114334106 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.3265947699546814, + "lr": 0.0001812879927456379, + "adult_val_accuracy": 0.8522537350654602, + "adult_val_precision": 0.847538948059082, + "adult_val_recall": 0.8522537350654602, + "adult_val_f1-score": 0.8491012454032898, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9061258435249329, + "adult_val_score": 0.8522537350654602 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3490794897079468, + "lr": 0.0001812881964724511, + "adult_test_accuracy": 0.8469042181968689, + "adult_test_precision": 0.8419470191001892, + "adult_test_recall": 0.8469042181968689, + "adult_test_f1-score": 0.8435306549072266, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.8982154726982117, + "adult_test_score": 0.8469042181968689 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.3282468318939209, + "lr": 0.0001812879927456379, + "adult_val_accuracy": 0.8535058498382568, + "adult_val_precision": 0.8495426177978516, + "adult_val_recall": 0.8535058498382568, + "adult_val_f1-score": 0.8510000109672546, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9047613739967346, + "adult_val_score": 0.8535058498382568 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3552078604698181, + "lr": 0.0001812881964724511, + "adult_test_accuracy": 0.8492658138275146, + "adult_test_precision": 0.8444361686706543, + "adult_test_recall": 0.8492658138275146, + "adult_test_f1-score": 0.8459572196006775, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.8962456583976746, + "adult_test_score": 0.8492658138275146 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.3272431194782257, + "lr": 0.00028711033519357443, + "adult_val_accuracy": 0.845158576965332, + "adult_val_precision": 0.8456279635429382, + "adult_val_recall": 0.845158576965332, + "adult_val_f1-score": 0.8453889489173889, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9061076641082764, + "adult_val_score": 0.845158576965332 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3443303108215332, + "lr": 0.0002871101605705917, + "adult_test_accuracy": 0.8438237905502319, + "adult_test_precision": 0.8424468636512756, + "adult_test_recall": 0.8438237905502319, + "adult_test_f1-score": 0.8430905938148499, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.8994778990745544, + "adult_test_score": 0.8438237905502319 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.3493582308292389, + "lr": 0.0001147433795267716, + "adult_val_accuracy": 0.8518363833427429, + "adult_val_precision": 0.8468301892280579, + "adult_val_recall": 0.8518363833427429, + "adult_val_f1-score": 0.8484076261520386, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9025565981864929, + "adult_val_score": 0.8518363833427429 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.38207849860191345, + "lr": 0.00011474348866613582, + "adult_test_accuracy": 0.8450559377670288, + "adult_test_precision": 0.839259684085846, + "adult_test_recall": 0.8450559377670288, + "adult_test_f1-score": 0.8408308029174805, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.8933817744255066, + "adult_test_score": 0.8450559377670288 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.3646900296211243, + "std": 0.021436320110340298, + "n": 5, + "ci95": 0.026612439883353347 + }, + "lr": { + "mean": 0.00016240329714491963, + "std": 8.904977238316988e-05, + "n": 5, + "ci95": 0.0001105521704273424 + }, + "adult_test_accuracy": { + "mean": 0.846904194355011, + "std": 0.002503585590198047, + "n": 5, + "ci95": 0.0031081137372937634 + }, + "adult_test_precision": { + "mean": 0.8427235126495362, + "std": 0.0024225481930568674, + "n": 5, + "ci95": 0.0030075086498243526 + }, + "adult_test_recall": { + "mean": 0.846904194355011, + "std": 0.002503585590198047, + "n": 5, + "ci95": 0.0031081137372937634 + }, + "adult_test_f1-score": { + "mean": 0.8440764307975769, + "std": 0.0024357055772409658, + "n": 5, + "ci95": 0.003023843081005598 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.8962172389030456, + "std": 0.0026775489241307932, + "n": 5, + "ci95": 0.003324083117409511 + }, + "adult_test_score": { + "mean": 0.846904194355011, + "std": 0.002503585590198047, + "n": 5, + "ci95": 0.0031081137372937634 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/downstream_results/fttransformer_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/downstream_results/fttransformer_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..8d5fefb9e77c7fc066f62a8bfcdfd8f617cef33a --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/downstream_results/fttransformer_adult_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "adult", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpcmgojpq8", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpcmgojpq8/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.30743104219436646, + "lr": 0.0, + "adult_val_accuracy": 0.8564273715019226, + "adult_val_precision": 0.8504680395126343, + "adult_val_recall": 0.8564273715019226, + "adult_val_f1-score": 0.8514736890792847, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9129331707954407, + "adult_val_score": 0.8564273715019226 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.31639957427978516, + "lr": 0.0, + "adult_test_accuracy": 0.8571721911430359, + "adult_test_precision": 0.8515150547027588, + "adult_test_recall": 0.8571721911430359, + "adult_test_f1-score": 0.8521527051925659, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9104266166687012, + "adult_test_score": 0.8571721911430359 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.30582520365715027, + "lr": 0.0, + "adult_val_accuracy": 0.8572620749473572, + "adult_val_precision": 0.8513126969337463, + "adult_val_recall": 0.8572620749473572, + "adult_val_f1-score": 0.8522243499755859, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9142816662788391, + "adult_val_score": 0.8572620749473572 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3195192515850067, + "lr": 0.0, + "adult_test_accuracy": 0.853065013885498, + "adult_test_precision": 0.8471779823303223, + "adult_test_recall": 0.853065013885498, + "adult_test_f1-score": 0.8481214642524719, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9084267020225525, + "adult_test_score": 0.853065013885498 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.30898499488830566, + "lr": 3.942650891985977e-06, + "adult_val_accuracy": 0.859557569026947, + "adult_val_precision": 0.8539456129074097, + "adult_val_recall": 0.859557569026947, + "adult_val_f1-score": 0.8549048900604248, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9134978652000427, + "adult_val_score": 0.859557569026947 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3249467611312866, + "lr": 3.9426490729965735e-06, + "adult_test_accuracy": 0.8548105359077454, + "adult_test_precision": 0.8491621613502502, + "adult_test_recall": 0.8548105359077454, + "adult_test_f1-score": 0.8501276969909668, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9066765904426575, + "adult_test_score": 0.8548105359077454 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.3104228377342224, + "lr": 0.0, + "adult_val_accuracy": 0.8612270355224609, + "adult_val_precision": 0.8557984232902527, + "adult_val_recall": 0.8612270355224609, + "adult_val_f1-score": 0.8567377328872681, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9124439358711243, + "adult_val_score": 0.8612270355224609 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3231554925441742, + "lr": 0.0, + "adult_test_accuracy": 0.8537837266921997, + "adult_test_precision": 0.8482478260993958, + "adult_test_recall": 0.8537837266921997, + "adult_test_f1-score": 0.8493890762329102, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9070814251899719, + "adult_test_score": 0.8537837266921997 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.306887149810791, + "lr": 0.0, + "adult_val_accuracy": 0.8622704744338989, + "adult_val_precision": 0.8569055199623108, + "adult_val_recall": 0.8622704744338989, + "adult_val_f1-score": 0.8577881455421448, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9134140014648438, + "adult_val_score": 0.8622704744338989 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3199262320995331, + "lr": 0.0, + "adult_test_accuracy": 0.8541944622993469, + "adult_test_precision": 0.8483118414878845, + "adult_test_recall": 0.8541944622993469, + "adult_test_f1-score": 0.8491114974021912, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9083704948425293, + "adult_test_score": 0.8541944622993469 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.32078946232795713, + "std": 0.0033358264028000547, + "n": 5, + "ci95": 0.004141311528698292 + }, + "lr": { + "mean": 7.885298145993147e-07, + "std": 1.7632062677293736e-06, + "n": 5, + "ci95": 2.188958765327697e-06 + }, + "adult_test_accuracy": { + "mean": 0.8546051859855652, + "std": 0.0015690206725791603, + "n": 5, + "ci95": 0.0019478841568805385 + }, + "adult_test_precision": { + "mean": 0.8488829731941223, + "std": 0.0016311342831048195, + "n": 5, + "ci95": 0.002024996026713774 + }, + "adult_test_recall": { + "mean": 0.8546051859855652, + "std": 0.0015690206725791603, + "n": 5, + "ci95": 0.0019478841568805385 + }, + "adult_test_f1-score": { + "mean": 0.8497804880142212, + "std": 0.0015083998042916729, + "n": 5, + "ci95": 0.0018726254742021042 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9081963658332824, + "std": 0.001467157484465253, + "n": 5, + "ci95": 0.0018214245800476452 + }, + "adult_test_score": { + "mean": 0.8546051859855652, + "std": 0.0015690206725791603, + "n": 5, + "ci95": 0.0019478841568805385 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/downstream_results/mlp_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/downstream_results/mlp_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..f9dccee9c9e96a6c62ae20311713d182b52244c2 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/downstream_results/mlp_adult_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "adult", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpyyd34cfg", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpyyd34cfg/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.3559983968734741, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8572620749473572, + "adult_val_precision": 0.8515556454658508, + "adult_val_recall": 0.8572620749473572, + "adult_val_f1-score": 0.8526722192764282, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9106177091598511, + "adult_val_score": 0.8572620749473572 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.33621934056282043, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8502926230430603, + "adult_test_precision": 0.8443757891654968, + "adult_test_recall": 0.8502926230430603, + "adult_test_f1-score": 0.8455743193626404, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9052691459655762, + "adult_test_score": 0.8502926230430603 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.3149104714393616, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8568447232246399, + "adult_val_precision": 0.8503187298774719, + "adult_val_recall": 0.8568447232246399, + "adult_val_f1-score": 0.8495827317237854, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9128347635269165, + "adult_val_score": 0.8568447232246399 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32636964321136475, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.848444402217865, + "adult_test_precision": 0.841575026512146, + "adult_test_recall": 0.848444402217865, + "adult_test_f1-score": 0.8418397307395935, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9041788578033447, + "adult_test_score": 0.848444402217865 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.3182305097579956, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8532971739768982, + "adult_val_precision": 0.8517521023750305, + "adult_val_recall": 0.8532971739768982, + "adult_val_f1-score": 0.8524627685546875, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9092016816139221, + "adult_val_score": 0.8532971739768982 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3303941488265991, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8463907837867737, + "adult_test_precision": 0.8443595767021179, + "adult_test_recall": 0.8463907837867737, + "adult_test_f1-score": 0.8452679514884949, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9036491513252258, + "adult_test_score": 0.8463907837867737 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.3132666051387787, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8537145256996155, + "adult_val_precision": 0.8468869924545288, + "adult_val_recall": 0.8537145256996155, + "adult_val_f1-score": 0.8469576835632324, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9107015132904053, + "adult_val_score": 0.8537145256996155 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3282977044582367, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8470068573951721, + "adult_test_precision": 0.8398596048355103, + "adult_test_recall": 0.8470068573951721, + "adult_test_f1-score": 0.8397834897041321, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.904646098613739, + "adult_test_score": 0.8470068573951721 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.32521042227745056, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8457846641540527, + "adult_val_precision": 0.8475684523582458, + "adult_val_recall": 0.8457846641540527, + "adult_val_f1-score": 0.8466175198554993, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.907246470451355, + "adult_val_score": 0.8457846641540527 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.33766505122184753, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8430023789405823, + "adult_test_precision": 0.8462325930595398, + "adult_test_recall": 0.8430023789405823, + "adult_test_f1-score": 0.8444317579269409, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9008655548095703, + "adult_test_score": 0.8430023789405823 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.3317891776561737, + "std": 0.0049411556761789184, + "n": 5, + "ci95": 0.006134271540532343 + }, + "lr": { + "mean": 0.0009999996982514858, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_accuracy": { + "mean": 0.8470274090766907, + "std": 0.002705368437767131, + "n": 5, + "ci95": 0.0033586200682676967 + }, + "adult_test_precision": { + "mean": 0.8432805180549622, + "std": 0.0025342692171444496, + "n": 5, + "ci95": 0.0031462063844137547 + }, + "adult_test_recall": { + "mean": 0.8470274090766907, + "std": 0.002705368437767131, + "n": 5, + "ci95": 0.0033586200682676967 + }, + "adult_test_f1-score": { + "mean": 0.8433794498443603, + "std": 0.002489623827736458, + "n": 5, + "ci95": 0.0030907806986816243 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9037217617034912, + "std": 0.0017044737978778874, + "n": 5, + "ci95": 0.0021160444631024016 + }, + "adult_test_score": { + "mean": 0.8470274090766907, + "std": 0.002705368437767131, + "n": 5, + "ci95": 0.0033586200682676967 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/downstream_results/resnet_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/downstream_results/resnet_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..bcef498dc34931f560cad8727b6f339a915f7feb --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/downstream_results/resnet_adult_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "adult", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpvh3jmcvm", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpvh3jmcvm/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.309451162815094, + "lr": 0.00020610730280168355, + "adult_val_accuracy": 0.8558013439178467, + "adult_val_precision": 0.8491974472999573, + "adult_val_recall": 0.8558013439178467, + "adult_val_f1-score": 0.8490793704986572, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9110930562019348, + "adult_val_score": 0.8558013439178467 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32797789573669434, + "lr": 0.00020610728824976832, + "adult_test_accuracy": 0.851422131061554, + "adult_test_precision": 0.84477299451828, + "adult_test_recall": 0.851422131061554, + "adult_test_f1-score": 0.8442701101303101, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9043971300125122, + "adult_test_score": 0.851422131061554 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.31741419434547424, + "lr": 0.0, + "adult_val_accuracy": 0.8503756523132324, + "adult_val_precision": 0.8433756828308105, + "adult_val_recall": 0.8503756523132324, + "adult_val_f1-score": 0.83997642993927, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9075667858123779, + "adult_val_score": 0.8503756523132324 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.33626243472099304, + "lr": 0.0, + "adult_test_accuracy": 0.8489577770233154, + "adult_test_precision": 0.8428980112075806, + "adult_test_recall": 0.8489577770233154, + "adult_test_f1-score": 0.8379392623901367, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9014104604721069, + "adult_test_score": 0.8489577770233154 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.3109719753265381, + "lr": 2.4471735741826706e-05, + "adult_val_accuracy": 0.85454922914505, + "adult_val_precision": 0.8477838039398193, + "adult_val_recall": 0.85454922914505, + "adult_val_f1-score": 0.8476442694664001, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9100755453109741, + "adult_val_score": 0.85454922914505 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3292672038078308, + "lr": 2.4471757569699548e-05, + "adult_test_accuracy": 0.8502926230430603, + "adult_test_precision": 0.8435102701187134, + "adult_test_recall": 0.8502926230430603, + "adult_test_f1-score": 0.8421924710273743, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9029139876365662, + "adult_test_score": 0.8502926230430603 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.31218716502189636, + "lr": 0.0, + "adult_val_accuracy": 0.855592668056488, + "adult_val_precision": 0.849566638469696, + "adult_val_recall": 0.855592668056488, + "adult_val_f1-score": 0.8506101965904236, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9097763895988464, + "adult_val_score": 0.855592668056488 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32591357827186584, + "lr": 0.0, + "adult_test_accuracy": 0.8509086966514587, + "adult_test_precision": 0.8444417715072632, + "adult_test_recall": 0.8509086966514587, + "adult_test_f1-score": 0.8449406623840332, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9043573141098022, + "adult_test_score": 0.8509086966514587 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.3094687759876251, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8580968379974365, + "adult_val_precision": 0.8517345786094666, + "adult_val_recall": 0.8580968379974365, + "adult_val_f1-score": 0.8512685894966125, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.910980224609375, + "adult_val_score": 0.8580968379974365 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3250436782836914, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8511140942573547, + "adult_test_precision": 0.8444275259971619, + "adult_test_recall": 0.8511140942573547, + "adult_test_f1-score": 0.8439319729804993, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9045629501342773, + "adult_test_score": 0.8511140942573547 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.3288929581642151, + "std": 0.004443541490479884, + "n": 5, + "ci95": 0.005516500974789044 + }, + "lr": { + "mean": 0.00024611574881419076, + "std": 0.00043017835505155204, + "n": 5, + "ci95": 0.0005340513462199611 + }, + "adult_test_accuracy": { + "mean": 0.8505390644073486, + "std": 0.0009757429447590387, + "n": 5, + "ci95": 0.0012113506574517124 + }, + "adult_test_precision": { + "mean": 0.8440101146697998, + "std": 0.000779288808896782, + "n": 5, + "ci95": 0.0009674597352430758 + }, + "adult_test_recall": { + "mean": 0.8505390644073486, + "std": 0.0009757429447590387, + "n": 5, + "ci95": 0.0012113506574517124 + }, + "adult_test_f1-score": { + "mean": 0.8426548957824707, + "std": 0.0028246950039354027, + "n": 5, + "ci95": 0.0035067598167083966 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.903528368473053, + "std": 0.0013578710181368938, + "n": 5, + "ci95": 0.0016857492635634203 + }, + "adult_test_score": { + "mean": 0.8505390644073486, + "std": 0.0009757429447590387, + "n": 5, + "ci95": 0.0012113506574517124 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..aaaf779351df12196b5ec98aa7990883ae64b4ec --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:171035382afa9de08fc1171f4691114fb12f5f906c2c5a69084f3dbcf60d2e75 +size 175049856 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..777aa2addadf2fd2e580fe18af9f44851dfbf4dd --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d508eca3089f268e2528a1b563aa3984135a80a5cedd8dab78204641eaaf3502 +size 15762582 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..1619d109a8a3dde54f12d98af95bcad43806969a --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.00025_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_211243/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: adult +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.001 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.3 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.1 +mask_min_trgt_share: 0.1 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 256 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.1 +recon_y_weight: 1.0 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.00025 +sw_s_y_weight: 0.0001 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 15 +vtjepa_kl_x_end_weight: 0.0001 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 15 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 15 +wjepa_joint_end_weight: 0.09 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..068c4eaeb7c159171791ecc72d10c3497a36fbd3 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-15T21:17:34.951060+00:00", + "spec": { + "dataset": "adult", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.0005, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0001, + "lambda_joint": 0.09, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..5f8b27c0cfee49ebe3be9d2610657d47c11edf0f --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "adult", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.0005, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0001, + "lambda_joint": 0.09, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "adult", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.09", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.0005", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "0.0001", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..1144db603fdb214f1dc0ff87da6b2851243fe4de --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "adult", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.09, + "s_x": 0.0005, + "z": 0.0, + "s_y": 0.0001 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 15 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "adult", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.09", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.0005", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "0.0001", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/autoint_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/autoint_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..dd4bfb25818537d4b908955058ff16d67b746aaf --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/autoint_adult_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "adult", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpc4tcvdhr", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpc4tcvdhr/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.3129625618457794, + "lr": 6.184666563058272e-05, + "adult_val_accuracy": 0.8568447232246399, + "adult_val_precision": 0.8516823649406433, + "adult_val_recall": 0.8568447232246399, + "adult_val_f1-score": 0.8530578017234802, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.911578357219696, + "adult_val_score": 0.8568447232246399 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32497116923332214, + "lr": 6.18466583546251e-05, + "adult_test_accuracy": 0.855734646320343, + "adult_test_precision": 0.8507261872291565, + "adult_test_recall": 0.855734646320343, + "adult_test_f1-score": 0.8519795536994934, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9085766673088074, + "adult_test_score": 0.855734646320343 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.3086968958377838, + "lr": 0.0001812879927456379, + "adult_val_accuracy": 0.8626878261566162, + "adult_val_precision": 0.8568668961524963, + "adult_val_recall": 0.8626878261566162, + "adult_val_f1-score": 0.8564330339431763, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9119505882263184, + "adult_val_score": 0.8626878261566162 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3204444944858551, + "lr": 0.0001812881964724511, + "adult_test_accuracy": 0.8570694923400879, + "adult_test_precision": 0.8510997891426086, + "adult_test_recall": 0.8570694923400879, + "adult_test_f1-score": 0.8507835865020752, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9077580571174622, + "adult_test_score": 0.8570694923400879 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.3128451406955719, + "lr": 1.570841959619429e-05, + "adult_val_accuracy": 0.8558013439178467, + "adult_val_precision": 0.8500397205352783, + "adult_val_recall": 0.8558013439178467, + "adult_val_f1-score": 0.8512482643127441, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9117394089698792, + "adult_val_score": 0.8558013439178467 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.33015069365501404, + "lr": 1.5708421415183693e-05, + "adult_test_accuracy": 0.8523462414741516, + "adult_test_precision": 0.846514880657196, + "adult_test_recall": 0.8523462414741516, + "adult_test_f1-score": 0.8475839495658875, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9068588614463806, + "adult_test_score": 0.8523462414741516 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.3171083331108093, + "lr": 4.758648356073536e-05, + "adult_val_accuracy": 0.8583055138587952, + "adult_val_precision": 0.8522410988807678, + "adult_val_recall": 0.8583055138587952, + "adult_val_f1-score": 0.852821409702301, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9097170233726501, + "adult_val_score": 0.8583055138587952 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3377856910228729, + "lr": 4.758644354296848e-05, + "adult_test_accuracy": 0.8550159335136414, + "adult_test_precision": 0.8490124344825745, + "adult_test_recall": 0.8550159335136414, + "adult_test_f1-score": 0.849465012550354, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9036782383918762, + "adult_test_score": 0.8550159335136414 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.314855694770813, + "lr": 2.4471735741826706e-05, + "adult_val_accuracy": 0.8628965020179749, + "adult_val_precision": 0.8576472997665405, + "adult_val_recall": 0.8628965020179749, + "adult_val_f1-score": 0.858567476272583, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9121567606925964, + "adult_val_score": 0.8628965020179749 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.33142223954200745, + "lr": 2.4471757569699548e-05, + "adult_test_accuracy": 0.8518328666687012, + "adult_test_precision": 0.845725417137146, + "adult_test_recall": 0.8518328666687012, + "adult_test_f1-score": 0.8465977907180786, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9062855839729309, + "adult_test_score": 0.8518328666687012 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.32895485758781434, + "std": 0.006591617210483101, + "n": 5, + "ci95": 0.008183261672018112 + }, + "lr": { + "mean": 6.618029547098559e-05, + "std": 6.689796861211155e-05, + "n": 5, + "ci95": 8.305148266327208e-05 + }, + "adult_test_accuracy": { + "mean": 0.854399836063385, + "std": 0.0022413690981297548, + "n": 5, + "ci95": 0.002782581155410685 + }, + "adult_test_precision": { + "mean": 0.8486157417297363, + "std": 0.0024263877674550677, + "n": 5, + "ci95": 0.0030122753468284935 + }, + "adult_test_recall": { + "mean": 0.854399836063385, + "std": 0.0022413690981297548, + "n": 5, + "ci95": 0.002782581155410685 + }, + "adult_test_f1-score": { + "mean": 0.8492819786071777, + "std": 0.002216608125160295, + "n": 5, + "ci95": 0.0027518412755613807 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9066314816474914, + "std": 0.0018672112927822933, + "n": 5, + "ci95": 0.002318077357629944 + }, + "adult_test_score": { + "mean": 0.854399836063385, + "std": 0.0022413690981297548, + "n": 5, + "ci95": 0.002782581155410685 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/dcnv2_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/dcnv2_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..5871f08de6a2f4bc14bf4c923c8ced29b10e8b6b --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/dcnv2_adult_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "adult", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpecp9jcm2", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpecp9jcm2/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.3154257535934448, + "lr": 0.000624345033429563, + "adult_val_accuracy": 0.8576794862747192, + "adult_val_precision": 0.8512991666793823, + "adult_val_recall": 0.8576794862747192, + "adult_val_f1-score": 0.8511360883712769, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9099377989768982, + "adult_val_score": 0.8576794862747192 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3306029140949249, + "lr": 0.0006243446841835976, + "adult_test_accuracy": 0.8479310274124146, + "adult_test_precision": 0.8408699631690979, + "adult_test_recall": 0.8479310274124146, + "adult_test_f1-score": 0.8406109809875488, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9038339853286743, + "adult_test_score": 0.8479310274124146 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.3332693874835968, + "lr": 0.00020610730280168355, + "adult_val_accuracy": 0.855592668056488, + "adult_val_precision": 0.8495977520942688, + "adult_val_recall": 0.855592668056488, + "adult_val_f1-score": 0.8506670594215393, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9069664478302002, + "adult_val_score": 0.855592668056488 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.36531636118888855, + "lr": 0.00020610728824976832, + "adult_test_accuracy": 0.8478283286094666, + "adult_test_precision": 0.8418262600898743, + "adult_test_recall": 0.8478283286094666, + "adult_test_f1-score": 0.8431715369224548, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.8969647288322449, + "adult_test_score": 0.8478283286094666 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.33715713024139404, + "lr": 0.0002320865314686671, + "adult_val_accuracy": 0.8572620749473572, + "adult_val_precision": 0.8519992232322693, + "adult_val_recall": 0.8572620749473572, + "adult_val_f1-score": 0.8533260822296143, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9056263566017151, + "adult_val_score": 0.8572620749473572 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.36276647448539734, + "lr": 0.00023208664788398892, + "adult_test_accuracy": 0.8520382046699524, + "adult_test_precision": 0.8464993238449097, + "adult_test_recall": 0.8520382046699524, + "adult_test_f1-score": 0.8477650284767151, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.900492250919342, + "adult_test_score": 0.8520382046699524 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.3125121295452118, + "lr": 0.0005626665661111474, + "adult_val_accuracy": 0.8572620749473572, + "adult_val_precision": 0.8531542420387268, + "adult_val_recall": 0.8572620749473572, + "adult_val_f1-score": 0.8545712828636169, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.911125123500824, + "adult_val_score": 0.8572620749473572 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32436859607696533, + "lr": 0.0005626667407341301, + "adult_test_accuracy": 0.8554266095161438, + "adult_test_precision": 0.8511304259300232, + "adult_test_recall": 0.8554266095161438, + "adult_test_f1-score": 0.8524986505508423, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9062830209732056, + "adult_test_score": 0.8554266095161438 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.3415168225765228, + "lr": 0.00013551567099057138, + "adult_val_accuracy": 0.8580968379974365, + "adult_val_precision": 0.8534587025642395, + "adult_val_recall": 0.8580968379974365, + "adult_val_f1-score": 0.8548644185066223, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9058709740638733, + "adult_val_score": 0.8580968379974365 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3774588108062744, + "lr": 0.00013551574375014752, + "adult_test_accuracy": 0.8477256298065186, + "adult_test_precision": 0.8428343534469604, + "adult_test_recall": 0.8477256298065186, + "adult_test_f1-score": 0.8443962335586548, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.8972837924957275, + "adult_test_score": 0.8477256298065186 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.35210263133049013, + "std": 0.023252261168886304, + "n": 5, + "ci95": 0.028866867042656553 + }, + "lr": { + "mean": 0.0003521442209603265, + "std": 0.00022421046319126377, + "n": 5, + "ci95": 0.00027834942948151345 + }, + "adult_test_accuracy": { + "mean": 0.8501899600028991, + "std": 0.003449329973974991, + "n": 5, + "ci95": 0.004282222233002518 + }, + "adult_test_precision": { + "mean": 0.8446320652961731, + "std": 0.0042122757678729715, + "n": 5, + "ci95": 0.005229392688092581 + }, + "adult_test_recall": { + "mean": 0.8501899600028991, + "std": 0.003449329973974991, + "n": 5, + "ci95": 0.004282222233002518 + }, + "adult_test_f1-score": { + "mean": 0.8456884860992432, + "std": 0.004595551546053417, + "n": 5, + "ci95": 0.005705216129479446 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9009715557098389, + "std": 0.004070907724307108, + "n": 5, + "ci95": 0.005053889218212549 + }, + "adult_test_score": { + "mean": 0.8501899600028991, + "std": 0.003449329973974991, + "n": 5, + "ci95": 0.004282222233002518 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/fttransformer_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/fttransformer_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..1e501221afb9611f2b8c630c4af776085ef368cf --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/fttransformer_adult_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "adult", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpgkc7dfpx", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpgkc7dfpx/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.3083151876926422, + "lr": 0.0, + "adult_val_accuracy": 0.8572620749473572, + "adult_val_precision": 0.8512853384017944, + "adult_val_recall": 0.8572620749473572, + "adult_val_f1-score": 0.8521676659584045, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.913361132144928, + "adult_val_score": 0.8572620749473572 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3239251375198364, + "lr": 0.0, + "adult_test_accuracy": 0.8548105359077454, + "adult_test_precision": 0.8489660024642944, + "adult_test_recall": 0.8548105359077454, + "adult_test_f1-score": 0.8497217297554016, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9074366688728333, + "adult_test_score": 0.8548105359077454 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.3046202063560486, + "lr": 0.0, + "adult_val_accuracy": 0.8651919960975647, + "adult_val_precision": 0.8604550361633301, + "adult_val_recall": 0.8651919960975647, + "adult_val_f1-score": 0.8615251779556274, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9156761169433594, + "adult_val_score": 0.8651919960975647 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3174950182437897, + "lr": 0.0, + "adult_test_accuracy": 0.8570694923400879, + "adult_test_precision": 0.8515889048576355, + "adult_test_recall": 0.8570694923400879, + "adult_test_f1-score": 0.8524595499038696, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9097620248794556, + "adult_test_score": 0.8570694923400879 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.30751922726631165, + "lr": 0.0001812879927456379, + "adult_val_accuracy": 0.8618530631065369, + "adult_val_precision": 0.8560426235198975, + "adult_val_recall": 0.8618530631065369, + "adult_val_f1-score": 0.8542932271957397, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9131970405578613, + "adult_val_score": 0.8618530631065369 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32495948672294617, + "lr": 0.0001812881964724511, + "adult_test_accuracy": 0.852859616279602, + "adult_test_precision": 0.8465269207954407, + "adult_test_recall": 0.852859616279602, + "adult_test_f1-score": 0.8443744778633118, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9071648120880127, + "adult_test_score": 0.852859616279602 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.3071005642414093, + "lr": 1.570841959619429e-05, + "adult_val_accuracy": 0.8603923320770264, + "adult_val_precision": 0.854596734046936, + "adult_val_recall": 0.8603923320770264, + "adult_val_f1-score": 0.8552144169807434, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9135773777961731, + "adult_val_score": 0.8603923320770264 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3209441900253296, + "lr": 1.5708421415183693e-05, + "adult_test_accuracy": 0.8556320071220398, + "adult_test_precision": 0.8497291803359985, + "adult_test_recall": 0.8556320071220398, + "adult_test_f1-score": 0.8502432107925415, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9095597267150879, + "adult_test_score": 0.8556320071220398 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.3103259801864624, + "lr": 9.866357686405536e-07, + "adult_val_accuracy": 0.8589315414428711, + "adult_val_precision": 0.8535484671592712, + "adult_val_recall": 0.8589315414428711, + "adult_val_f1-score": 0.8547211289405823, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9123598337173462, + "adult_val_score": 0.8589315414428711 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32588112354278564, + "lr": 9.866350865195272e-07, + "adult_test_accuracy": 0.856042742729187, + "adult_test_precision": 0.8503450155258179, + "adult_test_recall": 0.856042742729187, + "adult_test_f1-score": 0.8511052131652832, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9073761701583862, + "adult_test_score": 0.856042742729187 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.3226409912109375, + "std": 0.0034235611120530395, + "n": 5, + "ci95": 0.004250231094354165 + }, + "lr": { + "mean": 3.959665059483086e-05, + "std": 7.948846526499833e-05, + "n": 5, + "ci95": 9.868214284896717e-05 + }, + "adult_test_accuracy": { + "mean": 0.8552828788757324, + "std": 0.0015800775905302957, + "n": 5, + "ci95": 0.0019616109328735795 + }, + "adult_test_precision": { + "mean": 0.8494312047958374, + "std": 0.0018861858202254824, + "n": 5, + "ci95": 0.0023416335682247525 + }, + "adult_test_recall": { + "mean": 0.8552828788757324, + "std": 0.0015800775905302957, + "n": 5, + "ci95": 0.0019616109328735795 + }, + "adult_test_f1-score": { + "mean": 0.8495808362960815, + "std": 0.0030893152941199632, + "n": 5, + "ci95": 0.0038352766296783224 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9082598805427551, + "std": 0.001284896308804557, + "n": 5, + "ci95": 0.0015951537203397858 + }, + "adult_test_score": { + "mean": 0.8552828788757324, + "std": 0.0015800775905302957, + "n": 5, + "ci95": 0.0019616109328735795 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/mlp_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/mlp_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..0187526f6f8ef0aa43f690d39f6d577177dffcc9 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/mlp_adult_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "adult", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpkjj0wczg", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpkjj0wczg/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.31649187207221985, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8583055138587952, + "adult_val_precision": 0.8544908165931702, + "adult_val_recall": 0.8583055138587952, + "adult_val_f1-score": 0.8558571934700012, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9098939895629883, + "adult_val_score": 0.8583055138587952 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32713568210601807, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8506006598472595, + "adult_test_precision": 0.847040593624115, + "adult_test_recall": 0.8506006598472595, + "adult_test_f1-score": 0.8484002351760864, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9046164155006409, + "adult_test_score": 0.8506006598472595 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.3074006736278534, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8589315414428711, + "adult_val_precision": 0.8526650071144104, + "adult_val_recall": 0.8589315414428711, + "adult_val_f1-score": 0.8520825505256653, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9134336709976196, + "adult_val_score": 0.8589315414428711 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.322750985622406, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8499845862388611, + "adult_test_precision": 0.8432666659355164, + "adult_test_recall": 0.8499845862388611, + "adult_test_f1-score": 0.843402087688446, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9061656594276428, + "adult_test_score": 0.8499845862388611 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.3134402334690094, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8585141897201538, + "adult_val_precision": 0.8525142669677734, + "adult_val_recall": 0.8585141897201538, + "adult_val_f1-score": 0.849731981754303, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9106864333152771, + "adult_val_score": 0.8585141897201538 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32736653089523315, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8481363654136658, + "adult_test_precision": 0.8411210179328918, + "adult_test_recall": 0.8481363654136658, + "adult_test_f1-score": 0.8393124938011169, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.904833972454071, + "adult_test_score": 0.8481363654136658 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.31735482811927795, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8470367193222046, + "adult_val_precision": 0.8513215184211731, + "adult_val_recall": 0.8470367193222046, + "adult_val_f1-score": 0.848872184753418, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9130023717880249, + "adult_val_score": 0.8470367193222046 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3306499421596527, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8475202918052673, + "adult_test_precision": 0.8509243130683899, + "adult_test_recall": 0.8475202918052673, + "adult_test_f1-score": 0.8490062952041626, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.905820906162262, + "adult_test_score": 0.8475202918052673 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.3108050227165222, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8589315414428711, + "adult_val_precision": 0.8527976870536804, + "adult_val_recall": 0.8589315414428711, + "adult_val_f1-score": 0.8508283495903015, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9120212197303772, + "adult_val_score": 0.8589315414428711 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3234279751777649, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8518328666687012, + "adult_test_precision": 0.8452309370040894, + "adult_test_recall": 0.8518328666687012, + "adult_test_f1-score": 0.8441458344459534, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9069936871528625, + "adult_test_score": 0.8518328666687012 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.326266223192215, + "std": 0.0032247548276307944, + "n": 5, + "ci95": 0.004003420062172026 + }, + "lr": { + "mean": 0.0009999996982514858, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_accuracy": { + "mean": 0.849614953994751, + "std": 0.001774909471845842, + "n": 5, + "ci95": 0.002203487882936922 + }, + "adult_test_precision": { + "mean": 0.8455167055130005, + "std": 0.003742688388416551, + "n": 5, + "ci95": 0.004646416419710711 + }, + "adult_test_recall": { + "mean": 0.849614953994751, + "std": 0.001774909471845842, + "n": 5, + "ci95": 0.002203487882936922 + }, + "adult_test_f1-score": { + "mean": 0.8448533892631531, + "std": 0.00397288233799676, + "n": 5, + "ci95": 0.004932194137769697 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9056861281394959, + "std": 0.000978299594589643, + "n": 5, + "ci95": 0.001214524648583097 + }, + "adult_test_score": { + "mean": 0.849614953994751, + "std": 0.001774909471845842, + "n": 5, + "ci95": 0.002203487882936922 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/resnet_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/resnet_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..868d89fbb7d0f2c296bae99de02048694c4a2909 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/resnet_adult_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "adult", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpmfofm3hm", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpmfofm3hm/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.3041922152042389, + "lr": 0.00020610730280168355, + "adult_val_accuracy": 0.8583055138587952, + "adult_val_precision": 0.8523894548416138, + "adult_val_recall": 0.8583055138587952, + "adult_val_f1-score": 0.849209189414978, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9153968691825867, + "adult_val_score": 0.8583055138587952 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3245809078216553, + "lr": 0.00020610728824976832, + "adult_test_accuracy": 0.8508060574531555, + "adult_test_precision": 0.8444136381149292, + "adult_test_recall": 0.8508060574531555, + "adult_test_f1-score": 0.8413415551185608, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9065249562263489, + "adult_test_score": 0.8508060574531555 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.3081151247024536, + "lr": 9.54915230977349e-05, + "adult_val_accuracy": 0.8522537350654602, + "adult_val_precision": 0.8471648693084717, + "adult_val_recall": 0.8522537350654602, + "adult_val_f1-score": 0.8487269282341003, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9119558930397034, + "adult_val_score": 0.8522537350654602 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3286660611629486, + "lr": 9.54914212343283e-05, + "adult_test_accuracy": 0.8495738506317139, + "adult_test_precision": 0.8442495465278625, + "adult_test_recall": 0.8495738506317139, + "adult_test_f1-score": 0.8457114100456238, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9046452641487122, + "adult_test_score": 0.8495738506317139 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.31368014216423035, + "lr": 9.54915230977349e-05, + "adult_val_accuracy": 0.8564273715019226, + "adult_val_precision": 0.8509513735771179, + "adult_val_recall": 0.8564273715019226, + "adult_val_f1-score": 0.8522515296936035, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9105032086372375, + "adult_val_score": 0.8564273715019226 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3249024748802185, + "lr": 9.54914212343283e-05, + "adult_test_accuracy": 0.8540918231010437, + "adult_test_precision": 0.848552405834198, + "adult_test_recall": 0.8540918231010437, + "adult_test_f1-score": 0.8496665954589844, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9066640138626099, + "adult_test_score": 0.8540918231010437 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.31039074063301086, + "lr": 0.0, + "adult_val_accuracy": 0.8572620749473572, + "adult_val_precision": 0.8511360883712769, + "adult_val_recall": 0.8572620749473572, + "adult_val_f1-score": 0.8518243432044983, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9105858206748962, + "adult_val_score": 0.8572620749473572 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3238856792449951, + "lr": 0.0, + "adult_test_accuracy": 0.8510113954544067, + "adult_test_precision": 0.8444961905479431, + "adult_test_recall": 0.8510113954544067, + "adult_test_f1-score": 0.844857931137085, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9052449464797974, + "adult_test_score": 0.8510113954544067 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.3099907338619232, + "lr": 0.0007938927737995982, + "adult_val_accuracy": 0.8597663044929504, + "adult_val_precision": 0.8535987138748169, + "adult_val_recall": 0.8597663044929504, + "adult_val_f1-score": 0.8529577851295471, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9129458665847778, + "adult_val_score": 0.8597663044929504 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3275333046913147, + "lr": 0.0007938928320072591, + "adult_test_accuracy": 0.8553239703178406, + "adult_test_precision": 0.8491895198822021, + "adult_test_recall": 0.8553239703178406, + "adult_test_f1-score": 0.8476958870887756, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9070183038711548, + "adult_test_score": 0.8553239703178406 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.32591368556022643, + "std": 0.002068235403832118, + "n": 5, + "ci95": 0.002567641743815679 + }, + "lr": { + "mean": 0.0002381965925451368, + "std": 0.00031909841058062064, + "n": 5, + "ci95": 0.00039614948949908926 + }, + "adult_test_accuracy": { + "mean": 0.852161419391632, + "std": 0.002428129937314891, + "n": 5, + "ci95": 0.003014438189630919 + }, + "adult_test_precision": { + "mean": 0.846180260181427, + "std": 0.002468168258698613, + "n": 5, + "ci95": 0.00306414436192962 + }, + "adult_test_recall": { + "mean": 0.852161419391632, + "std": 0.002428129937314891, + "n": 5, + "ci95": 0.003014438189630919 + }, + "adult_test_f1-score": { + "mean": 0.8458546757698059, + "std": 0.0031346081805038763, + "n": 5, + "ci95": 0.0038915061602055336 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9060194969177247, + "std": 0.001019430063744018, + "n": 5, + "ci95": 0.001265586684049573 + }, + "adult_test_score": { + "mean": 0.852161419391632, + "std": 0.002428129937314891, + "n": 5, + "ci95": 0.003014438189630919 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..0b1bb38cfccedc65f32fcb4d82514f69f971d92f --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4d337efbf88479eb70a2fc4017fea04d027c3c2921e2ea90d751bf4baf518fe0 +size 175049856 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..fc443fa64fabe849a47b7424c21ecd431f6b2ec0 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:812ad992f570d69d17d62580bace2a6601d4972e64ab0097742f28201dff9d22 +size 15770774 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..e20655c389dc76bd957659fb3621271ffcf9e962 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.09_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: adult +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.001 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.3 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.1 +mask_min_trgt_share: 0.1 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 256 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.1 +recon_y_weight: 1.0 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.0005 +sw_s_y_weight: 0.0001 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 15 +vtjepa_kl_x_end_weight: 0.0001 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 15 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 15 +wjepa_joint_end_weight: 0.09 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..0f3a48659b04f1ddc39d3b5964666c7381239134 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-15T20:53:44.589601+00:00", + "spec": { + "dataset": "adult", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.001, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0002, + "lambda_joint": 0.18, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..87d2741b4911c6505f5e143c2c08e2392def8b78 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "adult", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.001, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0002, + "lambda_joint": 0.18, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "adult", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.18", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.001", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "0.0002", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..0c31b446413a8ea5051dd82aae4a42ba02a68f74 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "adult", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.18, + "s_x": 0.001, + "z": 0.0, + "s_y": 0.0002 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 15 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "adult", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.18", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.001", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "0.0002", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/autoint_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/autoint_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..5817409fedb62c45c4cfa864c2ecec69a4eb8cc4 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/autoint_adult_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "adult", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpz_vh56h5", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpz_vh56h5/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.3097549080848694, + "lr": 3.5111756005790085e-05, + "adult_val_accuracy": 0.8564273715019226, + "adult_val_precision": 0.8511545658111572, + "adult_val_recall": 0.8564273715019226, + "adult_val_f1-score": 0.8525221943855286, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9122442007064819, + "adult_val_score": 0.8564273715019226 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3297220766544342, + "lr": 3.511177419568412e-05, + "adult_test_accuracy": 0.8556320071220398, + "adult_test_precision": 0.8504593372344971, + "adult_test_recall": 0.8556320071220398, + "adult_test_f1-score": 0.8516566157341003, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.905756950378418, + "adult_test_score": 0.8556320071220398 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.31273287534713745, + "lr": 3.5111756005790085e-05, + "adult_val_accuracy": 0.8593488931655884, + "adult_val_precision": 0.8542547225952148, + "adult_val_recall": 0.8593488931655884, + "adult_val_f1-score": 0.8555231690406799, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9118042588233948, + "adult_val_score": 0.8593488931655884 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3315148651599884, + "lr": 3.511177419568412e-05, + "adult_test_accuracy": 0.8521409034729004, + "adult_test_precision": 0.8463577628135681, + "adult_test_recall": 0.8521409034729004, + "adult_test_f1-score": 0.847480833530426, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9067575335502625, + "adult_test_score": 0.8521409034729004 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.32173120975494385, + "lr": 9.54915230977349e-05, + "adult_val_accuracy": 0.8510016798973083, + "adult_val_precision": 0.8508240580558777, + "adult_val_recall": 0.8510016798973083, + "adult_val_f1-score": 0.8509121537208557, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9080141186714172, + "adult_val_score": 0.8510016798973083 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3291274905204773, + "lr": 9.54914212343283e-05, + "adult_test_accuracy": 0.8515247702598572, + "adult_test_precision": 0.850472629070282, + "adult_test_recall": 0.8515247702598572, + "adult_test_f1-score": 0.850969672203064, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9071711897850037, + "adult_test_score": 0.8515247702598572 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.3196982443332672, + "lr": 8.856373824528418e-06, + "adult_val_accuracy": 0.8568447232246399, + "adult_val_precision": 0.8511075377464294, + "adult_val_recall": 0.8568447232246399, + "adult_val_f1-score": 0.8522414565086365, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9082854390144348, + "adult_val_score": 0.8568447232246399 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3319469094276428, + "lr": 8.856371096044313e-06, + "adult_test_accuracy": 0.8549132347106934, + "adult_test_precision": 0.8490829467773438, + "adult_test_recall": 0.8549132347106934, + "adult_test_f1-score": 0.8498416543006897, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9064590930938721, + "adult_test_score": 0.8549132347106934 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.31897929310798645, + "lr": 1.570841959619429e-05, + "adult_val_accuracy": 0.859557569026947, + "adult_val_precision": 0.8540050983428955, + "adult_val_recall": 0.859557569026947, + "adult_val_f1-score": 0.8550142645835876, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9082602858543396, + "adult_val_score": 0.859557569026947 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3310595154762268, + "lr": 1.5708421415183693e-05, + "adult_test_accuracy": 0.8533730506896973, + "adult_test_precision": 0.8473863005638123, + "adult_test_recall": 0.8533730506896973, + "adult_test_f1-score": 0.8481785655021667, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9060685038566589, + "adult_test_score": 0.8533730506896973 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.3306741714477539, + "std": 0.0012014391671855244, + "n": 5, + "ci95": 0.0014915446049346812 + }, + "lr": { + "mean": 3.805595242738491e-05, + "std": 3.416210305389687e-05, + "n": 5, + "ci95": 4.2411053255927504e-05 + }, + "adult_test_accuracy": { + "mean": 0.8535167932510376, + "std": 0.001754000333041597, + "n": 5, + "ci95": 0.0021775299201626935 + }, + "adult_test_precision": { + "mean": 0.8487517952919006, + "std": 0.0018427287540902161, + "n": 5, + "ci95": 0.002287683144174413 + }, + "adult_test_recall": { + "mean": 0.8535167932510376, + "std": 0.001754000333041597, + "n": 5, + "ci95": 0.0021775299201626935 + }, + "adult_test_f1-score": { + "mean": 0.8496254682540894, + "std": 0.0017798993244143265, + "n": 5, + "ci95": 0.0022096826099619925 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.906442654132843, + "std": 0.0005568576010580454, + "n": 5, + "ci95": 0.0006913191889030033 + }, + "adult_test_score": { + "mean": 0.8535167932510376, + "std": 0.001754000333041597, + "n": 5, + "ci95": 0.0021775299201626935 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/dcnv2_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/dcnv2_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..d62a8c83e319aae44e423b7ad8b3535faa150153 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/dcnv2_adult_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "adult", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp3bcyimaa", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp3bcyimaa/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.3229980766773224, + "lr": 0.0003454915713518858, + "adult_val_accuracy": 0.8528798222541809, + "adult_val_precision": 0.8480056524276733, + "adult_val_recall": 0.8528798222541809, + "adult_val_f1-score": 0.8495551943778992, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9086211919784546, + "adult_val_score": 0.8528798222541809 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.35428568720817566, + "lr": 0.0003454913094174117, + "adult_test_accuracy": 0.8497792482376099, + "adult_test_precision": 0.8449870347976685, + "adult_test_recall": 0.8497792482376099, + "adult_test_f1-score": 0.846494734287262, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9001891613006592, + "adult_test_score": 0.8497792482376099 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.33625587821006775, + "lr": 0.0001812879927456379, + "adult_val_accuracy": 0.8503756523132324, + "adult_val_precision": 0.8463929295539856, + "adult_val_recall": 0.8503756523132324, + "adult_val_f1-score": 0.8478939533233643, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9053506851196289, + "adult_val_score": 0.8503756523132324 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3585028350353241, + "lr": 0.0001812881964724511, + "adult_test_accuracy": 0.8473148941993713, + "adult_test_precision": 0.8424565196037292, + "adult_test_recall": 0.8473148941993713, + "adult_test_f1-score": 0.8440285921096802, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.8968597650527954, + "adult_test_score": 0.8473148941993713 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.3385787904262543, + "lr": 0.0002320865314686671, + "adult_val_accuracy": 0.856218695640564, + "adult_val_precision": 0.8500114679336548, + "adult_val_recall": 0.856218695640564, + "adult_val_f1-score": 0.8507702350616455, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.906609833240509, + "adult_val_score": 0.856218695640564 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.35559284687042236, + "lr": 0.00023208664788398892, + "adult_test_accuracy": 0.8516274690628052, + "adult_test_precision": 0.8454770445823669, + "adult_test_recall": 0.8516274690628052, + "adult_test_f1-score": 0.8463290929794312, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9011832475662231, + "adult_test_score": 0.8516274690628052 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.32607176899909973, + "lr": 0.0002591231605038047, + "adult_val_accuracy": 0.8547579050064087, + "adult_val_precision": 0.8518714904785156, + "adult_val_recall": 0.8547579050064087, + "adult_val_f1-score": 0.8530622124671936, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9085204005241394, + "adult_val_score": 0.8547579050064087 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3481575846672058, + "lr": 0.0002591231605038047, + "adult_test_accuracy": 0.8472122550010681, + "adult_test_precision": 0.8431577086448669, + "adult_test_recall": 0.8472122550010681, + "adult_test_f1-score": 0.8446514010429382, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9009537100791931, + "adult_test_score": 0.8472122550010681 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.3388935327529907, + "lr": 7.783600449329242e-05, + "adult_val_accuracy": 0.8518363833427429, + "adult_val_precision": 0.8464736938476562, + "adult_val_recall": 0.8518363833427429, + "adult_val_f1-score": 0.8480270504951477, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9038204550743103, + "adult_val_score": 0.8518363833427429 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.37780502438545227, + "lr": 7.783606997691095e-05, + "adult_test_accuracy": 0.8488551378250122, + "adult_test_precision": 0.8433895707130432, + "adult_test_recall": 0.8488551378250122, + "adult_test_f1-score": 0.8448542356491089, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.8942604660987854, + "adult_test_score": 0.8488551378250122 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.35886879563331603, + "std": 0.011238120950660867, + "n": 5, + "ci95": 0.013951733164175459 + }, + "lr": { + "mean": 0.00021916507685091347, + "std": 9.890712655417639e-05, + "n": 5, + "ci95": 0.00012278973004273052 + }, + "adult_test_accuracy": { + "mean": 0.8489578008651734, + "std": 0.0018410945417859667, + "n": 5, + "ci95": 0.00228565432689236 + }, + "adult_test_precision": { + "mean": 0.843893575668335, + "std": 0.0012809838087613334, + "n": 5, + "ci95": 0.0015902964887040405 + }, + "adult_test_recall": { + "mean": 0.8489578008651734, + "std": 0.0018410945417859667, + "n": 5, + "ci95": 0.00228565432689236 + }, + "adult_test_f1-score": { + "mean": 0.8452716112136841, + "std": 0.0010860774890317307, + "n": 5, + "ci95": 0.001348327125959375 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.8986892700195312, + "std": 0.0030231886822580743, + "n": 5, + "ci95": 0.003753182759377539 + }, + "adult_test_score": { + "mean": 0.8489578008651734, + "std": 0.0018410945417859667, + "n": 5, + "ci95": 0.00228565432689236 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/fttransformer_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/fttransformer_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..34a556964cb2febf1c4ae73f5778aab13fc453b1 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/fttransformer_adult_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "adult", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpdqg0vskr", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpdqg0vskr/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.31012988090515137, + "lr": 0.0, + "adult_val_accuracy": 0.8558013439178467, + "adult_val_precision": 0.8495090007781982, + "adult_val_recall": 0.8558013439178467, + "adult_val_f1-score": 0.8502203226089478, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9119439125061035, + "adult_val_score": 0.8558013439178467 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32648965716362, + "lr": 0.0, + "adult_test_accuracy": 0.8541944622993469, + "adult_test_precision": 0.8482295870780945, + "adult_test_recall": 0.8541944622993469, + "adult_test_f1-score": 0.8489187359809875, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9071423411369324, + "adult_test_score": 0.8541944622993469 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.31407734751701355, + "lr": 0.00028711033519357443, + "adult_val_accuracy": 0.8549666404724121, + "adult_val_precision": 0.8482016921043396, + "adult_val_recall": 0.8549666404724121, + "adult_val_f1-score": 0.8473213911056519, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9095443487167358, + "adult_val_score": 0.8549666404724121 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32463371753692627, + "lr": 0.0002871101605705917, + "adult_test_accuracy": 0.8482390642166138, + "adult_test_precision": 0.8411728143692017, + "adult_test_recall": 0.8482390642166138, + "adult_test_f1-score": 0.8403176069259644, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9052990078926086, + "adult_test_score": 0.8482390642166138 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.31034132838249207, + "lr": 3.942650891985977e-06, + "adult_val_accuracy": 0.8610183596611023, + "adult_val_precision": 0.8556487560272217, + "adult_val_recall": 0.8610183596611023, + "adult_val_f1-score": 0.8566568493843079, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9122453927993774, + "adult_val_score": 0.8610183596611023 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3230978548526764, + "lr": 3.9426490729965735e-06, + "adult_test_accuracy": 0.8524489402770996, + "adult_test_precision": 0.8466174602508545, + "adult_test_recall": 0.8524489402770996, + "adult_test_f1-score": 0.8476762771606445, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9077522158622742, + "adult_test_score": 0.8524489402770996 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.3039909303188324, + "lr": 0.0, + "adult_val_accuracy": 0.8601836562156677, + "adult_val_precision": 0.8547541499137878, + "adult_val_recall": 0.8601836562156677, + "adult_val_f1-score": 0.8557959198951721, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9142337441444397, + "adult_val_score": 0.8601836562156677 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3211081624031067, + "lr": 0.0, + "adult_test_accuracy": 0.8541944622993469, + "adult_test_precision": 0.8484021425247192, + "adult_test_recall": 0.8541944622993469, + "adult_test_f1-score": 0.8493025302886963, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9090328216552734, + "adult_test_score": 0.8541944622993469 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.31253781914711, + "lr": 0.0, + "adult_val_accuracy": 0.8612270355224609, + "adult_val_precision": 0.8556633591651917, + "adult_val_recall": 0.8612270355224609, + "adult_val_f1-score": 0.8564663529396057, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.912572979927063, + "adult_val_score": 0.8612270355224609 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32445335388183594, + "lr": 0.0, + "adult_test_accuracy": 0.8555293083190918, + "adult_test_precision": 0.8496516942977905, + "adult_test_recall": 0.8555293083190918, + "adult_test_f1-score": 0.8502331972122192, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9084340929985046, + "adult_test_score": 0.8555293083190918 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.32395654916763306, + "std": 0.001998291215678178, + "n": 5, + "ci95": 0.0024808084863883095 + }, + "lr": { + "mean": 5.8210561928717655e-05, + "std": 0.00012797015392690045, + "n": 5, + "ci95": 0.00015887045960842622 + }, + "adult_test_accuracy": { + "mean": 0.8529212474822998, + "std": 0.0028368136834598058, + "n": 5, + "ci95": 0.0035218047324704644 + }, + "adult_test_precision": { + "mean": 0.8468147397041321, + "std": 0.0033331738943727257, + "n": 5, + "ci95": 0.004138018532479769 + }, + "adult_test_recall": { + "mean": 0.8529212474822998, + "std": 0.0028368136834598058, + "n": 5, + "ci95": 0.0035218047324704644 + }, + "adult_test_f1-score": { + "mean": 0.8472896695137024, + "std": 0.00400402745889337, + "n": 5, + "ci95": 0.004970859713449404 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9075320959091187, + "std": 0.001436386105834048, + "n": 5, + "ci95": 0.0017832229922874478 + }, + "adult_test_score": { + "mean": 0.8529212474822998, + "std": 0.0028368136834598058, + "n": 5, + "ci95": 0.0035218047324704644 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/mlp_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/mlp_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..51ddfb6c095ebe64ba119606b6e1b48437aea73a --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/mlp_adult_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "adult", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpgugnm2h5", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpgugnm2h5/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.31026405096054077, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8587228655815125, + "adult_val_precision": 0.8524975180625916, + "adult_val_recall": 0.8587228655815125, + "adult_val_f1-score": 0.852496325969696, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9122937917709351, + "adult_val_score": 0.8587228655815125 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3288733959197998, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8508060574531555, + "adult_test_precision": 0.8443951606750488, + "adult_test_recall": 0.8508060574531555, + "adult_test_f1-score": 0.8450220227241516, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9059605598449707, + "adult_test_score": 0.8508060574531555 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.3105933666229248, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8539232015609741, + "adult_val_precision": 0.853406548500061, + "adult_val_recall": 0.8539232015609741, + "adult_val_f1-score": 0.8536584377288818, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9126105904579163, + "adult_val_score": 0.8539232015609741 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.323831707239151, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.851422131061554, + "adult_test_precision": 0.8507850766181946, + "adult_test_recall": 0.851422131061554, + "adult_test_f1-score": 0.8510932922363281, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9068166613578796, + "adult_test_score": 0.851422131061554 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.3097812235355377, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8560100197792053, + "adult_val_precision": 0.8507546186447144, + "adult_val_recall": 0.8560100197792053, + "adult_val_f1-score": 0.8521473407745361, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9116249084472656, + "adult_val_score": 0.8560100197792053 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3222024738788605, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8498819470405579, + "adult_test_precision": 0.844456136226654, + "adult_test_recall": 0.8498819470405579, + "adult_test_f1-score": 0.8458816409111023, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9060293436050415, + "adult_test_score": 0.8498819470405579 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.315760999917984, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8572620749473572, + "adult_val_precision": 0.8508444428443909, + "adult_val_recall": 0.8572620749473572, + "adult_val_f1-score": 0.8507601618766785, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9086964130401611, + "adult_val_score": 0.8572620749473572 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.34755244851112366, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8492658138275146, + "adult_test_precision": 0.8424614667892456, + "adult_test_recall": 0.8492658138275146, + "adult_test_f1-score": 0.8426065444946289, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9033664464950562, + "adult_test_score": 0.8492658138275146 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.3196761906147003, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8574708104133606, + "adult_val_precision": 0.8518118858337402, + "adult_val_recall": 0.8574708104133606, + "adult_val_f1-score": 0.8474948406219482, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9098320007324219, + "adult_val_score": 0.8574708104133606 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.33166205883026123, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8469042181968689, + "adult_test_precision": 0.8401710987091064, + "adult_test_recall": 0.8469042181968689, + "adult_test_f1-score": 0.8363136053085327, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9049921035766602, + "adult_test_score": 0.8469042181968689 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.33082441687583924, + "std": 0.010094194597787048, + "n": 5, + "ci95": 0.012531588701873209 + }, + "lr": { + "mean": 0.0009999996982514858, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_accuracy": { + "mean": 0.8496560335159302, + "std": 0.0017476735588280467, + "n": 5, + "ci95": 0.0021696754517862653 + }, + "adult_test_precision": { + "mean": 0.8444537878036499, + "std": 0.003949665479931041, + "n": 5, + "ci95": 0.004903371222438429 + }, + "adult_test_recall": { + "mean": 0.8496560335159302, + "std": 0.0017476735588280467, + "n": 5, + "ci95": 0.0021696754517862653 + }, + "adult_test_f1-score": { + "mean": 0.8441834211349487, + "std": 0.005379453103051115, + "n": 5, + "ci95": 0.0066784024297719706 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9054330229759217, + "std": 0.0013241503767377441, + "n": 5, + "ci95": 0.0016438862694747047 + }, + "adult_test_score": { + "mean": 0.8496560335159302, + "std": 0.0017476735588280467, + "n": 5, + "ci95": 0.0021696754517862653 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/resnet_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/resnet_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..b4819f3e47a86b2c630dd5321cda1f200ea331e2 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/downstream_results/resnet_adult_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "adult", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpl4f3lbiu", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpl4f3lbiu/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.31318119168281555, + "lr": 0.0, + "adult_val_accuracy": 0.8566360473632812, + "adult_val_precision": 0.8505148887634277, + "adult_val_recall": 0.8566360473632812, + "adult_val_f1-score": 0.8513188362121582, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9108703136444092, + "adult_val_score": 0.8566360473632812 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.330771803855896, + "lr": 0.0, + "adult_test_accuracy": 0.8531676530838013, + "adult_test_precision": 0.8469684720039368, + "adult_test_recall": 0.8531676530838013, + "adult_test_f1-score": 0.8474610447883606, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9055724143981934, + "adult_test_score": 0.8531676530838013 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.31575214862823486, + "lr": 0.0, + "adult_val_accuracy": 0.855592668056488, + "adult_val_precision": 0.8498003482818604, + "adult_val_recall": 0.855592668056488, + "adult_val_f1-score": 0.8450923562049866, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9096784591674805, + "adult_val_score": 0.855592668056488 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.33381417393684387, + "lr": 0.0, + "adult_test_accuracy": 0.8502926230430603, + "adult_test_precision": 0.8448350429534912, + "adult_test_recall": 0.8502926230430603, + "adult_test_f1-score": 0.8388490676879883, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.904146671295166, + "adult_test_score": 0.8502926230430603 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.3271609842777252, + "lr": 2.4471735741826706e-05, + "adult_val_accuracy": 0.8457846641540527, + "adult_val_precision": 0.8561062216758728, + "adult_val_recall": 0.8457846641540527, + "adult_val_f1-score": 0.8495416641235352, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9126606583595276, + "adult_val_score": 0.8457846641540527 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.33690372109413147, + "lr": 2.4471757569699548e-05, + "adult_test_accuracy": 0.8412567973136902, + "adult_test_precision": 0.8505634665489197, + "adult_test_recall": 0.8412567973136902, + "adult_test_f1-score": 0.844720721244812, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9071280360221863, + "adult_test_score": 0.8412567973136902 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.3137013018131256, + "lr": 0.0007938927737995982, + "adult_val_accuracy": 0.8543405532836914, + "adult_val_precision": 0.8482469916343689, + "adult_val_recall": 0.8543405532836914, + "adult_val_f1-score": 0.8493722677230835, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9096468687057495, + "adult_val_score": 0.8543405532836914 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32899782061576843, + "lr": 0.0007938928320072591, + "adult_test_accuracy": 0.8522435426712036, + "adult_test_precision": 0.8462502360343933, + "adult_test_recall": 0.8522435426712036, + "adult_test_f1-score": 0.8471896052360535, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9038238525390625, + "adult_test_score": 0.8522435426712036 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.3165111839771271, + "lr": 0.000904508342500776, + "adult_val_accuracy": 0.8568447232246399, + "adult_val_precision": 0.8525021076202393, + "adult_val_recall": 0.8568447232246399, + "adult_val_f1-score": 0.8539435267448425, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.91126549243927, + "adult_val_score": 0.8568447232246399 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.329866498708725, + "lr": 0.0009045081678777933, + "adult_test_accuracy": 0.8536810874938965, + "adult_test_precision": 0.8490302562713623, + "adult_test_recall": 0.8536810874938965, + "adult_test_f1-score": 0.8504318594932556, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9066497683525085, + "adult_test_score": 0.8536810874938965 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.3320708036422729, + "std": 0.003255227198166325, + "n": 5, + "ci95": 0.004041250441864336 + }, + "lr": { + "mean": 0.0003445745514909504, + "std": 0.0004624234027687046, + "n": 5, + "ci95": 0.0005740824424851568 + }, + "adult_test_accuracy": { + "mean": 0.8501283407211304, + "std": 0.005125055601412339, + "n": 5, + "ci95": 0.006362576850381996 + }, + "adult_test_precision": { + "mean": 0.8475294947624207, + "std": 0.0022731360434104604, + "n": 5, + "ci95": 0.0028220187042627735 + }, + "adult_test_recall": { + "mean": 0.8501283407211304, + "std": 0.005125055601412339, + "n": 5, + "ci95": 0.006362576850381996 + }, + "adult_test_f1-score": { + "mean": 0.845730459690094, + "std": 0.004347422173838448, + "n": 5, + "ci95": 0.005397172213015454 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9054641485214233, + "std": 0.0014673321582540741, + "n": 5, + "ci95": 0.0018216414314325973 + }, + "adult_test_score": { + "mean": 0.8501283407211304, + "std": 0.005125055601412339, + "n": 5, + "ci95": 0.006362576850381996 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..b273adc59fd0acb7b1134e596c0b342e4a8572bc --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6160f1675fdd2ecdcb157dad93cea55d34a599874bcaa060b34f853312d44d19 +size 175049856 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..731db5194b1dcc2cfd9d23f7c9f655e5ae20c92c --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:61a1960ba84353627746a7dd5a775f5531412092267ccac5b7c2b3282de378ce +size 15766102 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..18ffce949beb5b66377af9c64032f4320ae7446f --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.18_sx0.001_sy0.0002/seed_42/wjepa_deterministic/adult/20260915_191033/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: adult +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.001 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.3 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.1 +mask_min_trgt_share: 0.1 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 256 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.1 +recon_y_weight: 1.0 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.001 +sw_s_y_weight: 0.0002 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 15 +vtjepa_kl_x_end_weight: 0.0001 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 15 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 15 +wjepa_joint_end_weight: 0.18 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..7d418a3f3d46c379f14e658612ca57d3eeb1a9bb --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-15T23:11:16.415299+00:00", + "spec": { + "dataset": "adult", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.002, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0004, + "lambda_joint": 0.36, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..6e36a089df5a1363e5a6f355f34ef901eb38f413 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "adult", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.002, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0004, + "lambda_joint": 0.36, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "adult", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.36", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.002", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "0.0004", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..1b1e9289d15a9e0b87a443e7a1125eb24da720a9 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "adult", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.36, + "s_x": 0.002, + "z": 0.0, + "s_y": 0.0004 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 15 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "adult", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.36", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.002", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "0.0004", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/downstream_results/autoint_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/downstream_results/autoint_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..05d9570983f271369e9ed59ad1ccc44076826b6e --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/downstream_results/autoint_adult_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "adult", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpavi2iibf", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpavi2iibf/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.31602638959884644, + "lr": 6.184666563058272e-05, + "adult_val_accuracy": 0.8578881621360779, + "adult_val_precision": 0.8522117137908936, + "adult_val_recall": 0.8578881621360779, + "adult_val_f1-score": 0.853290855884552, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9106281995773315, + "adult_val_score": 0.8578881621360779 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3371359407901764, + "lr": 6.18466583546251e-05, + "adult_test_accuracy": 0.8535783886909485, + "adult_test_precision": 0.8476585149765015, + "adult_test_recall": 0.8535783886909485, + "adult_test_f1-score": 0.848501443862915, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9044072031974792, + "adult_test_score": 0.8535783886909485 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.32009273767471313, + "lr": 4.758648356073536e-05, + "adult_val_accuracy": 0.8543405532836914, + "adult_val_precision": 0.8485678434371948, + "adult_val_recall": 0.8543405532836914, + "adult_val_f1-score": 0.8498815894126892, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.908366858959198, + "adult_val_score": 0.8543405532836914 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.331030011177063, + "lr": 4.758644354296848e-05, + "adult_test_accuracy": 0.8547078967094421, + "adult_test_precision": 0.8490305542945862, + "adult_test_recall": 0.8547078967094421, + "adult_test_f1-score": 0.8499814867973328, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.906295120716095, + "adult_test_score": 0.8547078967094421 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.31484881043434143, + "lr": 6.184666563058272e-05, + "adult_val_accuracy": 0.8568447232246399, + "adult_val_precision": 0.8508895635604858, + "adult_val_recall": 0.8568447232246399, + "adult_val_f1-score": 0.8518489599227905, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9097610712051392, + "adult_val_score": 0.8568447232246399 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32790663838386536, + "lr": 6.18466583546251e-05, + "adult_test_accuracy": 0.8492658138275146, + "adult_test_precision": 0.8427430987358093, + "adult_test_recall": 0.8492658138275146, + "adult_test_f1-score": 0.8434947729110718, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9061350226402283, + "adult_test_score": 0.8492658138275146 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.3138065040111542, + "lr": 8.856373824528418e-06, + "adult_val_accuracy": 0.8551753163337708, + "adult_val_precision": 0.8492802381515503, + "adult_val_recall": 0.8551753163337708, + "adult_val_f1-score": 0.8504620790481567, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.911631166934967, + "adult_val_score": 0.8551753163337708 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3378441333770752, + "lr": 8.856371096044313e-06, + "adult_test_accuracy": 0.8533730506896973, + "adult_test_precision": 0.8475449681282043, + "adult_test_recall": 0.8533730506896973, + "adult_test_f1-score": 0.8485080599784851, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9052463173866272, + "adult_test_score": 0.8533730506896973 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.3168400228023529, + "lr": 0.0001147433795267716, + "adult_val_accuracy": 0.8551753163337708, + "adult_val_precision": 0.8506505489349365, + "adult_val_recall": 0.8551753163337708, + "adult_val_f1-score": 0.8521369695663452, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9096913933753967, + "adult_val_score": 0.8551753163337708 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.33254772424697876, + "lr": 0.00011474348866613582, + "adult_test_accuracy": 0.8536810874938965, + "adult_test_precision": 0.8484154939651489, + "adult_test_recall": 0.8536810874938965, + "adult_test_f1-score": 0.8496910929679871, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9055430889129639, + "adult_test_score": 0.8536810874938965 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.3332928895950317, + "std": 0.004188390343924828, + "n": 5, + "ci95": 0.005199739771657463 + }, + "lr": { + "mean": 5.897592400287977e-05, + "std": 3.79743115429194e-05, + "n": 5, + "ci95": 4.7143776443242836e-05 + }, + "adult_test_accuracy": { + "mean": 0.8529212474822998, + "std": 0.002107583225813775, + "n": 5, + "ci95": 0.0026164906853148604 + }, + "adult_test_precision": { + "mean": 0.8470785260200501, + "std": 0.002497387777139062, + "n": 5, + "ci95": 0.003100419369669493 + }, + "adult_test_recall": { + "mean": 0.8529212474822998, + "std": 0.002107583225813775, + "n": 5, + "ci95": 0.0026164906853148604 + }, + "adult_test_f1-score": { + "mean": 0.8480353713035583, + "std": 0.0026261412624327906, + "n": 5, + "ci95": 0.003260262307707106 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9055253505706787, + "std": 0.0007570788968041613, + "n": 5, + "ci95": 0.0009398869080349992 + }, + "adult_test_score": { + "mean": 0.8529212474822998, + "std": 0.002107583225813775, + "n": 5, + "ci95": 0.0026164906853148604 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/downstream_results/dcnv2_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/downstream_results/dcnv2_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..64257712ac8c38d1c71499966712037cbb75be6a --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/downstream_results/dcnv2_adult_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "adult", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmptxo87g4d", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmptxo87g4d/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.3184669017791748, + "lr": 0.0003454915713518858, + "adult_val_accuracy": 0.8530884981155396, + "adult_val_precision": 0.8503730297088623, + "adult_val_recall": 0.8530884981155396, + "adult_val_f1-score": 0.851518988609314, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9111711382865906, + "adult_val_score": 0.8530884981155396 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3455669581890106, + "lr": 0.0003454913094174117, + "adult_test_accuracy": 0.8499845862388611, + "adult_test_precision": 0.8466554284095764, + "adult_test_recall": 0.8499845862388611, + "adult_test_f1-score": 0.8479675650596619, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9031599760055542, + "adult_test_score": 0.8499845862388611 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.31946510076522827, + "lr": 0.00028711033519357443, + "adult_val_accuracy": 0.8593488931655884, + "adult_val_precision": 0.8537064790725708, + "adult_val_recall": 0.8593488931655884, + "adult_val_f1-score": 0.854661762714386, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9112556576728821, + "adult_val_score": 0.8593488931655884 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.34291860461235046, + "lr": 0.0002871101605705917, + "adult_test_accuracy": 0.851216733455658, + "adult_test_precision": 0.844944417476654, + "adult_test_recall": 0.851216733455658, + "adult_test_f1-score": 0.8457058072090149, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9016596078872681, + "adult_test_score": 0.851216733455658 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.33075979351997375, + "lr": 0.0002320865314686671, + "adult_val_accuracy": 0.8551753163337708, + "adult_val_precision": 0.8493150472640991, + "adult_val_recall": 0.8551753163337708, + "adult_val_f1-score": 0.8505183458328247, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9074548482894897, + "adult_val_score": 0.8551753163337708 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.35528725385665894, + "lr": 0.00023208664788398892, + "adult_test_accuracy": 0.8492658138275146, + "adult_test_precision": 0.843364417552948, + "adult_test_recall": 0.8492658138275146, + "adult_test_f1-score": 0.8446530699729919, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9008036255836487, + "adult_test_score": 0.8492658138275146 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.32828187942504883, + "lr": 0.0001147433795267716, + "adult_val_accuracy": 0.8537145256996155, + "adult_val_precision": 0.8495492339134216, + "adult_val_recall": 0.8537145256996155, + "adult_val_f1-score": 0.8510338068008423, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9080153107643127, + "adult_val_score": 0.8537145256996155 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3575955629348755, + "lr": 0.00011474348866613582, + "adult_test_accuracy": 0.8502926230430603, + "adult_test_precision": 0.8461775779724121, + "adult_test_recall": 0.8502926230430603, + "adult_test_f1-score": 0.8476355671882629, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9010512828826904, + "adult_test_score": 0.8502926230430603 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.3409920930862427, + "lr": 7.783600449329242e-05, + "adult_val_accuracy": 0.8541318774223328, + "adult_val_precision": 0.8493201732635498, + "adult_val_recall": 0.8541318774223328, + "adult_val_f1-score": 0.8508355617523193, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.903526246547699, + "adult_val_score": 0.8541318774223328 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.36198142170906067, + "lr": 7.783606997691095e-05, + "adult_test_accuracy": 0.8496765494346619, + "adult_test_precision": 0.8445515036582947, + "adult_test_recall": 0.8496765494346619, + "adult_test_f1-score": 0.8460401296615601, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.89734947681427, + "adult_test_score": 0.8496765494346619 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.3526699602603912, + "std": 0.008114133921538105, + "n": 5, + "ci95": 0.010073412791043781 + }, + "lr": { + "mean": 0.00021145353530300782, + "std": 0.00011327205671731186, + "n": 5, + "ci95": 0.00014062328722172637 + }, + "adult_test_accuracy": { + "mean": 0.8500872611999511, + "std": 0.0007368655152877856, + "n": 5, + "ci95": 0.0009147927035411806 + }, + "adult_test_precision": { + "mean": 0.845138669013977, + "std": 0.0013143777947840238, + "n": 5, + "ci95": 0.0016317539515950574 + }, + "adult_test_recall": { + "mean": 0.8500872611999511, + "std": 0.0007368655152877856, + "n": 5, + "ci95": 0.0009147927035411806 + }, + "adult_test_f1-score": { + "mean": 0.8464004278182984, + "std": 0.001382665938382185, + "n": 5, + "ci95": 0.0017165312877655153 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9008047938346863, + "std": 0.0021376133917081572, + "n": 5, + "ci95": 0.0026537720834483897 + }, + "adult_test_score": { + "mean": 0.8500872611999511, + "std": 0.0007368655152877856, + "n": 5, + "ci95": 0.0009147927035411806 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/downstream_results/fttransformer_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/downstream_results/fttransformer_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..58f53422f372cba793ed3edfd88c0a446b086c9f --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/downstream_results/fttransformer_adult_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "adult", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp336any22", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp336any22/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.30802208185195923, + "lr": 9.866357686405536e-07, + "adult_val_accuracy": 0.8547579050064087, + "adult_val_precision": 0.8494271636009216, + "adult_val_recall": 0.8547579050064087, + "adult_val_f1-score": 0.8508616089820862, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9123208522796631, + "adult_val_score": 0.8547579050064087 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.323154479265213, + "lr": 9.866350865195272e-07, + "adult_test_accuracy": 0.8526542782783508, + "adult_test_precision": 0.8468689322471619, + "adult_test_recall": 0.8526542782783508, + "adult_test_f1-score": 0.8479426503181458, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9077871441841125, + "adult_test_score": 0.8526542782783508 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.3077408969402313, + "lr": 8.856373824528418e-06, + "adult_val_accuracy": 0.8593488931655884, + "adult_val_precision": 0.8538922071456909, + "adult_val_recall": 0.8593488931655884, + "adult_val_f1-score": 0.854989230632782, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9120140671730042, + "adult_val_score": 0.8593488931655884 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32276657223701477, + "lr": 8.856371096044313e-06, + "adult_test_accuracy": 0.85296231508255, + "adult_test_precision": 0.8473031520843506, + "adult_test_recall": 0.85296231508255, + "adult_test_f1-score": 0.8484358191490173, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9076554179191589, + "adult_test_score": 0.85296231508255 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.30735018849372864, + "lr": 1.570841959619429e-05, + "adult_val_accuracy": 0.8589315414428711, + "adult_val_precision": 0.8538520336151123, + "adult_val_recall": 0.8589315414428711, + "adult_val_f1-score": 0.8551472425460815, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9137529134750366, + "adult_val_score": 0.8589315414428711 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3216746151447296, + "lr": 1.5708421415183693e-05, + "adult_test_accuracy": 0.8536810874938965, + "adult_test_precision": 0.848214864730835, + "adult_test_recall": 0.8536810874938965, + "adult_test_f1-score": 0.8494026064872742, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9077582955360413, + "adult_test_score": 0.8536810874938965 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.3095518946647644, + "lr": 0.0, + "adult_val_accuracy": 0.8576794862747192, + "adult_val_precision": 0.8522859215736389, + "adult_val_recall": 0.8576794862747192, + "adult_val_f1-score": 0.8535400032997131, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9127209782600403, + "adult_val_score": 0.8576794862747192 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.322591096162796, + "lr": 0.0, + "adult_test_accuracy": 0.8517301678657532, + "adult_test_precision": 0.8458373546600342, + "adult_test_recall": 0.8517301678657532, + "adult_test_f1-score": 0.8469206094741821, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9081859588623047, + "adult_test_score": 0.8517301678657532 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.31376832723617554, + "lr": 9.866357686405536e-07, + "adult_val_accuracy": 0.8551753163337708, + "adult_val_precision": 0.8494234085083008, + "adult_val_recall": 0.8551753163337708, + "adult_val_f1-score": 0.8506861925125122, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9110867977142334, + "adult_val_score": 0.8551753163337708 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32766711711883545, + "lr": 9.866350865195272e-07, + "adult_test_accuracy": 0.8520382046699524, + "adult_test_precision": 0.8463738560676575, + "adult_test_recall": 0.8520382046699524, + "adult_test_f1-score": 0.8475776314735413, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9062919020652771, + "adult_test_score": 0.8520382046699524 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.32357077598571776, + "std": 0.002353468493423043, + "n": 5, + "ci95": 0.0029217486245866967 + }, + "lr": { + "mean": 5.307612536853412e-06, + "std": 6.824275707519004e-06, + "n": 5, + "ci95": 8.472099039339038e-06 + }, + "adult_test_accuracy": { + "mean": 0.8526132106781006, + "std": 0.0007704410821983181, + "n": 5, + "ci95": 0.0009564755927384287 + }, + "adult_test_precision": { + "mean": 0.8469196319580078, + "std": 0.0009078107671745471, + "n": 5, + "ci95": 0.0011270152406074514 + }, + "adult_test_recall": { + "mean": 0.8526132106781006, + "std": 0.0007704410821983181, + "n": 5, + "ci95": 0.0009564755927384287 + }, + "adult_test_f1-score": { + "mean": 0.8480558633804322, + "std": 0.0009338639847167716, + "n": 5, + "ci95": 0.0011593593967891799 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9075357437133789, + "std": 0.0007240468236636543, + "n": 5, + "ci95": 0.0008988787472989483 + }, + "adult_test_score": { + "mean": 0.8526132106781006, + "std": 0.0007704410821983181, + "n": 5, + "ci95": 0.0009564755927384287 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/downstream_results/mlp_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/downstream_results/mlp_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..1d7ca8d7b3dca522286c002d569a6d6901d820b4 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/downstream_results/mlp_adult_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "adult", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp85sk7n7c", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp85sk7n7c/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.3251536190509796, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8583055138587952, + "adult_val_precision": 0.8521308302879333, + "adult_val_recall": 0.8583055138587952, + "adult_val_f1-score": 0.8524736166000366, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.911405086517334, + "adult_val_score": 0.8583055138587952 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32946667075157166, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8499845862388611, + "adult_test_precision": 0.8434247970581055, + "adult_test_recall": 0.8499845862388611, + "adult_test_f1-score": 0.8439649343490601, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9052516222000122, + "adult_test_score": 0.8499845862388611 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.3070836365222931, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8622704744338989, + "adult_val_precision": 0.8568785190582275, + "adult_val_recall": 0.8622704744338989, + "adult_val_f1-score": 0.8577345013618469, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9135227203369141, + "adult_val_score": 0.8622704744338989 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3215703070163727, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.851319432258606, + "adult_test_precision": 0.8454440236091614, + "adult_test_recall": 0.851319432258606, + "adult_test_f1-score": 0.84657883644104, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9062159061431885, + "adult_test_score": 0.851319432258606 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.31316646933555603, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8580968379974365, + "adult_val_precision": 0.8517271876335144, + "adult_val_recall": 0.8580968379974365, + "adult_val_f1-score": 0.8508983254432678, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9109476804733276, + "adult_val_score": 0.8580968379974365 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32460489869117737, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8492658138275146, + "adult_test_precision": 0.8423492312431335, + "adult_test_recall": 0.8492658138275146, + "adult_test_f1-score": 0.8418701887130737, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9060997366905212, + "adult_test_score": 0.8492658138275146 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.3275885283946991, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8541318774223328, + "adult_val_precision": 0.8484793305397034, + "adult_val_recall": 0.8541318774223328, + "adult_val_f1-score": 0.842825710773468, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9097819328308105, + "adult_val_score": 0.8541318774223328 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3439422845840454, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8451586365699768, + "adult_test_precision": 0.8386726975440979, + "adult_test_recall": 0.8451586365699768, + "adult_test_f1-score": 0.8332472443580627, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9046995639801025, + "adult_test_score": 0.8451586365699768 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.31227654218673706, + "lr": 0.0009999998146668077, + "adult_val_accuracy": 0.8528798222541809, + "adult_val_precision": 0.8494364023208618, + "adult_val_recall": 0.8528798222541809, + "adult_val_f1-score": 0.8507920503616333, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9111184477806091, + "adult_val_score": 0.8528798222541809 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32541224360466003, + "lr": 0.0009999996982514858, + "adult_test_accuracy": 0.8497792482376099, + "adult_test_precision": 0.8460406064987183, + "adult_test_recall": 0.8497792482376099, + "adult_test_f1-score": 0.8474451899528503, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9047033786773682, + "adult_test_score": 0.8497792482376099 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.32899928092956543, + "std": 0.008815879095327652, + "n": 5, + "ci95": 0.010944604821895161 + }, + "lr": { + "mean": 0.0009999996982514858, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_accuracy": { + "mean": 0.8491015434265137, + "std": 0.0023308570599090517, + "n": 5, + "ci95": 0.0028936773226108845 + }, + "adult_test_precision": { + "mean": 0.8431862711906433, + "std": 0.002931490813571333, + "n": 5, + "ci95": 0.003639343070228635 + }, + "adult_test_recall": { + "mean": 0.8491015434265137, + "std": 0.0023308570599090517, + "n": 5, + "ci95": 0.0028936773226108845 + }, + "adult_test_f1-score": { + "mean": 0.8426212787628173, + "std": 0.005682746905427274, + "n": 5, + "ci95": 0.007054931052277275 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9053940415382385, + "std": 0.0007336662406820011, + "n": 5, + "ci95": 0.0009108209162811225 + }, + "adult_test_score": { + "mean": 0.8491015434265137, + "std": 0.0023308570599090517, + "n": 5, + "ci95": 0.0028936773226108845 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/downstream_results/resnet_adult_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/downstream_results/resnet_adult_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..4403842fb8d35930cd4d3c4b262febd500ccf09a --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/downstream_results/resnet_adult_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "adult", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp9m1is_c2", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "adult", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp9m1is_c2/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "adult_val_loss": 0.30838286876678467, + "lr": 0.00020610730280168355, + "adult_val_accuracy": 0.8583055138587952, + "adult_val_precision": 0.8526843190193176, + "adult_val_recall": 0.8583055138587952, + "adult_val_f1-score": 0.8485957980155945, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9123663306236267, + "adult_val_score": 0.8583055138587952 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32302045822143555, + "lr": 0.00020610728824976832, + "adult_test_accuracy": 0.8516274690628052, + "adult_test_precision": 0.84550940990448, + "adult_test_recall": 0.8516274690628052, + "adult_test_f1-score": 0.8418778777122498, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9073248505592346, + "adult_test_score": 0.8516274690628052 + } + } + }, + { + "seed": 1, + "val_metrics": { + "adult_val_loss": 0.3170086145401001, + "lr": 0.0, + "adult_val_accuracy": 0.8510016798973083, + "adult_val_precision": 0.8443875312805176, + "adult_val_recall": 0.8510016798973083, + "adult_val_f1-score": 0.840018630027771, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9078263640403748, + "adult_val_score": 0.8510016798973083 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3321130871772766, + "lr": 0.0, + "adult_test_accuracy": 0.848444402217865, + "adult_test_precision": 0.8430157899856567, + "adult_test_recall": 0.848444402217865, + "adult_test_f1-score": 0.8362640142440796, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9025493264198303, + "adult_test_score": 0.848444402217865 + } + } + }, + { + "seed": 2, + "val_metrics": { + "adult_val_loss": 0.3102990686893463, + "lr": 0.0009755284409038723, + "adult_val_accuracy": 0.8547579050064087, + "adult_val_precision": 0.8555600643157959, + "adult_val_recall": 0.8547579050064087, + "adult_val_f1-score": 0.8551446199417114, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9138563275337219, + "adult_val_score": 0.8547579050064087 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.32634368538856506, + "lr": 0.0009755276259966195, + "adult_test_accuracy": 0.8524489402770996, + "adult_test_precision": 0.8521803617477417, + "adult_test_recall": 0.8524489402770996, + "adult_test_f1-score": 0.8523128032684326, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9077268838882446, + "adult_test_score": 0.8524489402770996 + } + } + }, + { + "seed": 3, + "val_metrics": { + "adult_val_loss": 0.31046244502067566, + "lr": 0.0, + "adult_val_accuracy": 0.8568447232246399, + "adult_val_precision": 0.8517257571220398, + "adult_val_recall": 0.8568447232246399, + "adult_val_f1-score": 0.8531110882759094, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.9110794067382812, + "adult_val_score": 0.8568447232246399 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3221278190612793, + "lr": 0.0, + "adult_test_accuracy": 0.851422131061554, + "adult_test_precision": 0.8459579944610596, + "adult_test_recall": 0.851422131061554, + "adult_test_f1-score": 0.8472910523414612, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9058641195297241, + "adult_test_score": 0.851422131061554 + } + } + }, + { + "seed": 4, + "val_metrics": { + "adult_val_loss": 0.3071945309638977, + "lr": 0.000904508342500776, + "adult_val_accuracy": 0.8574708104133606, + "adult_val_precision": 0.8524536490440369, + "adult_val_recall": 0.8574708104133606, + "adult_val_f1-score": 0.8538326621055603, + "adult_val_support": 4792.0, + "adult_val_roc_auc": 0.912779688835144, + "adult_val_score": 0.8574708104133606 + }, + "test_metrics": { + "0": { + "adult_test_loss": 0.3212665319442749, + "lr": 0.0009045081678777933, + "adult_test_accuracy": 0.8534756898880005, + "adult_test_precision": 0.8479958772659302, + "adult_test_recall": 0.8534756898880005, + "adult_test_f1-score": 0.8491911888122559, + "adult_test_support": 9739.0, + "adult_test_roc_auc": 0.9076801538467407, + "adult_test_score": 0.8534756898880005 + } + } + } + ], + "summary": { + "0": { + "adult_test_loss": { + "mean": 0.3249743163585663, + "std": 0.004430152200792488, + "n": 5, + "ci95": 0.005499878641055805 + }, + "lr": { + "mean": 0.0004172286164248362, + "std": 0.00048525014098169003, + "n": 5, + "ci95": 0.0006024210376964257 + }, + "adult_test_accuracy": { + "mean": 0.8514837265014649, + "std": 0.0018815886764255846, + "n": 5, + "ci95": 0.002335926375367948 + }, + "adult_test_precision": { + "mean": 0.8469318866729736, + "std": 0.0034273271675646553, + "n": 5, + "ci95": 0.004254906520238103 + }, + "adult_test_recall": { + "mean": 0.8514837265014649, + "std": 0.0018815886764255846, + "n": 5, + "ci95": 0.002335926375367948 + }, + "adult_test_f1-score": { + "mean": 0.8453873872756958, + "std": 0.006356233679124539, + "n": 5, + "ci95": 0.007891041270122289 + }, + "adult_test_support": { + "mean": 9739.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "adult_test_roc_auc": { + "mean": 0.9062290668487549, + "std": 0.0021922321654874836, + "n": 5, + "ci95": 0.002721579376221726 + }, + "adult_test_score": { + "mean": 0.8514837265014649, + "std": 0.0018815886764255846, + "n": 5, + "ci95": 0.002335926375367948 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..3d65c47de48e4104a09eba079116fd6684b497ca --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:94cd166a131698bfa133ab20d3a0821507e82c64b5810207b1c0be293e8c6d10 +size 175049856 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..406efba713cfec3cf48e161d34217b63a6ebf056 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2ae049e9f7ad0d2c40cf04e796b05c6015a361011c6bccd141d974265d936bba +size 15769622 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4fa62dd118177f2f9e3a1ead2ed9733d9a568eb6 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/adult/deterministic_sw/sw_f0.36_sx0.002_sy0.0004/seed_42/wjepa_deterministic/adult/20260915_211715/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: adult +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.001 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.3 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.1 +mask_min_trgt_share: 0.1 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 256 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.1 +recon_y_weight: 1.0 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.002 +sw_s_y_weight: 0.0004 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 15 +vtjepa_kl_x_end_weight: 0.0001 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 15 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 15 +wjepa_joint_end_weight: 0.36 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..4bd235998632a02240cd853e2e7d6cde6bba33d1 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-16T04:18:33.587878+00:00", + "spec": { + "dataset": "bank_marketing", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.00025, + "sw_z_weight": 0.0, + "sw_s_y_weight": 2.5e-05, + "lambda_joint": 0.001, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..4352d7f6c6480f0160df520c1352327148057ffd --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "bank_marketing", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.00025, + "sw_z_weight": 0.0, + "sw_s_y_weight": 2.5e-05, + "lambda_joint": 0.001, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "bank_marketing", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.001", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.00025", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "2.5e-05", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..544e9bdd3ba4b8931bc10268edec7384750b0318 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "bank_marketing", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.001, + "s_x": 0.00025, + "z": 0.0, + "s_y": 2.5e-05 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 15 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "bank_marketing", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.001", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.00025", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "2.5e-05", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/downstream_results/autoint_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/downstream_results/autoint_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..3950111e1ed773f2b6b3ef8e8d87c90bfb937634 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/downstream_results/autoint_bank_marketing_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "bank_marketing", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp6zeicrxs", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp6zeicrxs/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.22611165046691895, + "lr": 3.942650891985977e-06, + "bank_marketing_val_accuracy": 0.8958191871643066, + "bank_marketing_val_precision": 0.886978268623352, + "bank_marketing_val_recall": 0.8958191871643066, + "bank_marketing_val_f1-score": 0.8903791308403015, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.916971743106842, + "bank_marketing_val_score": 0.8958191871643066 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22128261625766754, + "lr": 3.942649982491275e-06, + "bank_marketing_test_accuracy": 0.8994227647781372, + "bank_marketing_test_precision": 0.8907265663146973, + "bank_marketing_test_recall": 0.8994227647781372, + "bank_marketing_test_f1-score": 0.8938032984733582, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.922081708908081, + "bank_marketing_test_score": 0.8994227647781372 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.21729792654514313, + "lr": 9.866358823273913e-07, + "bank_marketing_val_accuracy": 0.897175133228302, + "bank_marketing_val_precision": 0.8896446824073792, + "bank_marketing_val_recall": 0.897175133228302, + "bank_marketing_val_f1-score": 0.8926685452461243, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9221920967102051, + "bank_marketing_val_score": 0.897175133228302 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2156086415052414, + "lr": 9.86635200206365e-07, + "bank_marketing_test_accuracy": 0.9013099670410156, + "bank_marketing_test_precision": 0.8926609754562378, + "bank_marketing_test_recall": 0.9013099670410156, + "bank_marketing_test_f1-score": 0.8956167697906494, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9255392551422119, + "bank_marketing_test_score": 0.9013099670410156 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.22027043998241425, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8989830613136292, + "bank_marketing_val_precision": 0.8910719752311707, + "bank_marketing_val_recall": 0.8989830613136292, + "bank_marketing_val_f1-score": 0.8941360116004944, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9212469458580017, + "bank_marketing_val_score": 0.8989830613136292 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.216365247964859, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.9020870327949524, + "bank_marketing_test_precision": 0.8938513398170471, + "bank_marketing_test_recall": 0.9020870327949524, + "bank_marketing_test_f1-score": 0.8967174291610718, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.925674319267273, + "bank_marketing_test_score": 0.9020870327949524 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.21749942004680634, + "lr": 6.184667290654033e-05, + "bank_marketing_val_accuracy": 0.9023728966712952, + "bank_marketing_val_precision": 0.8956892490386963, + "bank_marketing_val_recall": 0.9023728966712952, + "bank_marketing_val_f1-score": 0.8983437418937683, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9211104512214661, + "bank_marketing_val_score": 0.9023728966712952 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21675556898117065, + "lr": 6.184665107866749e-05, + "bank_marketing_test_accuracy": 0.8997557759284973, + "bank_marketing_test_precision": 0.8926168084144592, + "bank_marketing_test_recall": 0.8997557759284973, + "bank_marketing_test_f1-score": 0.8953888416290283, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9248812198638916, + "bank_marketing_test_score": 0.8997557759284973 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.22956803441047668, + "lr": 9.54915230977349e-05, + "bank_marketing_val_accuracy": 0.9016948938369751, + "bank_marketing_val_precision": 0.892932116985321, + "bank_marketing_val_recall": 0.9016948938369751, + "bank_marketing_val_f1-score": 0.8960297703742981, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9153397679328918, + "bank_marketing_val_score": 0.9016948938369751 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22201068699359894, + "lr": 9.549142851028591e-05, + "bank_marketing_test_accuracy": 0.8994227647781372, + "bank_marketing_test_precision": 0.8904885649681091, + "bank_marketing_test_recall": 0.8994227647781372, + "bank_marketing_test_f1-score": 0.8935944437980652, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9219614863395691, + "bank_marketing_test_score": 0.8994227647781372 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.2184045523405075, + "std": 0.002999270115467796, + "n": 5, + "ci95": 0.003723488697265861 + }, + "lr": { + "mean": 3.245347295433021e-05, + "std": 4.385772330073975e-05, + "n": 5, + "ci95": 5.44478258746787e-05 + }, + "bank_marketing_test_accuracy": { + "mean": 0.9003996610641479, + "std": 0.0012246551662196583, + "n": 5, + "ci95": 0.0015203664538083529 + }, + "bank_marketing_test_precision": { + "mean": 0.8920688509941102, + "std": 0.0014254093448568936, + "n": 5, + "ci95": 0.0017695957283673895 + }, + "bank_marketing_test_recall": { + "mean": 0.9003996610641479, + "std": 0.0012246551662196583, + "n": 5, + "ci95": 0.0015203664538083529 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8950241565704345, + "std": 0.0013120528563677772, + "n": 5, + "ci95": 0.0016288676220610524 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9240275979042053, + "std": 0.0018561229946868249, + "n": 5, + "ci95": 0.002304311624287867 + }, + "bank_marketing_test_score": { + "mean": 0.9003996610641479, + "std": 0.0012246551662196583, + "n": 5, + "ci95": 0.0015203664538083529 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/downstream_results/dcnv2_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/downstream_results/dcnv2_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..922934529c3263f7c1f0818b03df5059751004f8 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/downstream_results/dcnv2_bank_marketing_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "bank_marketing", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpxieoo8c7", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpxieoo8c7/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.24487870931625366, + "lr": 0.00023208656057249755, + "bank_marketing_val_accuracy": 0.8987570405006409, + "bank_marketing_val_precision": 0.89217209815979, + "bank_marketing_val_recall": 0.8987570405006409, + "bank_marketing_val_f1-score": 0.8948853015899658, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.906031608581543, + "bank_marketing_val_score": 0.8987570405006409 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23201237618923187, + "lr": 0.00023208656057249755, + "bank_marketing_test_accuracy": 0.8965364098548889, + "bank_marketing_test_precision": 0.8905078172683716, + "bank_marketing_test_recall": 0.8965364098548889, + "bank_marketing_test_f1-score": 0.8930413722991943, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9149787425994873, + "bank_marketing_test_score": 0.8965364098548889 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.26167696714401245, + "lr": 0.00015772644837852567, + "bank_marketing_val_accuracy": 0.8942372798919678, + "bank_marketing_val_precision": 0.8885349035263062, + "bank_marketing_val_recall": 0.8942372798919678, + "bank_marketing_val_f1-score": 0.8910248875617981, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9040787220001221, + "bank_marketing_val_score": 0.8942372798919678 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.24023163318634033, + "lr": 0.00015772631741128862, + "bank_marketing_test_accuracy": 0.8938721418380737, + "bank_marketing_test_precision": 0.8879433870315552, + "bank_marketing_test_recall": 0.8938721418380737, + "bank_marketing_test_f1-score": 0.890484094619751, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9127835035324097, + "bank_marketing_test_score": 0.8938721418380737 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.245857834815979, + "lr": 0.00013551567099057138, + "bank_marketing_val_accuracy": 0.8998870253562927, + "bank_marketing_val_precision": 0.8944838643074036, + "bank_marketing_val_recall": 0.8998870253562927, + "bank_marketing_val_f1-score": 0.8967978358268738, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9065213203430176, + "bank_marketing_val_score": 0.8998870253562927 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23830512166023254, + "lr": 0.00013551574375014752, + "bank_marketing_test_accuracy": 0.8946492075920105, + "bank_marketing_test_precision": 0.8885173201560974, + "bank_marketing_test_recall": 0.8946492075920105, + "bank_marketing_test_f1-score": 0.8911149501800537, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9128586053848267, + "bank_marketing_test_score": 0.8946492075920105 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.24853940308094025, + "lr": 9.54915230977349e-05, + "bank_marketing_val_accuracy": 0.900564968585968, + "bank_marketing_val_precision": 0.8931647539138794, + "bank_marketing_val_recall": 0.900564968585968, + "bank_marketing_val_f1-score": 0.8960529565811157, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9065751433372498, + "bank_marketing_val_score": 0.900564968585968 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.25023913383483887, + "lr": 9.549142851028591e-05, + "bank_marketing_test_accuracy": 0.8928729891777039, + "bank_marketing_test_precision": 0.8842678070068359, + "bank_marketing_test_recall": 0.8928729891777039, + "bank_marketing_test_f1-score": 0.8876234292984009, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9084752798080444, + "bank_marketing_test_score": 0.8928729891777039 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.24205172061920166, + "lr": 0.00020610730280168355, + "bank_marketing_val_accuracy": 0.8974011540412903, + "bank_marketing_val_precision": 0.8905527591705322, + "bank_marketing_val_recall": 0.8974011540412903, + "bank_marketing_val_f1-score": 0.8933743834495544, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9060702919960022, + "bank_marketing_val_score": 0.8974011540412903 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23046530783176422, + "lr": 0.00020610728824976832, + "bank_marketing_test_accuracy": 0.8984236121177673, + "bank_marketing_test_precision": 0.8912854790687561, + "bank_marketing_test_recall": 0.8984236121177673, + "bank_marketing_test_f1-score": 0.8940976858139038, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9146462678909302, + "bank_marketing_test_score": 0.8984236121177673 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.23825071454048158, + "std": 0.00786095762293505, + "n": 5, + "ci95": 0.009759103292408627 + }, + "lr": { + "mean": 0.0001653854676987976, + "std": 5.4642799538255315e-05, + "n": 5, + "ci95": 6.783711991073e-05 + }, + "bank_marketing_test_accuracy": { + "mean": 0.8952708721160889, + "std": 0.0022155197496209833, + "n": 5, + "ci95": 0.002750490095486566 + }, + "bank_marketing_test_precision": { + "mean": 0.8885043621063232, + "std": 0.002739146956701429, + "n": 5, + "ci95": 0.003400554915287177 + }, + "bank_marketing_test_recall": { + "mean": 0.8952708721160889, + "std": 0.0022155197496209833, + "n": 5, + "ci95": 0.002750490095486566 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8912723064422607, + "std": 0.0025036326548266426, + "n": 5, + "ci95": 0.003108172166380131 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9127484798431397, + "std": 0.002590813784815869, + "n": 5, + "ci95": 0.003216404482787924 + }, + "bank_marketing_test_score": { + "mean": 0.8952708721160889, + "std": 0.0022155197496209833, + "n": 5, + "ci95": 0.002750490095486566 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/downstream_results/fttransformer_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/downstream_results/fttransformer_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..5894d8beb51e4ce252f4159110d3c2eb51896e93 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/downstream_results/fttransformer_bank_marketing_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "bank_marketing", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpjqlksr1e", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpjqlksr1e/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.22565948963165283, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9021468758583069, + "bank_marketing_val_precision": 0.8941959142684937, + "bank_marketing_val_recall": 0.9021468758583069, + "bank_marketing_val_f1-score": 0.8971415758132935, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9094438552856445, + "bank_marketing_val_score": 0.9021468758583069 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2238014042377472, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.899533748626709, + "bank_marketing_test_precision": 0.8900641202926636, + "bank_marketing_test_recall": 0.899533748626709, + "bank_marketing_test_f1-score": 0.8932085037231445, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.91472327709198, + "bank_marketing_test_score": 0.899533748626709 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.22002482414245605, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8996610045433044, + "bank_marketing_val_precision": 0.8899199366569519, + "bank_marketing_val_recall": 0.8996610045433044, + "bank_marketing_val_f1-score": 0.8932660222053528, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9157553315162659, + "bank_marketing_val_score": 0.8996610045433044 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22054523229599, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8985346555709839, + "bank_marketing_test_precision": 0.8873898386955261, + "bank_marketing_test_recall": 0.8985346555709839, + "bank_marketing_test_f1-score": 0.8906089663505554, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9182788133621216, + "bank_marketing_test_score": 0.8985346555709839 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.22208501398563385, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9007909893989563, + "bank_marketing_val_precision": 0.8914057612419128, + "bank_marketing_val_recall": 0.9007909893989563, + "bank_marketing_val_f1-score": 0.8946347832679749, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9142630100250244, + "bank_marketing_val_score": 0.9007909893989563 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2201967090368271, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.9021980166435242, + "bank_marketing_test_precision": 0.8913424015045166, + "bank_marketing_test_recall": 0.9021980166435242, + "bank_marketing_test_f1-score": 0.894023597240448, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9188998341560364, + "bank_marketing_test_score": 0.9021980166435242 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.22542229294776917, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.900564968585968, + "bank_marketing_val_precision": 0.8903577923774719, + "bank_marketing_val_recall": 0.900564968585968, + "bank_marketing_val_f1-score": 0.8936613202095032, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9104665517807007, + "bank_marketing_val_score": 0.900564968585968 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22330550849437714, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.9007548689842224, + "bank_marketing_test_precision": 0.8899936676025391, + "bank_marketing_test_recall": 0.9007548689842224, + "bank_marketing_test_f1-score": 0.8929461240768433, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9152087569236755, + "bank_marketing_test_score": 0.9007548689842224 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.22800232470035553, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8962711691856384, + "bank_marketing_val_precision": 0.885810911655426, + "bank_marketing_val_recall": 0.8962711691856384, + "bank_marketing_val_f1-score": 0.8894844651222229, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9084431529045105, + "bank_marketing_val_score": 0.8962711691856384 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22737419605255127, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8992007374763489, + "bank_marketing_test_precision": 0.8878336548805237, + "bank_marketing_test_recall": 0.8992007374763489, + "bank_marketing_test_f1-score": 0.8909220695495605, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9117336869239807, + "bank_marketing_test_score": 0.8992007374763489 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.22304461002349854, + "std": 0.0029044599210716286, + "n": 5, + "ci95": 0.003605785164863396 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_accuracy": { + "mean": 0.9000444054603577, + "std": 0.0014486796009176627, + "n": 5, + "ci95": 0.0017984849354374377 + }, + "bank_marketing_test_precision": { + "mean": 0.8893247365951538, + "std": 0.001660744049586403, + "n": 5, + "ci95": 0.002061755513715051 + }, + "bank_marketing_test_recall": { + "mean": 0.9000444054603577, + "std": 0.0014486796009176627, + "n": 5, + "ci95": 0.0017984849354374377 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8923418521881104, + "std": 0.001496927844493347, + "n": 5, + "ci95": 0.0018583834383066836 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9157688736915588, + "std": 0.0029065635152159723, + "n": 5, + "ci95": 0.0036083967032439187 + }, + "bank_marketing_test_score": { + "mean": 0.9000444054603577, + "std": 0.0014486796009176627, + "n": 5, + "ci95": 0.0017984849354374377 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/downstream_results/mlp_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/downstream_results/mlp_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..5125295d6dfc6ac6ff5eac5183707c961b66c463 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/downstream_results/mlp_bank_marketing_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "bank_marketing", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpx8h7c71g", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpx8h7c71g/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.23036538064479828, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8996610045433044, + "bank_marketing_val_precision": 0.8884727954864502, + "bank_marketing_val_recall": 0.8996610045433044, + "bank_marketing_val_f1-score": 0.8918704390525818, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9046750664710999, + "bank_marketing_val_score": 0.8996610045433044 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23671753704547882, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8990896940231323, + "bank_marketing_test_precision": 0.8875277638435364, + "bank_marketing_test_recall": 0.8990896940231323, + "bank_marketing_test_f1-score": 0.8905965685844421, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9076991677284241, + "bank_marketing_test_score": 0.8990896940231323 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.2288859784603119, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8989830613136292, + "bank_marketing_val_precision": 0.8856574296951294, + "bank_marketing_val_recall": 0.8989830613136292, + "bank_marketing_val_f1-score": 0.8888009786605835, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9097251296043396, + "bank_marketing_val_score": 0.8989830613136292 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22853031754493713, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8994227647781372, + "bank_marketing_test_precision": 0.8855958580970764, + "bank_marketing_test_recall": 0.8994227647781372, + "bank_marketing_test_f1-score": 0.887127697467804, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9139648079872131, + "bank_marketing_test_score": 0.8994227647781372 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.22585520148277283, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.897175133228302, + "bank_marketing_val_precision": 0.883800208568573, + "bank_marketing_val_recall": 0.897175133228302, + "bank_marketing_val_f1-score": 0.8873447775840759, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.908288836479187, + "bank_marketing_val_score": 0.897175133228302 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2233278453350067, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8973134756088257, + "bank_marketing_test_precision": 0.8842926621437073, + "bank_marketing_test_recall": 0.8973134756088257, + "bank_marketing_test_f1-score": 0.8873161673545837, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9146900773048401, + "bank_marketing_test_score": 0.8973134756088257 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.2245796173810959, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8987570405006409, + "bank_marketing_val_precision": 0.8897973895072937, + "bank_marketing_val_recall": 0.8987570405006409, + "bank_marketing_val_f1-score": 0.8930883407592773, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9098792672157288, + "bank_marketing_val_score": 0.8987570405006409 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22117182612419128, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.9010879397392273, + "bank_marketing_test_precision": 0.8910074830055237, + "bank_marketing_test_recall": 0.9010879397392273, + "bank_marketing_test_f1-score": 0.8940004110336304, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9165617227554321, + "bank_marketing_test_score": 0.9010879397392273 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.22811979055404663, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.9003390073776245, + "bank_marketing_val_precision": 0.8861464858055115, + "bank_marketing_val_recall": 0.9003390073776245, + "bank_marketing_val_f1-score": 0.8882543444633484, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9083015322685242, + "bank_marketing_val_score": 0.9003390073776245 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22718489170074463, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8990896940231323, + "bank_marketing_test_precision": 0.8846362829208374, + "bank_marketing_test_recall": 0.8990896940231323, + "bank_marketing_test_f1-score": 0.8852412700653076, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.913707435131073, + "bank_marketing_test_score": 0.8990896940231323 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.2273864835500717, + "std": 0.0059898260804782735, + "n": 5, + "ci95": 0.007436159082247422 + }, + "lr": { + "mean": 0.0009999994654208422, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_accuracy": { + "mean": 0.899200713634491, + "std": 0.001341384280745296, + "n": 5, + "ci95": 0.001665281557098499 + }, + "bank_marketing_test_precision": { + "mean": 0.8866120100021362, + "std": 0.0027601732148490373, + "n": 5, + "ci95": 0.0034266582776201165 + }, + "bank_marketing_test_recall": { + "mean": 0.899200713634491, + "std": 0.001341384280745296, + "n": 5, + "ci95": 0.001665281557098499 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8888564229011535, + "std": 0.003461175006601948, + "n": 5, + "ci95": 0.004296927425735165 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9133246421813965, + "std": 0.003337117365385752, + "n": 5, + "ci95": 0.004142914213488717 + }, + "bank_marketing_test_score": { + "mean": 0.899200713634491, + "std": 0.001341384280745296, + "n": 5, + "ci95": 0.001665281557098499 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/downstream_results/resnet_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/downstream_results/resnet_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..9ee507cdbf00931df05493bc38c9f534f23ae263 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/downstream_results/resnet_bank_marketing_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "bank_marketing", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp6zdiq6kq", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp6zdiq6kq/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.21832960844039917, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9014689326286316, + "bank_marketing_val_precision": 0.8897938132286072, + "bank_marketing_val_recall": 0.9014689326286316, + "bank_marketing_val_f1-score": 0.8928584456443787, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9228784441947937, + "bank_marketing_val_score": 0.9014689326286316 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21502749621868134, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.900865912437439, + "bank_marketing_test_precision": 0.8898667097091675, + "bank_marketing_test_recall": 0.900865912437439, + "bank_marketing_test_f1-score": 0.8927526473999023, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9280371069908142, + "bank_marketing_test_score": 0.900865912437439 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.2157387137413025, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9014689326286316, + "bank_marketing_val_precision": 0.8933734893798828, + "bank_marketing_val_recall": 0.9014689326286316, + "bank_marketing_val_f1-score": 0.8963764309883118, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9216016530990601, + "bank_marketing_val_score": 0.9014689326286316 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2096211016178131, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.9004218578338623, + "bank_marketing_test_precision": 0.8919440507888794, + "bank_marketing_test_recall": 0.9004218578338623, + "bank_marketing_test_f1-score": 0.8949353098869324, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9289332032203674, + "bank_marketing_test_score": 0.9004218578338623 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.21583814918994904, + "lr": 0.0004999999655410647, + "bank_marketing_val_accuracy": 0.900564968585968, + "bank_marketing_val_precision": 0.8888975381851196, + "bank_marketing_val_recall": 0.900564968585968, + "bank_marketing_val_f1-score": 0.892121434211731, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9213452935218811, + "bank_marketing_val_score": 0.900564968585968 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21264716982841492, + "lr": 0.0004999997327104211, + "bank_marketing_test_accuracy": 0.9006438851356506, + "bank_marketing_test_precision": 0.8897886276245117, + "bank_marketing_test_recall": 0.9006438851356506, + "bank_marketing_test_f1-score": 0.8927411437034607, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9267732501029968, + "bank_marketing_test_score": 0.9006438851356506 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.21576574444770813, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9012429118156433, + "bank_marketing_val_precision": 0.893185555934906, + "bank_marketing_val_recall": 0.9012429118156433, + "bank_marketing_val_f1-score": 0.8961913585662842, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9206241965293884, + "bank_marketing_val_score": 0.9012429118156433 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21562933921813965, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8978685736656189, + "bank_marketing_test_precision": 0.8903787136077881, + "bank_marketing_test_recall": 0.8978685736656189, + "bank_marketing_test_f1-score": 0.8932943940162659, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9236408472061157, + "bank_marketing_test_score": 0.8978685736656189 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.22064602375030518, + "lr": 0.0007938927737995982, + "bank_marketing_val_accuracy": 0.9014689326286316, + "bank_marketing_val_precision": 0.8910717368125916, + "bank_marketing_val_recall": 0.9014689326286316, + "bank_marketing_val_f1-score": 0.8942846059799194, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9185978770256042, + "bank_marketing_val_score": 0.9014689326286316 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21673761308193207, + "lr": 0.0007938928320072591, + "bank_marketing_test_accuracy": 0.899533748626709, + "bank_marketing_test_precision": 0.8890221118927002, + "bank_marketing_test_recall": 0.899533748626709, + "bank_marketing_test_f1-score": 0.892167329788208, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9247206449508667, + "bank_marketing_test_score": 0.899533748626709 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.21393254399299622, + "std": 0.002836686848411513, + "n": 5, + "ci95": 0.003521647271204706 + }, + "lr": { + "mean": 0.00025877851294353604, + "std": 0.0003692675072412476, + "n": 5, + "ci95": 0.0004584326641303102 + }, + "bank_marketing_test_accuracy": { + "mean": 0.8998667955398559, + "std": 0.0012261737101846558, + "n": 5, + "ci95": 0.0015222516729024283 + }, + "bank_marketing_test_precision": { + "mean": 0.8902000427246094, + "std": 0.001088735704988695, + "n": 5, + "ci95": 0.0013516272078758398 + }, + "bank_marketing_test_recall": { + "mean": 0.8998667955398559, + "std": 0.0012261737101846558, + "n": 5, + "ci95": 0.0015222516729024283 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8931781649589539, + "std": 0.0010600599045472658, + "n": 5, + "ci95": 0.0013160272069695994 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9264210104942322, + "std": 0.0022174029748305416, + "n": 5, + "ci95": 0.002752828053560443 + }, + "bank_marketing_test_score": { + "mean": 0.8998667955398559, + "std": 0.0012261737101846558, + "n": 5, + "ci95": 0.0015222516729024283 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..0eb6178f1e2d11f5fa100e22bf22493e50f3f570 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:16bd4a80faf4cd5c6c31b41c20ff461322b949ec994278cc4567ca36ecf90fd8 +size 185184384 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..8cdffb7bd5798b50a2f30dfa5e7575c7b2173d6d --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3aaff4065f0097bdf175fedd36d761d20c089d9faaaa1cd7e25b9592dca1d772 +size 15472814 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..52b3745494ff9e189feaf6177f1b0e92543de88e --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.001_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_024237/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: bank_marketing +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.001 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.3 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.1 +mask_min_trgt_share: 0.1 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 256 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.1 +recon_y_weight: 1.0 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.00025 +sw_s_y_weight: 2.5e-05 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 15 +vtjepa_kl_x_end_weight: 0.0001 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 15 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 15 +wjepa_joint_end_weight: 0.001 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..fb3764f2f62e5a13f7126eb2d4a42489c7fb0f36 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-16T03:13:12.250591+00:00", + "spec": { + "dataset": "bank_marketing", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.0005, + "sw_z_weight": 0.0, + "sw_s_y_weight": 5e-05, + "lambda_joint": 0.002, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..19fbfed403fc4b9024841aa505dbf5dc10be3a3b --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "bank_marketing", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.0005, + "sw_z_weight": 0.0, + "sw_s_y_weight": 5e-05, + "lambda_joint": 0.002, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "bank_marketing", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.002", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.0005", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "5e-05", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..1a543c98ac126cd4134010c2df3cd92ee905c70f --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "bank_marketing", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.002, + "s_x": 0.0005, + "z": 0.0, + "s_y": 5e-05 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 15 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "bank_marketing", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.002", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.0005", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "5e-05", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/downstream_results/autoint_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/downstream_results/autoint_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..f2f441201973e95440e09b122d7ad7db7ffdee32 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/downstream_results/autoint_bank_marketing_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "bank_marketing", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmptkq5m2a8", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmptkq5m2a8/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.220062255859375, + "lr": 3.942650891985977e-06, + "bank_marketing_val_accuracy": 0.9021468758583069, + "bank_marketing_val_precision": 0.8930810689926147, + "bank_marketing_val_recall": 0.9021468758583069, + "bank_marketing_val_f1-score": 0.8961839079856873, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9195704460144043, + "bank_marketing_val_score": 0.9021468758583069 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2171637862920761, + "lr": 3.942649982491275e-06, + "bank_marketing_test_accuracy": 0.903863251209259, + "bank_marketing_test_precision": 0.8950685858726501, + "bank_marketing_test_recall": 0.903863251209259, + "bank_marketing_test_f1-score": 0.8978362083435059, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9241321086883545, + "bank_marketing_test_score": 0.903863251209259 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.22556912899017334, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8996610045433044, + "bank_marketing_val_precision": 0.8909509778022766, + "bank_marketing_val_recall": 0.8996610045433044, + "bank_marketing_val_f1-score": 0.8941517472267151, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9172124862670898, + "bank_marketing_val_score": 0.8996610045433044 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22024722397327423, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8996447324752808, + "bank_marketing_test_precision": 0.8906821608543396, + "bank_marketing_test_recall": 0.8996447324752808, + "bank_marketing_test_f1-score": 0.8937769532203674, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9235501289367676, + "bank_marketing_test_score": 0.8996447324752808 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.2250065803527832, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8992090225219727, + "bank_marketing_val_precision": 0.8903083801269531, + "bank_marketing_val_recall": 0.8992090225219727, + "bank_marketing_val_f1-score": 0.8935655951499939, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9184932112693787, + "bank_marketing_val_score": 0.8992090225219727 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22188028693199158, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8992007374763489, + "bank_marketing_test_precision": 0.8890897631645203, + "bank_marketing_test_recall": 0.8992007374763489, + "bank_marketing_test_f1-score": 0.892283022403717, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9224686026573181, + "bank_marketing_test_score": 0.8992007374763489 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.2179114669561386, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9035028219223022, + "bank_marketing_val_precision": 0.8935216665267944, + "bank_marketing_val_recall": 0.9035028219223022, + "bank_marketing_val_f1-score": 0.8965232372283936, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9210816025733948, + "bank_marketing_val_score": 0.9035028219223022 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21467138826847076, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.9014209508895874, + "bank_marketing_test_precision": 0.8918351531028748, + "bank_marketing_test_recall": 0.9014209508895874, + "bank_marketing_test_f1-score": 0.8948167562484741, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9255882501602173, + "bank_marketing_test_score": 0.9014209508895874 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.21751318871974945, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9007909893989563, + "bank_marketing_val_precision": 0.8925493359565735, + "bank_marketing_val_recall": 0.9007909893989563, + "bank_marketing_val_f1-score": 0.8956104516983032, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9217963218688965, + "bank_marketing_val_score": 0.9007909893989563 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2203984260559082, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8977575302124023, + "bank_marketing_test_precision": 0.8888751864433289, + "bank_marketing_test_recall": 0.8977575302124023, + "bank_marketing_test_f1-score": 0.8920714855194092, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9226546883583069, + "bank_marketing_test_score": 0.8977575302124023 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.21887222230434417, + "std": 0.0029083560859151578, + "n": 5, + "ci95": 0.003610622116921415 + }, + "lr": { + "mean": 7.885299964982551e-07, + "std": 1.7632066744677695e-06, + "n": 5, + "ci95": 2.1889592702791558e-06 + }, + "bank_marketing_test_accuracy": { + "mean": 0.9003774404525757, + "std": 0.002346551256865035, + "n": 5, + "ci95": 0.00291316111791058 + }, + "bank_marketing_test_precision": { + "mean": 0.8911101698875428, + "std": 0.0025226495906345695, + "n": 5, + "ci95": 0.003131781025472971 + }, + "bank_marketing_test_recall": { + "mean": 0.9003774404525757, + "std": 0.002346551256865035, + "n": 5, + "ci95": 0.00291316111791058 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8941568851470947, + "std": 0.0023440056799846586, + "n": 5, + "ci95": 0.0029100008734586976 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9236787557601929, + "std": 0.0012630743423196343, + "n": 5, + "ci95": 0.001568062514002723 + }, + "bank_marketing_test_score": { + "mean": 0.9003774404525757, + "std": 0.002346551256865035, + "n": 5, + "ci95": 0.00291316111791058 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/downstream_results/dcnv2_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/downstream_results/dcnv2_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..65adfad40ab29bddbd072b2959ceb78abe60f157 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/downstream_results/dcnv2_bank_marketing_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "bank_marketing", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpopyq4rxj", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpopyq4rxj/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.2575227916240692, + "lr": 7.783600449329242e-05, + "bank_marketing_val_accuracy": 0.8962711691856384, + "bank_marketing_val_precision": 0.8910936713218689, + "bank_marketing_val_recall": 0.8962711691856384, + "bank_marketing_val_f1-score": 0.8933694362640381, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9018504023551941, + "bank_marketing_val_score": 0.8962711691856384 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.24581386148929596, + "lr": 7.783605542499572e-05, + "bank_marketing_test_accuracy": 0.897535502910614, + "bank_marketing_test_precision": 0.8910602927207947, + "bank_marketing_test_recall": 0.897535502910614, + "bank_marketing_test_f1-score": 0.8937128782272339, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9108147025108337, + "bank_marketing_test_score": 0.897535502910614 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.24404685199260712, + "lr": 0.00023208656057249755, + "bank_marketing_val_accuracy": 0.8964971899986267, + "bank_marketing_val_precision": 0.8917258977890015, + "bank_marketing_val_recall": 0.8964971899986267, + "bank_marketing_val_f1-score": 0.8938471674919128, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9033644795417786, + "bank_marketing_val_score": 0.8964971899986267 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23978230357170105, + "lr": 0.00023208656057249755, + "bank_marketing_test_accuracy": 0.8932060599327087, + "bank_marketing_test_precision": 0.8875333070755005, + "bank_marketing_test_recall": 0.8932060599327087, + "bank_marketing_test_f1-score": 0.8899931907653809, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9079089164733887, + "bank_marketing_test_score": 0.8932060599327087 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.23973488807678223, + "lr": 0.00015772644837852567, + "bank_marketing_val_accuracy": 0.8958191871643066, + "bank_marketing_val_precision": 0.8883044123649597, + "bank_marketing_val_recall": 0.8958191871643066, + "bank_marketing_val_f1-score": 0.8913601636886597, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9050081968307495, + "bank_marketing_val_score": 0.8958191871643066 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23744255304336548, + "lr": 0.00015772631741128862, + "bank_marketing_test_accuracy": 0.8959813714027405, + "bank_marketing_test_precision": 0.8874585032463074, + "bank_marketing_test_recall": 0.8959813714027405, + "bank_marketing_test_f1-score": 0.8906747102737427, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9102327823638916, + "bank_marketing_test_score": 0.8959813714027405 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.2536405920982361, + "lr": 4.758648356073536e-05, + "bank_marketing_val_accuracy": 0.8996610045433044, + "bank_marketing_val_precision": 0.8950634598731995, + "bank_marketing_val_recall": 0.8996610045433044, + "bank_marketing_val_f1-score": 0.8970919847488403, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9018797278404236, + "bank_marketing_val_score": 0.8996610045433044 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2399340569972992, + "lr": 4.758644718094729e-05, + "bank_marketing_test_accuracy": 0.8944271802902222, + "bank_marketing_test_precision": 0.8893968462944031, + "bank_marketing_test_recall": 0.8944271802902222, + "bank_marketing_test_f1-score": 0.8916105628013611, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9127699136734009, + "bank_marketing_test_score": 0.8944271802902222 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.23666277527809143, + "lr": 0.00020610730280168355, + "bank_marketing_val_accuracy": 0.9010169506072998, + "bank_marketing_val_precision": 0.8935287594795227, + "bank_marketing_val_recall": 0.9010169506072998, + "bank_marketing_val_f1-score": 0.8964226245880127, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9097002148628235, + "bank_marketing_val_score": 0.9010169506072998 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23276656866073608, + "lr": 0.00020610728824976832, + "bank_marketing_test_accuracy": 0.8965364098548889, + "bank_marketing_test_precision": 0.8885753154754639, + "bank_marketing_test_recall": 0.8965364098548889, + "bank_marketing_test_f1-score": 0.8916469812393188, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.913848876953125, + "bank_marketing_test_score": 0.8965364098548889 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.23914786875247956, + "std": 0.004719246528633006, + "n": 5, + "ci95": 0.005858779113742956 + }, + "lr": { + "mean": 0.0001442685337678995, + "std": 7.980891142895875e-05, + "n": 5, + "ci95": 9.907996552703651e-05 + }, + "bank_marketing_test_accuracy": { + "mean": 0.8955373048782349, + "std": 0.0017215749189417194, + "n": 5, + "ci95": 0.00213727490535679 + }, + "bank_marketing_test_precision": { + "mean": 0.8888048529624939, + "std": 0.0014932881978461379, + "n": 5, + "ci95": 0.0018538649445961524 + }, + "bank_marketing_test_recall": { + "mean": 0.8955373048782349, + "std": 0.0017215749189417194, + "n": 5, + "ci95": 0.00213727490535679 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8915276646614074, + "std": 0.0014034251138436069, + "n": 5, + "ci95": 0.0017423030763071775 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.911115038394928, + "std": 0.002310859178581375, + "n": 5, + "ci95": 0.002868850654046138 + }, + "bank_marketing_test_score": { + "mean": 0.8955373048782349, + "std": 0.0017215749189417194, + "n": 5, + "ci95": 0.00213727490535679 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/downstream_results/fttransformer_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/downstream_results/fttransformer_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..af78457ca20a12f7c8765edd49de1b1da9e10281 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/downstream_results/fttransformer_bank_marketing_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "bank_marketing", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp7t2nndiq", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp7t2nndiq/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.22926878929138184, + "lr": 2.447173756081611e-05, + "bank_marketing_val_accuracy": 0.8983050584793091, + "bank_marketing_val_precision": 0.8869761228561401, + "bank_marketing_val_recall": 0.8983050584793091, + "bank_marketing_val_f1-score": 0.8905303478240967, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9073823094367981, + "bank_marketing_val_score": 0.8983050584793091 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22579815983772278, + "lr": 2.447175393172074e-05, + "bank_marketing_test_accuracy": 0.9009768962860107, + "bank_marketing_test_precision": 0.8899774551391602, + "bank_marketing_test_recall": 0.9009768962860107, + "bank_marketing_test_f1-score": 0.8928441405296326, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9136890769004822, + "bank_marketing_test_score": 0.9009768962860107 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.224252849817276, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8998870253562927, + "bank_marketing_val_precision": 0.8924860954284668, + "bank_marketing_val_recall": 0.8998870253562927, + "bank_marketing_val_f1-score": 0.8953960537910461, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9106707572937012, + "bank_marketing_val_score": 0.8998870253562927 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2207651287317276, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8993117213249207, + "bank_marketing_test_precision": 0.8901563286781311, + "bank_marketing_test_recall": 0.8993117213249207, + "bank_marketing_test_f1-score": 0.8932924270629883, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9177462458610535, + "bank_marketing_test_score": 0.8993117213249207 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.22439514100551605, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8976271152496338, + "bank_marketing_val_precision": 0.887643039226532, + "bank_marketing_val_recall": 0.8976271152496338, + "bank_marketing_val_f1-score": 0.8911600112915039, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9112979173660278, + "bank_marketing_val_score": 0.8976271152496338 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2207329124212265, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.9006438851356506, + "bank_marketing_test_precision": 0.8900318145751953, + "bank_marketing_test_recall": 0.9006438851356506, + "bank_marketing_test_f1-score": 0.8930239677429199, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9178270697593689, + "bank_marketing_test_score": 0.9006438851356506 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.22807899117469788, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9007909893989563, + "bank_marketing_val_precision": 0.8909189701080322, + "bank_marketing_val_recall": 0.9007909893989563, + "bank_marketing_val_f1-score": 0.8941872119903564, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9068520069122314, + "bank_marketing_val_score": 0.9007909893989563 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22919899225234985, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8958703279495239, + "bank_marketing_test_precision": 0.8839811682701111, + "bank_marketing_test_recall": 0.8958703279495239, + "bank_marketing_test_f1-score": 0.8874983191490173, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9098312258720398, + "bank_marketing_test_score": 0.8958703279495239 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.22654274106025696, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8980790972709656, + "bank_marketing_val_precision": 0.8861681222915649, + "bank_marketing_val_recall": 0.8980790972709656, + "bank_marketing_val_f1-score": 0.8897360563278198, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9096811413764954, + "bank_marketing_val_score": 0.8980790972709656 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22256548702716827, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8973134756088257, + "bank_marketing_test_precision": 0.8845266699790955, + "bank_marketing_test_recall": 0.8973134756088257, + "bank_marketing_test_f1-score": 0.8876304626464844, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9163182377815247, + "bank_marketing_test_score": 0.8973134756088257 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.223812136054039, + "std": 0.003649743681751011, + "n": 5, + "ci95": 0.004531028824923887 + }, + "lr": { + "mean": 4.894350786344148e-06, + "std": 1.0944101063995064e-05, + "n": 5, + "ci95": 1.3586717782891353e-05 + }, + "bank_marketing_test_accuracy": { + "mean": 0.8988232612609863, + "std": 0.002190358201612396, + "n": 5, + "ci95": 0.002719252915769902 + }, + "bank_marketing_test_precision": { + "mean": 0.8877346873283386, + "std": 0.003184000176745278, + "n": 5, + "ci95": 0.0039528245919105654 + }, + "bank_marketing_test_recall": { + "mean": 0.8988232612609863, + "std": 0.002190358201612396, + "n": 5, + "ci95": 0.002719252915769902 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8908578634262085, + "std": 0.0030111068682199225, + "n": 5, + "ci95": 0.003738183610824189 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9150823712348938, + "std": 0.003379081267865609, + "n": 5, + "ci95": 0.0041950109272095295 + }, + "bank_marketing_test_score": { + "mean": 0.8988232612609863, + "std": 0.002190358201612396, + "n": 5, + "ci95": 0.002719252915769902 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/downstream_results/mlp_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/downstream_results/mlp_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..74ac4caaba72892e2b13d8328f157cfa6c984694 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/downstream_results/mlp_bank_marketing_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "bank_marketing", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpby14sayo", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpby14sayo/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.2332422137260437, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8946892619132996, + "bank_marketing_val_precision": 0.8869893550872803, + "bank_marketing_val_recall": 0.8946892619132996, + "bank_marketing_val_f1-score": 0.8901278376579285, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9031198620796204, + "bank_marketing_val_score": 0.8946892619132996 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22771266102790833, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8986456394195557, + "bank_marketing_test_precision": 0.8907520771026611, + "bank_marketing_test_recall": 0.8986456394195557, + "bank_marketing_test_f1-score": 0.8937296271324158, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9125028848648071, + "bank_marketing_test_score": 0.8986456394195557 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.23242446780204773, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8994350433349609, + "bank_marketing_val_precision": 0.8859468698501587, + "bank_marketing_val_recall": 0.8994350433349609, + "bank_marketing_val_f1-score": 0.8888997435569763, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9030782580375671, + "bank_marketing_val_score": 0.8994350433349609 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23005063831806183, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8993117213249207, + "bank_marketing_test_precision": 0.885762095451355, + "bank_marketing_test_recall": 0.8993117213249207, + "bank_marketing_test_f1-score": 0.8877106308937073, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9097352027893066, + "bank_marketing_test_score": 0.8993117213249207 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.23707500100135803, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8933333158493042, + "bank_marketing_val_precision": 0.8742707371711731, + "bank_marketing_val_recall": 0.8933333158493042, + "bank_marketing_val_f1-score": 0.8760882019996643, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.8975174427032471, + "bank_marketing_val_score": 0.8933333158493042 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23643742501735687, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8959813714027405, + "bank_marketing_test_precision": 0.8795642852783203, + "bank_marketing_test_recall": 0.8959813714027405, + "bank_marketing_test_f1-score": 0.8765068054199219, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.902720034122467, + "bank_marketing_test_score": 0.8959813714027405 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.23853863775730133, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8964971899986267, + "bank_marketing_val_precision": 0.8794101476669312, + "bank_marketing_val_recall": 0.8964971899986267, + "bank_marketing_val_f1-score": 0.8807644844055176, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.898530125617981, + "bank_marketing_val_score": 0.8964971899986267 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23879991471767426, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8955373167991638, + "bank_marketing_test_precision": 0.8788213729858398, + "bank_marketing_test_recall": 0.8955373167991638, + "bank_marketing_test_f1-score": 0.8779602646827698, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9033033847808838, + "bank_marketing_test_score": 0.8955373167991638 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.23447182774543762, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8967231512069702, + "bank_marketing_val_precision": 0.8821591138839722, + "bank_marketing_val_recall": 0.8967231512069702, + "bank_marketing_val_f1-score": 0.8854878544807434, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9018567204475403, + "bank_marketing_val_score": 0.8967231512069702 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23027409613132477, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8979795575141907, + "bank_marketing_test_precision": 0.8839319944381714, + "bank_marketing_test_recall": 0.8979795575141907, + "bank_marketing_test_f1-score": 0.8860899209976196, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9101659059524536, + "bank_marketing_test_score": 0.8979795575141907 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.2326549470424652, + "std": 0.004715528427511359, + "n": 5, + "ci95": 0.005854163221552937 + }, + "lr": { + "mean": 0.0009999994654208422, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_accuracy": { + "mean": 0.8974911212921143, + "std": 0.0016570126171984334, + "n": 5, + "ci95": 0.0020571230712252723 + }, + "bank_marketing_test_precision": { + "mean": 0.8837663650512695, + "std": 0.004871369590114778, + "n": 5, + "ci95": 0.006047634561306575 + }, + "bank_marketing_test_recall": { + "mean": 0.8974911212921143, + "std": 0.0016570126171984334, + "n": 5, + "ci95": 0.0020571230712252723 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8843994498252868, + "std": 0.007152460458540983, + "n": 5, + "ci95": 0.008879528901939045 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9076854825019837, + "std": 0.004399422613314011, + "n": 5, + "ci95": 0.005461728935546567 + }, + "bank_marketing_test_score": { + "mean": 0.8974911212921143, + "std": 0.0016570126171984334, + "n": 5, + "ci95": 0.0020571230712252723 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/downstream_results/resnet_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/downstream_results/resnet_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..cd0eb8d7a2d0c296ec525af6d4da36245d1e1e3c --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/downstream_results/resnet_bank_marketing_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "bank_marketing", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp0_ncwm2i", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp0_ncwm2i/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.22039537131786346, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8996610045433044, + "bank_marketing_val_precision": 0.889180600643158, + "bank_marketing_val_recall": 0.8996610045433044, + "bank_marketing_val_f1-score": 0.8925786018371582, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9192377924919128, + "bank_marketing_val_score": 0.8996610045433044 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21513909101486206, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.9004218578338623, + "bank_marketing_test_precision": 0.8904873132705688, + "bank_marketing_test_recall": 0.9004218578338623, + "bank_marketing_test_f1-score": 0.8935607075691223, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9258755445480347, + "bank_marketing_test_score": 0.9004218578338623 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.2140064686536789, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9012429118156433, + "bank_marketing_val_precision": 0.8933160305023193, + "bank_marketing_val_recall": 0.9012429118156433, + "bank_marketing_val_f1-score": 0.8962962031364441, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9229562282562256, + "bank_marketing_val_score": 0.9012429118156433 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21069926023483276, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8996447324752808, + "bank_marketing_test_precision": 0.891583263874054, + "bank_marketing_test_recall": 0.8996447324752808, + "bank_marketing_test_f1-score": 0.8945503830909729, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9281417727470398, + "bank_marketing_test_score": 0.8996447324752808 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.21945235133171082, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9003390073776245, + "bank_marketing_val_precision": 0.8876591324806213, + "bank_marketing_val_recall": 0.9003390073776245, + "bank_marketing_val_f1-score": 0.8906826972961426, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9197818040847778, + "bank_marketing_val_score": 0.9003390073776245 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21502326428890228, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8997557759284973, + "bank_marketing_test_precision": 0.8876996040344238, + "bank_marketing_test_recall": 0.8997557759284973, + "bank_marketing_test_f1-score": 0.8904853463172913, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.925521194934845, + "bank_marketing_test_score": 0.8997557759284973 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.21634548902511597, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9003390073776245, + "bank_marketing_val_precision": 0.8932591676712036, + "bank_marketing_val_recall": 0.9003390073776245, + "bank_marketing_val_f1-score": 0.8960733413696289, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9209128022193909, + "bank_marketing_val_score": 0.9003390073776245 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.20960387587547302, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.9018650054931641, + "bank_marketing_test_precision": 0.8962315917015076, + "bank_marketing_test_recall": 0.9018650054931641, + "bank_marketing_test_f1-score": 0.8985499739646912, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9284915328025818, + "bank_marketing_test_score": 0.9018650054931641 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.21936513483524323, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9019209146499634, + "bank_marketing_val_precision": 0.8937519192695618, + "bank_marketing_val_recall": 0.9019209146499634, + "bank_marketing_val_f1-score": 0.8967469334602356, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.917718768119812, + "bank_marketing_val_score": 0.9019209146499634 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21318331360816956, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8985346555709839, + "bank_marketing_test_precision": 0.8904697299003601, + "bank_marketing_test_recall": 0.8985346555709839, + "bank_marketing_test_f1-score": 0.8934861421585083, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9253090620040894, + "bank_marketing_test_score": 0.8985346555709839 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.21272976100444793, + "std": 0.0025082279399100665, + "n": 5, + "ci95": 0.003113877051705598 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_accuracy": { + "mean": 0.9000444054603577, + "std": 0.0012226451860459441, + "n": 5, + "ci95": 0.0015178711338903648 + }, + "bank_marketing_test_precision": { + "mean": 0.8912943005561829, + "std": 0.0031110019279128533, + "n": 5, + "ci95": 0.003862199825222841 + }, + "bank_marketing_test_recall": { + "mean": 0.9000444054603577, + "std": 0.0012226451860459441, + "n": 5, + "ci95": 0.0015178711338903648 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8941265106201172, + "std": 0.0029040910011418016, + "n": 5, + "ci95": 0.00360532716370444 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9266678214073181, + "std": 0.0015237402838145294, + "n": 5, + "ci95": 0.0018916701417095137 + }, + "bank_marketing_test_score": { + "mean": 0.9000444054603577, + "std": 0.0012226451860459441, + "n": 5, + "ci95": 0.0015178711338903648 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..18df7b502b9bd183bdcf7bb8c48b596985685541 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6134aea7e7ece0977a47ad67313adc20e184ecad60523f73ac5d9d2061ad4dd7 +size 185184384 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..9c7a75a120d614f1663c4eedabdd678e5c0d8412 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dcee140dc90f9f7d54743e1bc9561d964e329f403ff43440a5536ce85ccac8cd +size 15464302 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..7295cf636173f8c4666f6d93e43b40bd78436b39 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.002_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/bank_marketing/20260916_020700/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: bank_marketing +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.001 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.3 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.1 +mask_min_trgt_share: 0.1 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 256 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.1 +recon_y_weight: 1.0 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.0005 +sw_s_y_weight: 5.0e-05 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 15 +vtjepa_kl_x_end_weight: 0.0001 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 15 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 15 +wjepa_joint_end_weight: 0.002 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..7540b8a42207602340397b66dba86308a63eb5ad --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-16T04:44:37.024342+00:00", + "spec": { + "dataset": "bank_marketing", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.0005, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0001, + "lambda_joint": 0.004, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..5bb97d13f9f1c46d61870ea4e97668ed7ffc9513 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "bank_marketing", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.0005, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0001, + "lambda_joint": 0.004, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "bank_marketing", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.004", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.0005", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "0.0001", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..a55d2769d0d680b747e11df9e5e324f5ba354c5b --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "bank_marketing", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.004, + "s_x": 0.0005, + "z": 0.0, + "s_y": 0.0001 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 15 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "bank_marketing", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.004", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.0005", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "0.0001", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/downstream_results/autoint_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/downstream_results/autoint_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..273f11e49cf933811a29f082b8d509f2711df222 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/downstream_results/autoint_bank_marketing_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "bank_marketing", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp85ae9lh2", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp85ae9lh2/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.22496601939201355, + "lr": 3.942650891985977e-06, + "bank_marketing_val_accuracy": 0.8974011540412903, + "bank_marketing_val_precision": 0.8899657130241394, + "bank_marketing_val_recall": 0.8974011540412903, + "bank_marketing_val_f1-score": 0.8929571509361267, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9186492562294006, + "bank_marketing_val_score": 0.8974011540412903 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2213181108236313, + "lr": 3.942649982491275e-06, + "bank_marketing_test_accuracy": 0.8996447324752808, + "bank_marketing_test_precision": 0.8914605379104614, + "bank_marketing_test_recall": 0.8996447324752808, + "bank_marketing_test_f1-score": 0.8944488167762756, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9236972332000732, + "bank_marketing_test_score": 0.8996447324752808 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.22405768930912018, + "lr": 9.54915230977349e-05, + "bank_marketing_val_accuracy": 0.8996610045433044, + "bank_marketing_val_precision": 0.8925860524177551, + "bank_marketing_val_recall": 0.8996610045433044, + "bank_marketing_val_f1-score": 0.8954176902770996, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9172192811965942, + "bank_marketing_val_score": 0.8996610045433044 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21699170768260956, + "lr": 9.549142851028591e-05, + "bank_marketing_test_accuracy": 0.8994227647781372, + "bank_marketing_test_precision": 0.8906071782112122, + "bank_marketing_test_recall": 0.8994227647781372, + "bank_marketing_test_f1-score": 0.8936991095542908, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9239392876625061, + "bank_marketing_test_score": 0.8994227647781372 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.2253960818052292, + "lr": 9.866358823273913e-07, + "bank_marketing_val_accuracy": 0.897175133228302, + "bank_marketing_val_precision": 0.8912844657897949, + "bank_marketing_val_recall": 0.897175133228302, + "bank_marketing_val_f1-score": 0.8938024044036865, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9182994961738586, + "bank_marketing_val_score": 0.897175133228302 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22066986560821533, + "lr": 9.86635200206365e-07, + "bank_marketing_test_accuracy": 0.9020870327949524, + "bank_marketing_test_precision": 0.8946049213409424, + "bank_marketing_test_recall": 0.9020870327949524, + "bank_marketing_test_f1-score": 0.8973624110221863, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9236867427825928, + "bank_marketing_test_score": 0.9020870327949524 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.23002375662326813, + "lr": 8.856373824528418e-06, + "bank_marketing_val_accuracy": 0.8985310792922974, + "bank_marketing_val_precision": 0.8884251117706299, + "bank_marketing_val_recall": 0.8985310792922974, + "bank_marketing_val_f1-score": 0.8918921947479248, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9170133471488953, + "bank_marketing_val_score": 0.8985310792922974 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2247307151556015, + "lr": 8.856372005539015e-06, + "bank_marketing_test_accuracy": 0.900865912437439, + "bank_marketing_test_precision": 0.8912705779075623, + "bank_marketing_test_recall": 0.900865912437439, + "bank_marketing_test_f1-score": 0.8943050503730774, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9225965738296509, + "bank_marketing_test_score": 0.900865912437439 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.22508326172828674, + "lr": 2.447173756081611e-05, + "bank_marketing_val_accuracy": 0.8987570405006409, + "bank_marketing_val_precision": 0.8896638751029968, + "bank_marketing_val_recall": 0.8987570405006409, + "bank_marketing_val_f1-score": 0.8929779529571533, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9187059998512268, + "bank_marketing_val_score": 0.8987570405006409 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2235657125711441, + "lr": 2.447175393172074e-05, + "bank_marketing_test_accuracy": 0.8985346555709839, + "bank_marketing_test_precision": 0.8893592357635498, + "bank_marketing_test_recall": 0.8985346555709839, + "bank_marketing_test_f1-score": 0.8925487995147705, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9223669171333313, + "bank_marketing_test_score": 0.8985346555709839 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.22145522236824036, + "std": 0.0029892126915640966, + "n": 5, + "ci95": 0.003711002758091558 + }, + "lr": { + "mean": 2.6749767926048662e-05, + "std": 3.948000285570419e-05, + "n": 5, + "ci95": 4.901303942019586e-05 + }, + "bank_marketing_test_accuracy": { + "mean": 0.9001110196113586, + "std": 0.0013829798694408792, + "n": 5, + "ci95": 0.0017169210221688088 + }, + "bank_marketing_test_precision": { + "mean": 0.8914604902267456, + "std": 0.0019408068836268943, + "n": 5, + "ci95": 0.0024094437034836366 + }, + "bank_marketing_test_recall": { + "mean": 0.9001110196113586, + "std": 0.0013829798694408792, + "n": 5, + "ci95": 0.0017169210221688088 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8944728374481201, + "std": 0.0017803750839237907, + "n": 5, + "ci95": 0.0022102732487133915 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9232573509216309, + "std": 0.0007197934086136407, + "n": 5, + "ci95": 0.0008935982816343759 + }, + "bank_marketing_test_score": { + "mean": 0.9001110196113586, + "std": 0.0013829798694408792, + "n": 5, + "ci95": 0.0017169210221688088 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/downstream_results/dcnv2_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/downstream_results/dcnv2_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..cab5e94f5c265441d3eeb4163082866599c9ae7f --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/downstream_results/dcnv2_bank_marketing_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "bank_marketing", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmppqthg7ol", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmppqthg7ol/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.24375221133232117, + "lr": 0.000315937795676291, + "bank_marketing_val_accuracy": 0.8994350433349609, + "bank_marketing_val_precision": 0.8942989706993103, + "bank_marketing_val_recall": 0.8994350433349609, + "bank_marketing_val_f1-score": 0.8965256214141846, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9070912599563599, + "bank_marketing_val_score": 0.8994350433349609 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23108889162540436, + "lr": 0.0003159379994031042, + "bank_marketing_test_accuracy": 0.8956483006477356, + "bank_marketing_test_precision": 0.8892665505409241, + "bank_marketing_test_recall": 0.8956483006477356, + "bank_marketing_test_f1-score": 0.8919277787208557, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9163992404937744, + "bank_marketing_test_score": 0.8956483006477356 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.25239297747612, + "lr": 0.00011474337225081399, + "bank_marketing_val_accuracy": 0.8996610045433044, + "bank_marketing_val_precision": 0.89551842212677, + "bank_marketing_val_recall": 0.8996610045433044, + "bank_marketing_val_f1-score": 0.8973739147186279, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.90987628698349, + "bank_marketing_val_score": 0.8996610045433044 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.24037134647369385, + "lr": 0.0001147434813901782, + "bank_marketing_test_accuracy": 0.8938721418380737, + "bank_marketing_test_precision": 0.8885444402694702, + "bank_marketing_test_recall": 0.8938721418380737, + "bank_marketing_test_f1-score": 0.8908728361129761, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9145432710647583, + "bank_marketing_test_score": 0.8938721418380737 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.2437855303287506, + "lr": 0.00023208656057249755, + "bank_marketing_val_accuracy": 0.8953672051429749, + "bank_marketing_val_precision": 0.8869123458862305, + "bank_marketing_val_recall": 0.8953672051429749, + "bank_marketing_val_f1-score": 0.8902366757392883, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.909464418888092, + "bank_marketing_val_score": 0.8953672051429749 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23420105874538422, + "lr": 0.00023208656057249755, + "bank_marketing_test_accuracy": 0.896869421005249, + "bank_marketing_test_precision": 0.8888448476791382, + "bank_marketing_test_recall": 0.896869421005249, + "bank_marketing_test_f1-score": 0.8919187784194946, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.914148211479187, + "bank_marketing_test_score": 0.896869421005249 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.2561049163341522, + "lr": 0.00015772644837852567, + "bank_marketing_val_accuracy": 0.8955932259559631, + "bank_marketing_val_precision": 0.8878315091133118, + "bank_marketing_val_recall": 0.8955932259559631, + "bank_marketing_val_f1-score": 0.8909635543823242, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9041873216629028, + "bank_marketing_val_score": 0.8955932259559631 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.24152587354183197, + "lr": 0.00015772631741128862, + "bank_marketing_test_accuracy": 0.8964254260063171, + "bank_marketing_test_precision": 0.8889591693878174, + "bank_marketing_test_recall": 0.8964254260063171, + "bank_marketing_test_f1-score": 0.8919134140014648, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9111571907997131, + "bank_marketing_test_score": 0.8964254260063171 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.25932633876800537, + "lr": 0.00011474337225081399, + "bank_marketing_val_accuracy": 0.897175133228302, + "bank_marketing_val_precision": 0.8897908329963684, + "bank_marketing_val_recall": 0.897175133228302, + "bank_marketing_val_f1-score": 0.8927741050720215, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9013753533363342, + "bank_marketing_val_score": 0.897175133228302 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.24880704283714294, + "lr": 0.0001147434813901782, + "bank_marketing_test_accuracy": 0.8979795575141907, + "bank_marketing_test_precision": 0.890141487121582, + "bank_marketing_test_recall": 0.8979795575141907, + "bank_marketing_test_f1-score": 0.8931329846382141, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9103535413742065, + "bank_marketing_test_score": 0.8979795575141907 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.23919884264469146, + "std": 0.006889421209414242, + "n": 5, + "ci95": 0.008552974896012852 + }, + "lr": { + "mean": 0.00018704756803344935, + "std": 8.654009443066119e-05, + "n": 5, + "ci95": 0.00010743649323583143 + }, + "bank_marketing_test_accuracy": { + "mean": 0.8961589694023132, + "std": 0.0015313918963240796, + "n": 5, + "ci95": 0.0019011693503830627 + }, + "bank_marketing_test_precision": { + "mean": 0.8891512989997864, + "std": 0.0006109130774020028, + "n": 5, + "ci95": 0.000758427167658913 + }, + "bank_marketing_test_recall": { + "mean": 0.8961589694023132, + "std": 0.0015313918963240796, + "n": 5, + "ci95": 0.0019011693503830627 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8919531583786011, + "std": 0.0008003892685136092, + "n": 5, + "ci95": 0.0009936552160986294 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9133202910423279, + "std": 0.002507073411494659, + "n": 5, + "ci95": 0.0031124437451543565 + }, + "bank_marketing_test_score": { + "mean": 0.8961589694023132, + "std": 0.0015313918963240796, + "n": 5, + "ci95": 0.0019011693503830627 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/downstream_results/fttransformer_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/downstream_results/fttransformer_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..eaf53e4f200f24123536492933b8035604cfde3f --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/downstream_results/fttransformer_bank_marketing_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "bank_marketing", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpr8dtfu7i", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpr8dtfu7i/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.22502751648426056, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8976271152496338, + "bank_marketing_val_precision": 0.8868613839149475, + "bank_marketing_val_recall": 0.8976271152496338, + "bank_marketing_val_f1-score": 0.8904603719711304, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9110836982727051, + "bank_marketing_val_score": 0.8976271152496338 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22315052151679993, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8983126282691956, + "bank_marketing_test_precision": 0.8880540132522583, + "bank_marketing_test_recall": 0.8983126282691956, + "bank_marketing_test_f1-score": 0.8913339972496033, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9161393642425537, + "bank_marketing_test_score": 0.8983126282691956 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.22669024765491486, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.902824878692627, + "bank_marketing_val_precision": 0.8941544890403748, + "bank_marketing_val_recall": 0.902824878692627, + "bank_marketing_val_f1-score": 0.8971715569496155, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9106824994087219, + "bank_marketing_val_score": 0.902824878692627 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.220283642411232, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8990896940231323, + "bank_marketing_test_precision": 0.8892680406570435, + "bank_marketing_test_recall": 0.8990896940231323, + "bank_marketing_test_f1-score": 0.8924658298492432, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9190824031829834, + "bank_marketing_test_score": 0.8990896940231323 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.22623814642429352, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8967231512069702, + "bank_marketing_val_precision": 0.8872887492179871, + "bank_marketing_val_recall": 0.8967231512069702, + "bank_marketing_val_f1-score": 0.8907715082168579, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9136490821838379, + "bank_marketing_val_score": 0.8967231512069702 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2187533974647522, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.9010879397392273, + "bank_marketing_test_precision": 0.8913168907165527, + "bank_marketing_test_recall": 0.9010879397392273, + "bank_marketing_test_f1-score": 0.8943268060684204, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9211629033088684, + "bank_marketing_test_score": 0.9010879397392273 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.23159025609493256, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8980790972709656, + "bank_marketing_val_precision": 0.8876446485519409, + "bank_marketing_val_recall": 0.8980790972709656, + "bank_marketing_val_f1-score": 0.8911784887313843, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9058002233505249, + "bank_marketing_val_score": 0.8980790972709656 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22212037444114685, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.9011989235877991, + "bank_marketing_test_precision": 0.8901998400688171, + "bank_marketing_test_recall": 0.9011989235877991, + "bank_marketing_test_f1-score": 0.893027126789093, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9169424176216125, + "bank_marketing_test_score": 0.9011989235877991 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.22729137539863586, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9003390073776245, + "bank_marketing_val_precision": 0.8901535272598267, + "bank_marketing_val_recall": 0.9003390073776245, + "bank_marketing_val_f1-score": 0.8934770226478577, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9077252745628357, + "bank_marketing_val_score": 0.9003390073776245 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22731122374534607, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8987566828727722, + "bank_marketing_test_precision": 0.886952817440033, + "bank_marketing_test_recall": 0.8987566828727722, + "bank_marketing_test_f1-score": 0.8900277614593506, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9115653038024902, + "bank_marketing_test_score": 0.8987566828727722 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.22232383191585542, + "std": 0.0032599564140335913, + "n": 5, + "ci95": 0.0040471215975624785 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_accuracy": { + "mean": 0.8996891736984253, + "std": 0.0013564368309208915, + "n": 5, + "ci95": 0.001683968770415768 + }, + "bank_marketing_test_precision": { + "mean": 0.8891583204269409, + "std": 0.0017204625213745575, + "n": 5, + "ci95": 0.002135893902776585 + }, + "bank_marketing_test_recall": { + "mean": 0.8996891736984253, + "std": 0.0013564368309208915, + "n": 5, + "ci95": 0.001683968770415768 + }, + "bank_marketing_test_f1-score": { + "mean": 0.892236304283142, + "std": 0.001638604537629423, + "n": 5, + "ci95": 0.002034270085807222 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9169784784317017, + "std": 0.0036036017913505594, + "n": 5, + "ci95": 0.004473745285675284 + }, + "bank_marketing_test_score": { + "mean": 0.8996891736984253, + "std": 0.0013564368309208915, + "n": 5, + "ci95": 0.001683968770415768 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/downstream_results/mlp_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/downstream_results/mlp_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..ba367767ddc070a98a3e8753d5fee83629f6dacc --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/downstream_results/mlp_bank_marketing_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "bank_marketing", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpm4duq74g", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpm4duq74g/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.22812308371067047, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.9007909893989563, + "bank_marketing_val_precision": 0.8895459175109863, + "bank_marketing_val_recall": 0.9007909893989563, + "bank_marketing_val_f1-score": 0.8927901387214661, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9083749055862427, + "bank_marketing_val_score": 0.9007909893989563 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22574743628501892, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8973134756088257, + "bank_marketing_test_precision": 0.8847188353538513, + "bank_marketing_test_recall": 0.8973134756088257, + "bank_marketing_test_f1-score": 0.8878791332244873, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9138630032539368, + "bank_marketing_test_score": 0.8973134756088257 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.22816883027553558, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8958191871643066, + "bank_marketing_val_precision": 0.8814221024513245, + "bank_marketing_val_recall": 0.8958191871643066, + "bank_marketing_val_f1-score": 0.8850435614585876, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9074811339378357, + "bank_marketing_val_score": 0.8958191871643066 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22445723414421082, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.9019760489463806, + "bank_marketing_test_precision": 0.8897424936294556, + "bank_marketing_test_recall": 0.9019760489463806, + "bank_marketing_test_f1-score": 0.8916972279548645, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9152436852455139, + "bank_marketing_test_score": 0.9019760489463806 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.22417841851711273, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8962711691856384, + "bank_marketing_val_precision": 0.8830186724662781, + "bank_marketing_val_recall": 0.8962711691856384, + "bank_marketing_val_f1-score": 0.8867511749267578, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9101238250732422, + "bank_marketing_val_score": 0.8962711691856384 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22482462227344513, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8949822187423706, + "bank_marketing_test_precision": 0.8802840709686279, + "bank_marketing_test_recall": 0.8949822187423706, + "bank_marketing_test_f1-score": 0.8832611441612244, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9133269786834717, + "bank_marketing_test_score": 0.8949822187423706 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.22408631443977356, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8987570405006409, + "bank_marketing_val_precision": 0.8897973895072937, + "bank_marketing_val_recall": 0.8987570405006409, + "bank_marketing_val_f1-score": 0.8930883407592773, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9114716053009033, + "bank_marketing_val_score": 0.8987570405006409 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21894444525241852, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.9011989235877991, + "bank_marketing_test_precision": 0.8918437957763672, + "bank_marketing_test_recall": 0.9011989235877991, + "bank_marketing_test_f1-score": 0.8948464393615723, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.918743908405304, + "bank_marketing_test_score": 0.9011989235877991 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.22795964777469635, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8978531360626221, + "bank_marketing_val_precision": 0.8837701082229614, + "bank_marketing_val_recall": 0.8978531360626221, + "bank_marketing_val_f1-score": 0.8869472146034241, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9069811701774597, + "bank_marketing_val_score": 0.8978531360626221 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.224864661693573, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8972024917602539, + "bank_marketing_test_precision": 0.8826704025268555, + "bank_marketing_test_recall": 0.8972024917602539, + "bank_marketing_test_f1-score": 0.8847751021385193, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9141396880149841, + "bank_marketing_test_score": 0.8972024917602539 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.2237676799297333, + "std": 0.0027376545208251373, + "n": 5, + "ci95": 0.003398702108469909 + }, + "lr": { + "mean": 0.0009999994654208422, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_accuracy": { + "mean": 0.898534631729126, + "std": 0.002950731297775602, + "n": 5, + "ci95": 0.0036632294568181754 + }, + "bank_marketing_test_precision": { + "mean": 0.8858519196510315, + "std": 0.004833373965695264, + "n": 5, + "ci95": 0.006000464325674246 + }, + "bank_marketing_test_recall": { + "mean": 0.898534631729126, + "std": 0.002950731297775602, + "n": 5, + "ci95": 0.0036632294568181754 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8884918093681335, + "std": 0.00480117625858762, + "n": 5, + "ci95": 0.005960492001116047 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.915063452720642, + "std": 0.0021730163393932894, + "n": 5, + "ci95": 0.0026977236018113577 + }, + "bank_marketing_test_score": { + "mean": 0.898534631729126, + "std": 0.002950731297775602, + "n": 5, + "ci95": 0.0036632294568181754 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/downstream_results/resnet_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/downstream_results/resnet_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..0d96926652182777c2f5d6756b46bbed53ffd918 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/downstream_results/resnet_bank_marketing_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "bank_marketing", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpwsgif2yb", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpwsgif2yb/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.216755211353302, + "lr": 0.0004999999655410647, + "bank_marketing_val_accuracy": 0.9001129865646362, + "bank_marketing_val_precision": 0.8914607167243958, + "bank_marketing_val_recall": 0.9001129865646362, + "bank_marketing_val_f1-score": 0.8946285247802734, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9197025299072266, + "bank_marketing_val_score": 0.9001129865646362 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21670807898044586, + "lr": 0.0004999997327104211, + "bank_marketing_test_accuracy": 0.8973134756088257, + "bank_marketing_test_precision": 0.8890118598937988, + "bank_marketing_test_recall": 0.8973134756088257, + "bank_marketing_test_f1-score": 0.8921266794204712, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9227942228317261, + "bank_marketing_test_score": 0.8973134756088257 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.2195689082145691, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9007909893989563, + "bank_marketing_val_precision": 0.8937546610832214, + "bank_marketing_val_recall": 0.9007909893989563, + "bank_marketing_val_f1-score": 0.8965446949005127, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9194051027297974, + "bank_marketing_val_score": 0.9007909893989563 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21504178643226624, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8989787101745605, + "bank_marketing_test_precision": 0.8920576572418213, + "bank_marketing_test_recall": 0.8989787101745605, + "bank_marketing_test_f1-score": 0.8947988152503967, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9253376126289368, + "bank_marketing_test_score": 0.8989787101745605 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.21602363884449005, + "lr": 0.0004999999655410647, + "bank_marketing_val_accuracy": 0.9035028219223022, + "bank_marketing_val_precision": 0.8908135890960693, + "bank_marketing_val_recall": 0.9035028219223022, + "bank_marketing_val_f1-score": 0.8927425742149353, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9220688343048096, + "bank_marketing_val_score": 0.9035028219223022 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21810175478458405, + "lr": 0.0004999997327104211, + "bank_marketing_test_accuracy": 0.9000887870788574, + "bank_marketing_test_precision": 0.8873394727706909, + "bank_marketing_test_recall": 0.9000887870788574, + "bank_marketing_test_f1-score": 0.8896437883377075, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9236680269241333, + "bank_marketing_test_score": 0.9000887870788574 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.2160715013742447, + "lr": 0.00020610730280168355, + "bank_marketing_val_accuracy": 0.9030508399009705, + "bank_marketing_val_precision": 0.8930979371070862, + "bank_marketing_val_recall": 0.9030508399009705, + "bank_marketing_val_f1-score": 0.8961514830589294, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9199990034103394, + "bank_marketing_val_score": 0.9030508399009705 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2124871462583542, + "lr": 0.00020610728824976832, + "bank_marketing_test_accuracy": 0.8985346555709839, + "bank_marketing_test_precision": 0.8873376250267029, + "bank_marketing_test_recall": 0.8985346555709839, + "bank_marketing_test_f1-score": 0.8905511498451233, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9265192151069641, + "bank_marketing_test_score": 0.8985346555709839 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.21505101025104523, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9025988578796387, + "bank_marketing_val_precision": 0.8944501280784607, + "bank_marketing_val_recall": 0.9025988578796387, + "bank_marketing_val_f1-score": 0.8974084258079529, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9206488728523254, + "bank_marketing_val_score": 0.9025988578796387 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.20904773473739624, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8998667597770691, + "bank_marketing_test_precision": 0.8916483521461487, + "bank_marketing_test_recall": 0.8998667597770691, + "bank_marketing_test_f1-score": 0.8946314454078674, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.928045928478241, + "bank_marketing_test_score": 0.8998667597770691 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.2142773002386093, + "std": 0.00359418223003455, + "n": 5, + "ci95": 0.004462051230540842 + }, + "lr": { + "mean": 0.0002412213507341221, + "std": 0.0002507693060567254, + "n": 5, + "ci95": 0.0003113213017753773 + }, + "bank_marketing_test_accuracy": { + "mean": 0.8989564776420593, + "std": 0.0011167611710139317, + "n": 5, + "ci95": 0.0013864198414043813 + }, + "bank_marketing_test_precision": { + "mean": 0.8894789934158325, + "std": 0.002276886323722618, + "n": 5, + "ci95": 0.0028266745457896443 + }, + "bank_marketing_test_recall": { + "mean": 0.8989564776420593, + "std": 0.0011167611710139317, + "n": 5, + "ci95": 0.0013864198414043813 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8923503756523132, + "std": 0.0023351126362635356, + "n": 5, + "ci95": 0.0028989604714591837 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9252730011940002, + "std": 0.0021193533243827846, + "n": 5, + "ci95": 0.00263110285004167 + }, + "bank_marketing_test_score": { + "mean": 0.8989564776420593, + "std": 0.0011167611710139317, + "n": 5, + "ci95": 0.0013864198414043813 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..d1b9f887166e752fe5c51fa06827dc55dfea0ee2 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e0fce5d1868559a0e46564cee1f7277d8bd80cc2a84d75d56e167a2a7bc21a8c +size 185184384 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..af0ba03d43c3e02b861b03bde640171a32ec7bc4 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:170429016689d73cd7f2934ca4988dfac2ee5e2ef14cc8b26c724e795f2d6f50 +size 15473582 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..72f0c7fd794a394de5830e788bad69812d677f49 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_031325/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: bank_marketing +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.001 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.3 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.1 +mask_min_trgt_share: 0.1 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 256 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.1 +recon_y_weight: 1.0 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.0005 +sw_s_y_weight: 0.0001 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 15 +vtjepa_kl_x_end_weight: 0.0001 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 15 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 15 +wjepa_joint_end_weight: 0.004 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..b7b4d7af63eb2b52a9c2ad7bc507c1ace64e0236 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-16T02:38:13.665106+00:00", + "spec": { + "dataset": "bank_marketing", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.001, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0001, + "lambda_joint": 0.004, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..03c635bbb69721b87d4036e31d73585fc3a37603 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "bank_marketing", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.001, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0001, + "lambda_joint": 0.004, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "bank_marketing", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.004", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.001", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "0.0001", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..1fbdf3d554888d76dc5539ee214f9a499be4f423 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "bank_marketing", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.004, + "s_x": 0.001, + "z": 0.0, + "s_y": 0.0001 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 15 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "bank_marketing", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.004", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.001", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "0.0001", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/downstream_results/autoint_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/downstream_results/autoint_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..c75fee95cdab62456f7d062f8e3bb3dd7112c933 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/downstream_results/autoint_bank_marketing_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "bank_marketing", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpj1fdn3ft", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpj1fdn3ft/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.21948149800300598, + "lr": 3.942650891985977e-06, + "bank_marketing_val_accuracy": 0.9023728966712952, + "bank_marketing_val_precision": 0.8952900767326355, + "bank_marketing_val_recall": 0.9023728966712952, + "bank_marketing_val_f1-score": 0.8980438113212585, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9229846000671387, + "bank_marketing_val_score": 0.9023728966712952 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22231349349021912, + "lr": 3.942649982491275e-06, + "bank_marketing_test_accuracy": 0.8984236121177673, + "bank_marketing_test_precision": 0.8893837332725525, + "bank_marketing_test_recall": 0.8984236121177673, + "bank_marketing_test_f1-score": 0.8925639390945435, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9235153198242188, + "bank_marketing_test_score": 0.8984236121177673 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.21828456223011017, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9007909893989563, + "bank_marketing_val_precision": 0.8930777907371521, + "bank_marketing_val_recall": 0.9007909893989563, + "bank_marketing_val_f1-score": 0.8960306644439697, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9228183031082153, + "bank_marketing_val_score": 0.9007909893989563 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21871507167816162, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8996447324752808, + "bank_marketing_test_precision": 0.8903918862342834, + "bank_marketing_test_recall": 0.8996447324752808, + "bank_marketing_test_f1-score": 0.8935130834579468, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9251676797866821, + "bank_marketing_test_score": 0.8996447324752808 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.22860261797904968, + "lr": 9.866358823273913e-07, + "bank_marketing_val_accuracy": 0.8960452079772949, + "bank_marketing_val_precision": 0.887739360332489, + "bank_marketing_val_recall": 0.8960452079772949, + "bank_marketing_val_f1-score": 0.8910025954246521, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.916431188583374, + "bank_marketing_val_score": 0.8960452079772949 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21838077902793884, + "lr": 9.86635200206365e-07, + "bank_marketing_test_accuracy": 0.9023090600967407, + "bank_marketing_test_precision": 0.8939895629882812, + "bank_marketing_test_recall": 0.9023090600967407, + "bank_marketing_test_f1-score": 0.8968508839607239, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9256624579429626, + "bank_marketing_test_score": 0.9023090600967407 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.2225060611963272, + "lr": 3.942650891985977e-06, + "bank_marketing_val_accuracy": 0.9032768607139587, + "bank_marketing_val_precision": 0.8956463932991028, + "bank_marketing_val_recall": 0.9032768607139587, + "bank_marketing_val_f1-score": 0.8984830975532532, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9205772280693054, + "bank_marketing_val_score": 0.9032768607139587 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2180434763431549, + "lr": 3.942649982491275e-06, + "bank_marketing_test_accuracy": 0.9030861258506775, + "bank_marketing_test_precision": 0.894201397895813, + "bank_marketing_test_recall": 0.9030861258506775, + "bank_marketing_test_f1-score": 0.8970361948013306, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9257867336273193, + "bank_marketing_test_score": 0.9030861258506775 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.21735332906246185, + "lr": 9.54915230977349e-05, + "bank_marketing_val_accuracy": 0.9003390073776245, + "bank_marketing_val_precision": 0.89178067445755, + "bank_marketing_val_recall": 0.9003390073776245, + "bank_marketing_val_f1-score": 0.8949207663536072, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9221661686897278, + "bank_marketing_val_score": 0.9003390073776245 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21332308650016785, + "lr": 9.549142851028591e-05, + "bank_marketing_test_accuracy": 0.9020870327949524, + "bank_marketing_test_precision": 0.8941370248794556, + "bank_marketing_test_recall": 0.9020870327949524, + "bank_marketing_test_f1-score": 0.8969678282737732, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9272551536560059, + "bank_marketing_test_score": 0.9020870327949524 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.21815518140792847, + "std": 0.003202247257539722, + "n": 5, + "ci95": 0.003975477702994431 + }, + "lr": { + "mean": 2.0872672735094966e-05, + "std": 4.1750248532764225e-05, + "n": 5, + "ci95": 5.183146983596762e-05 + }, + "bank_marketing_test_accuracy": { + "mean": 0.9011101126670837, + "std": 0.0019787054028491827, + "n": 5, + "ci95": 0.0024564933864180106 + }, + "bank_marketing_test_precision": { + "mean": 0.8924207210540771, + "std": 0.002340792566704576, + "n": 5, + "ci95": 0.002906011905969667 + }, + "bank_marketing_test_recall": { + "mean": 0.9011101126670837, + "std": 0.0019787054028491827, + "n": 5, + "ci95": 0.0024564933864180106 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8953863859176636, + "std": 0.0021704292185748286, + "n": 5, + "ci95": 0.002694511782016828 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9254774689674378, + "std": 0.0013449932155623699, + "n": 5, + "ci95": 0.0016697619231486399 + }, + "bank_marketing_test_score": { + "mean": 0.9011101126670837, + "std": 0.0019787054028491827, + "n": 5, + "ci95": 0.0024564933864180106 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/downstream_results/dcnv2_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/downstream_results/dcnv2_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..6b3ac067a58b59b5e28a8f6ff51d226fec8b03fa --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/downstream_results/dcnv2_bank_marketing_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "bank_marketing", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp7e7nf9r6", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp7e7nf9r6/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.24878370761871338, + "lr": 9.54915230977349e-05, + "bank_marketing_val_accuracy": 0.8996610045433044, + "bank_marketing_val_precision": 0.8923066258430481, + "bank_marketing_val_recall": 0.8996610045433044, + "bank_marketing_val_f1-score": 0.8952116966247559, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9083406925201416, + "bank_marketing_val_score": 0.8996610045433044 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.24591408669948578, + "lr": 9.549142851028591e-05, + "bank_marketing_test_accuracy": 0.8958703279495239, + "bank_marketing_test_precision": 0.88845294713974, + "bank_marketing_test_recall": 0.8958703279495239, + "bank_marketing_test_f1-score": 0.8914102911949158, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9115240573883057, + "bank_marketing_test_score": 0.8958703279495239 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.24530549347400665, + "lr": 0.000315937795676291, + "bank_marketing_val_accuracy": 0.8955932259559631, + "bank_marketing_val_precision": 0.8950775861740112, + "bank_marketing_val_recall": 0.8955932259559631, + "bank_marketing_val_f1-score": 0.8953327536582947, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9046961069107056, + "bank_marketing_val_score": 0.8955932259559631 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2277309000492096, + "lr": 0.0003159379994031042, + "bank_marketing_test_accuracy": 0.8948712348937988, + "bank_marketing_test_precision": 0.8941749334335327, + "bank_marketing_test_recall": 0.8948712348937988, + "bank_marketing_test_f1-score": 0.8945179581642151, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9153793454170227, + "bank_marketing_test_score": 0.8948712348937988 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.2553221881389618, + "lr": 0.00015772644837852567, + "bank_marketing_val_accuracy": 0.8964971899986267, + "bank_marketing_val_precision": 0.8871020674705505, + "bank_marketing_val_recall": 0.8964971899986267, + "bank_marketing_val_f1-score": 0.8905891180038452, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9006679654121399, + "bank_marketing_val_score": 0.8964971899986267 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.24071477353572845, + "lr": 0.00015772631741128862, + "bank_marketing_test_accuracy": 0.8957593441009521, + "bank_marketing_test_precision": 0.8868793845176697, + "bank_marketing_test_recall": 0.8957593441009521, + "bank_marketing_test_f1-score": 0.8901763558387756, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9093999266624451, + "bank_marketing_test_score": 0.8957593441009521 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.2640564441680908, + "lr": 6.184667290654033e-05, + "bank_marketing_val_accuracy": 0.894463300704956, + "bank_marketing_val_precision": 0.8883780241012573, + "bank_marketing_val_recall": 0.894463300704956, + "bank_marketing_val_f1-score": 0.8910015821456909, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.8993784189224243, + "bank_marketing_val_score": 0.894463300704956 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2508843243122101, + "lr": 6.184665107866749e-05, + "bank_marketing_test_accuracy": 0.8948712348937988, + "bank_marketing_test_precision": 0.8876044750213623, + "bank_marketing_test_recall": 0.8948712348937988, + "bank_marketing_test_f1-score": 0.8905467987060547, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.908989429473877, + "bank_marketing_test_score": 0.8948712348937988 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.2521127462387085, + "lr": 0.00020610730280168355, + "bank_marketing_val_accuracy": 0.8978531360626221, + "bank_marketing_val_precision": 0.8879690170288086, + "bank_marketing_val_recall": 0.8978531360626221, + "bank_marketing_val_f1-score": 0.8914575576782227, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9041525721549988, + "bank_marketing_val_score": 0.8978531360626221 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23682673275470734, + "lr": 0.00020610728824976832, + "bank_marketing_test_accuracy": 0.8948712348937988, + "bank_marketing_test_precision": 0.8856142163276672, + "bank_marketing_test_recall": 0.8948712348937988, + "bank_marketing_test_f1-score": 0.8890246748924255, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9119967222213745, + "bank_marketing_test_score": 0.8948712348937988 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.24041416347026826, + "std": 0.008855769590241565, + "n": 5, + "ci95": 0.010994127472814225 + }, + "lr": { + "mean": 0.0001674219369306229, + "std": 9.995964345825973e-05, + "n": 5, + "ci95": 0.00012409639287907343 + }, + "bank_marketing_test_accuracy": { + "mean": 0.8952486753463745, + "std": 0.0005183190189511494, + "n": 5, + "ci95": 0.0006434748903372845 + }, + "bank_marketing_test_precision": { + "mean": 0.8885451912879944, + "std": 0.0033148521504196933, + "n": 5, + "ci95": 0.004115272729702125 + }, + "bank_marketing_test_recall": { + "mean": 0.8952486753463745, + "std": 0.0005183190189511494, + "n": 5, + "ci95": 0.0006434748903372845 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8911352157592773, + "std": 0.0020759722059661, + "n": 5, + "ci95": 0.002577246712421306 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.911457896232605, + "std": 0.002549560812243796, + "n": 5, + "ci95": 0.0031651903636232106 + }, + "bank_marketing_test_score": { + "mean": 0.8952486753463745, + "std": 0.0005183190189511494, + "n": 5, + "ci95": 0.0006434748903372845 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/downstream_results/fttransformer_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/downstream_results/fttransformer_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..3f931c8446c55c25b9821e6d81a2fd65a3102144 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/downstream_results/fttransformer_bank_marketing_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "bank_marketing", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpthvldpip", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpthvldpip/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.22677680850028992, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.897175133228302, + "bank_marketing_val_precision": 0.8865931034088135, + "bank_marketing_val_recall": 0.897175133228302, + "bank_marketing_val_f1-score": 0.8902133107185364, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9102813601493835, + "bank_marketing_val_score": 0.897175133228302 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22284826636314392, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.9004218578338623, + "bank_marketing_test_precision": 0.890540599822998, + "bank_marketing_test_recall": 0.9004218578338623, + "bank_marketing_test_f1-score": 0.893615186214447, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9157853722572327, + "bank_marketing_test_score": 0.9004218578338623 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.23019227385520935, + "lr": 2.447173756081611e-05, + "bank_marketing_val_accuracy": 0.9010169506072998, + "bank_marketing_val_precision": 0.8917264938354492, + "bank_marketing_val_recall": 0.9010169506072998, + "bank_marketing_val_f1-score": 0.8949300050735474, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9059411287307739, + "bank_marketing_val_score": 0.9010169506072998 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22492770850658417, + "lr": 2.447175393172074e-05, + "bank_marketing_test_accuracy": 0.9003108143806458, + "bank_marketing_test_precision": 0.8894136548042297, + "bank_marketing_test_recall": 0.9003108143806458, + "bank_marketing_test_f1-score": 0.892410159111023, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.913726806640625, + "bank_marketing_test_score": 0.9003108143806458 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.22250057756900787, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8969491720199585, + "bank_marketing_val_precision": 0.8862643837928772, + "bank_marketing_val_recall": 0.8969491720199585, + "bank_marketing_val_f1-score": 0.8899129629135132, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9142277836799622, + "bank_marketing_val_score": 0.8969491720199585 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2215438336133957, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8990896940231323, + "bank_marketing_test_precision": 0.8880785703659058, + "bank_marketing_test_recall": 0.8990896940231323, + "bank_marketing_test_f1-score": 0.8912361264228821, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9173547625541687, + "bank_marketing_test_score": 0.8990896940231323 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.2246096283197403, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8994350433349609, + "bank_marketing_val_precision": 0.8892220258712769, + "bank_marketing_val_recall": 0.8994350433349609, + "bank_marketing_val_f1-score": 0.8926262259483337, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9116213321685791, + "bank_marketing_val_score": 0.8994350433349609 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22113539278507233, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8993117213249207, + "bank_marketing_test_precision": 0.8892467617988586, + "bank_marketing_test_recall": 0.8993117213249207, + "bank_marketing_test_f1-score": 0.8924291729927063, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9176443219184875, + "bank_marketing_test_score": 0.8993117213249207 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.22259756922721863, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.897175133228302, + "bank_marketing_val_precision": 0.8878017663955688, + "bank_marketing_val_recall": 0.897175133228302, + "bank_marketing_val_f1-score": 0.8912495374679565, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9147111177444458, + "bank_marketing_val_score": 0.897175133228302 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22052712738513947, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8998667597770691, + "bank_marketing_test_precision": 0.8896520137786865, + "bank_marketing_test_recall": 0.8998667597770691, + "bank_marketing_test_f1-score": 0.8927748203277588, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9190995693206787, + "bank_marketing_test_score": 0.8998667597770691 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.22219646573066712, + "std": 0.0017480722348999643, + "n": 5, + "ci95": 0.00217017039415241 + }, + "lr": { + "mean": 4.894350786344148e-06, + "std": 1.0944101063995064e-05, + "n": 5, + "ci95": 1.3586717782891353e-05 + }, + "bank_marketing_test_accuracy": { + "mean": 0.8998001694679261, + "std": 0.0005905602989024556, + "n": 5, + "ci95": 0.0007331599066975909 + }, + "bank_marketing_test_precision": { + "mean": 0.8893863201141358, + "std": 0.0008850689072488599, + "n": 5, + "ci95": 0.0010987820188141245 + }, + "bank_marketing_test_recall": { + "mean": 0.8998001694679261, + "std": 0.0005905602989024556, + "n": 5, + "ci95": 0.0007331599066975909 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8924930930137634, + "std": 0.0008557733698794005, + "n": 5, + "ci95": 0.0010624126362390246 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9167221665382386, + "std": 0.0020464821460132437, + "n": 5, + "ci95": 0.0025406358368786656 + }, + "bank_marketing_test_score": { + "mean": 0.8998001694679261, + "std": 0.0005905602989024556, + "n": 5, + "ci95": 0.0007331599066975909 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/downstream_results/mlp_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/downstream_results/mlp_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..5c3b97cccb8b0689ae243df2504d3e9e361bd1f0 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/downstream_results/mlp_bank_marketing_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "bank_marketing", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpmw_unsu9", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpmw_unsu9/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.22977767884731293, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8978531360626221, + "bank_marketing_val_precision": 0.8853709697723389, + "bank_marketing_val_recall": 0.8978531360626221, + "bank_marketing_val_f1-score": 0.88892662525177, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9060643911361694, + "bank_marketing_val_score": 0.8978531360626221 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22559453547000885, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8974245190620422, + "bank_marketing_test_precision": 0.884318470954895, + "bank_marketing_test_recall": 0.8974245190620422, + "bank_marketing_test_f1-score": 0.8872795104980469, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9136142730712891, + "bank_marketing_test_score": 0.8974245190620422 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.22912684082984924, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8967231512069702, + "bank_marketing_val_precision": 0.8824706077575684, + "bank_marketing_val_recall": 0.8967231512069702, + "bank_marketing_val_f1-score": 0.8859037756919861, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9075892567634583, + "bank_marketing_val_score": 0.8967231512069702 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22638066112995148, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8996447324752808, + "bank_marketing_test_precision": 0.8857904672622681, + "bank_marketing_test_recall": 0.8996447324752808, + "bank_marketing_test_f1-score": 0.8871031999588013, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9144390821456909, + "bank_marketing_test_score": 0.8996447324752808 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.2247917205095291, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8974011540412903, + "bank_marketing_val_precision": 0.8834943771362305, + "bank_marketing_val_recall": 0.8974011540412903, + "bank_marketing_val_f1-score": 0.8868569731712341, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.909329354763031, + "bank_marketing_val_score": 0.8974011540412903 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.223946213722229, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8984236121177673, + "bank_marketing_test_precision": 0.8844873905181885, + "bank_marketing_test_recall": 0.8984236121177673, + "bank_marketing_test_f1-score": 0.8865182399749756, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.914363443851471, + "bank_marketing_test_score": 0.8984236121177673 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.22808919847011566, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8985310792922974, + "bank_marketing_val_precision": 0.8898804187774658, + "bank_marketing_val_recall": 0.8985310792922974, + "bank_marketing_val_f1-score": 0.8931238651275635, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.907413125038147, + "bank_marketing_val_score": 0.8985310792922974 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2233138233423233, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8992007374763489, + "bank_marketing_test_precision": 0.8892562389373779, + "bank_marketing_test_recall": 0.8992007374763489, + "bank_marketing_test_f1-score": 0.8924477696418762, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9150081872940063, + "bank_marketing_test_score": 0.8992007374763489 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.22907376289367676, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8983050584793091, + "bank_marketing_val_precision": 0.8845475912094116, + "bank_marketing_val_recall": 0.8983050584793091, + "bank_marketing_val_f1-score": 0.8877190947532654, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9051838517189026, + "bank_marketing_val_score": 0.8983050584793091 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22502531111240387, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8949822187423706, + "bank_marketing_test_precision": 0.8797292113304138, + "bank_marketing_test_recall": 0.8949822187423706, + "bank_marketing_test_f1-score": 0.8823571801185608, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9136731624603271, + "bank_marketing_test_score": 0.8949822187423706 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.2248521089553833, + "std": 0.001235378147369437, + "n": 5, + "ci95": 0.0015336786589699638 + }, + "lr": { + "mean": 0.0009999994654208422, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_accuracy": { + "mean": 0.8979351639747619, + "std": 0.0018536131005875956, + "n": 5, + "ci95": 0.00230119567875778 + }, + "bank_marketing_test_precision": { + "mean": 0.8847163558006287, + "std": 0.0034222288696545197, + "n": 5, + "ci95": 0.004248577162123346 + }, + "bank_marketing_test_recall": { + "mean": 0.8979351639747619, + "std": 0.0018536131005875956, + "n": 5, + "ci95": 0.00230119567875778 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8871411800384521, + "std": 0.003586613325195341, + "n": 5, + "ci95": 0.004452654700540383 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9142196297645568, + "std": 0.0005821829731940016, + "n": 5, + "ci95": 0.0007227597505303035 + }, + "bank_marketing_test_score": { + "mean": 0.8979351639747619, + "std": 0.0018536131005875956, + "n": 5, + "ci95": 0.00230119567875778 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/downstream_results/resnet_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/downstream_results/resnet_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..7603746613b8ed674456071653db1c2279974bf7 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/downstream_results/resnet_bank_marketing_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "bank_marketing", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp0nih1wtg", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp0nih1wtg/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.21889719367027283, + "lr": 0.0004999999655410647, + "bank_marketing_val_accuracy": 0.9010169506072998, + "bank_marketing_val_precision": 0.8943459391593933, + "bank_marketing_val_recall": 0.9010169506072998, + "bank_marketing_val_f1-score": 0.8970322012901306, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9175759553909302, + "bank_marketing_val_score": 0.9010169506072998 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21612519025802612, + "lr": 0.0004999997327104211, + "bank_marketing_test_accuracy": 0.8986456394195557, + "bank_marketing_test_precision": 0.891594648361206, + "bank_marketing_test_recall": 0.8986456394195557, + "bank_marketing_test_f1-score": 0.8943783640861511, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9227721095085144, + "bank_marketing_test_score": 0.8986456394195557 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.21721623837947845, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9010169506072998, + "bank_marketing_val_precision": 0.8924801349639893, + "bank_marketing_val_recall": 0.9010169506072998, + "bank_marketing_val_f1-score": 0.8955821394920349, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9208208322525024, + "bank_marketing_val_score": 0.9010169506072998 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21192345023155212, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8997557759284973, + "bank_marketing_test_precision": 0.8920477032661438, + "bank_marketing_test_recall": 0.8997557759284973, + "bank_marketing_test_f1-score": 0.8949436545372009, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9275190234184265, + "bank_marketing_test_score": 0.8997557759284973 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.2241469770669937, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9007909893989563, + "bank_marketing_val_precision": 0.8896549940109253, + "bank_marketing_val_recall": 0.9007909893989563, + "bank_marketing_val_f1-score": 0.8929097652435303, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9155095219612122, + "bank_marketing_val_score": 0.9007909893989563 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21976067125797272, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8957593441009521, + "bank_marketing_test_precision": 0.8836522102355957, + "bank_marketing_test_recall": 0.8957593441009521, + "bank_marketing_test_f1-score": 0.8871678113937378, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.921647310256958, + "bank_marketing_test_score": 0.8957593441009521 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.21638795733451843, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9030508399009705, + "bank_marketing_val_precision": 0.8936535716056824, + "bank_marketing_val_recall": 0.9030508399009705, + "bank_marketing_val_f1-score": 0.8967077136039734, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9203115701675415, + "bank_marketing_val_score": 0.9030508399009705 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2147349715232849, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.9017540216445923, + "bank_marketing_test_precision": 0.8928338885307312, + "bank_marketing_test_recall": 0.9017540216445923, + "bank_marketing_test_f1-score": 0.89577716588974, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.924019992351532, + "bank_marketing_test_score": 0.9017540216445923 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.21550880372524261, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9016948938369751, + "bank_marketing_val_precision": 0.8917353749275208, + "bank_marketing_val_recall": 0.9016948938369751, + "bank_marketing_val_f1-score": 0.8949263095855713, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9207606911659241, + "bank_marketing_val_score": 0.9016948938369751 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21002539992332458, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8985346555709839, + "bank_marketing_test_precision": 0.8884825706481934, + "bank_marketing_test_recall": 0.8985346555709839, + "bank_marketing_test_f1-score": 0.8917370438575745, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9276644587516785, + "bank_marketing_test_score": 0.8985346555709839 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.2145139366388321, + "std": 0.0037759369857679314, + "n": 5, + "ci95": 0.004687693387663463 + }, + "lr": { + "mean": 9.999994654208422e-05, + "std": 0.00022360667821444535, + "n": 5, + "ci95": 0.0002775998516008258 + }, + "bank_marketing_test_accuracy": { + "mean": 0.8988898873329163, + "std": 0.002175958179190269, + "n": 5, + "ci95": 0.002701375792781664 + }, + "bank_marketing_test_precision": { + "mean": 0.889722204208374, + "std": 0.0037741916514751122, + "n": 5, + "ci95": 0.004685526616328414 + }, + "bank_marketing_test_recall": { + "mean": 0.8988898873329163, + "std": 0.002175958179190269, + "n": 5, + "ci95": 0.002701375792781664 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8928008079528809, + "std": 0.0034929008607329456, + "n": 5, + "ci95": 0.0043363139613655 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9247245788574219, + "std": 0.002749089619278021, + "n": 5, + "ci95": 0.003412898382297281 + }, + "bank_marketing_test_score": { + "mean": 0.8988898873329163, + "std": 0.002175958179190269, + "n": 5, + "ci95": 0.002701375792781664 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..fc39c7a2fbc214298b908a7ff98073ec8731e66b --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9f05be00ee66ac8205c7c52ed8a2f20d4a98b8e6e86ea9ab4332f2dc9655bea7 +size 185184384 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..017caf0915946977ad1fcab42176c5ae793d5064 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cb265cceaa72dc8e09f8f607cea12b79241de7ca17be12baa77f2c88d3394230 +size 15472046 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f4332e21be8ce926cf7dad008fe55f826053ed1f --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.004_sx0.001_sy0.0001/seed_42/wjepa_deterministic/bank_marketing/20260916_010201/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: bank_marketing +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.001 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.3 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.1 +mask_min_trgt_share: 0.1 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 256 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.1 +recon_y_weight: 1.0 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.001 +sw_s_y_weight: 0.0001 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 15 +vtjepa_kl_x_end_weight: 0.0001 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 15 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 15 +wjepa_joint_end_weight: 0.004 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..fe0aa7d3414aead029238b9f9588cb2928fdc8b3 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-16T01:55:01.850484+00:00", + "spec": { + "dataset": "bank_marketing", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.002, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0002, + "lambda_joint": 0.008, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..d903ef894485ff2eddd9d6942c01be25500e24fe --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "bank_marketing", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.002, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0002, + "lambda_joint": 0.008, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "bank_marketing", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.008", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.002", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "0.0002", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..ea0e596976a3bff6d1f6032f659fbc40334b3167 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "bank_marketing", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.008, + "s_x": 0.002, + "z": 0.0, + "s_y": 0.0002 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 15 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "bank_marketing", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.008", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.002", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "0.0002", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/downstream_results/autoint_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/downstream_results/autoint_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..697133edea4100e2e15cd0864177ef4151b403b9 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/downstream_results/autoint_bank_marketing_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "bank_marketing", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpe44vwqdh", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpe44vwqdh/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.22757205367088318, + "lr": 1.5708417777204886e-05, + "bank_marketing_val_accuracy": 0.9003390073776245, + "bank_marketing_val_precision": 0.8919113874435425, + "bank_marketing_val_recall": 0.9003390073776245, + "bank_marketing_val_f1-score": 0.8950281143188477, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9164072275161743, + "bank_marketing_val_score": 0.9003390073776245 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21744327247142792, + "lr": 1.57084250531625e-05, + "bank_marketing_test_accuracy": 0.9010879397392273, + "bank_marketing_test_precision": 0.8934049010276794, + "bank_marketing_test_recall": 0.9010879397392273, + "bank_marketing_test_f1-score": 0.8962408900260925, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9262787699699402, + "bank_marketing_test_score": 0.9010879397392273 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.2200772911310196, + "lr": 3.942650891985977e-06, + "bank_marketing_val_accuracy": 0.8989830613136292, + "bank_marketing_val_precision": 0.890253484249115, + "bank_marketing_val_recall": 0.8989830613136292, + "bank_marketing_val_f1-score": 0.8934911489486694, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.919632077217102, + "bank_marketing_val_score": 0.8989830613136292 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21413947641849518, + "lr": 3.942649982491275e-06, + "bank_marketing_test_accuracy": 0.8993117213249207, + "bank_marketing_test_precision": 0.8900396227836609, + "bank_marketing_test_recall": 0.8993117213249207, + "bank_marketing_test_f1-score": 0.8931862711906433, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9255951642990112, + "bank_marketing_test_score": 0.8993117213249207 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.22331640124320984, + "lr": 9.866358823273913e-07, + "bank_marketing_val_accuracy": 0.9012429118156433, + "bank_marketing_val_precision": 0.8921700119972229, + "bank_marketing_val_recall": 0.9012429118156433, + "bank_marketing_val_f1-score": 0.8953343629837036, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9210175275802612, + "bank_marketing_val_score": 0.9012429118156433 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2156042903661728, + "lr": 9.86635200206365e-07, + "bank_marketing_test_accuracy": 0.903197169303894, + "bank_marketing_test_precision": 0.8956077098846436, + "bank_marketing_test_recall": 0.903197169303894, + "bank_marketing_test_f1-score": 0.8983319401741028, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9273210167884827, + "bank_marketing_test_score": 0.903197169303894 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.2209695726633072, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9046327471733093, + "bank_marketing_val_precision": 0.8970348834991455, + "bank_marketing_val_recall": 0.9046327471733093, + "bank_marketing_val_f1-score": 0.8998053073883057, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9214191436767578, + "bank_marketing_val_score": 0.9046327471733093 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21803201735019684, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8976465463638306, + "bank_marketing_test_precision": 0.8894157409667969, + "bank_marketing_test_recall": 0.8976465463638306, + "bank_marketing_test_f1-score": 0.8925023674964905, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9263976216316223, + "bank_marketing_test_score": 0.8976465463638306 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.21899819374084473, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9016948938369751, + "bank_marketing_val_precision": 0.8935623168945312, + "bank_marketing_val_recall": 0.9016948938369751, + "bank_marketing_val_f1-score": 0.8965616226196289, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9229053854942322, + "bank_marketing_val_score": 0.9016948938369751 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2184963971376419, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.9016429781913757, + "bank_marketing_test_precision": 0.8933506011962891, + "bank_marketing_test_recall": 0.9016429781913757, + "bank_marketing_test_f1-score": 0.8962490558624268, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9264887571334839, + "bank_marketing_test_score": 0.9016429781913757 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.21674309074878692, + "std": 0.0018235503430180475, + "n": 5, + "ci95": 0.0022638738192021605 + }, + "lr": { + "mean": 4.127542047172028e-06, + "std": 6.672509625277484e-06, + "n": 5, + "ci95": 8.283686768986896e-06 + }, + "bank_marketing_test_accuracy": { + "mean": 0.9005772709846497, + "std": 0.0021477397639649413, + "n": 5, + "ci95": 0.002666343619585795 + }, + "bank_marketing_test_precision": { + "mean": 0.892363715171814, + "std": 0.0025823182922880624, + "n": 5, + "ci95": 0.0032058576266572087 + }, + "bank_marketing_test_recall": { + "mean": 0.9005772709846497, + "std": 0.0021477397639649413, + "n": 5, + "ci95": 0.002666343619585795 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8953021049499512, + "std": 0.0024121161477358077, + "n": 5, + "ci95": 0.0029945576312942084 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9264162659645081, + "std": 0.0006158915748357907, + "n": 5, + "ci95": 0.0007646077976823562 + }, + "bank_marketing_test_score": { + "mean": 0.9005772709846497, + "std": 0.0021477397639649413, + "n": 5, + "ci95": 0.002666343619585795 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/downstream_results/dcnv2_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/downstream_results/dcnv2_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..1e419a28ddfe7e32eeb73da5f26405bf239f2826 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/downstream_results/dcnv2_bank_marketing_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "bank_marketing", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp6wrdplh4", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp6wrdplh4/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.24813303351402283, + "lr": 7.783600449329242e-05, + "bank_marketing_val_accuracy": 0.8967231512069702, + "bank_marketing_val_precision": 0.8877082467079163, + "bank_marketing_val_recall": 0.8967231512069702, + "bank_marketing_val_f1-score": 0.8911083936691284, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9048277139663696, + "bank_marketing_val_score": 0.8967231512069702 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.24406498670578003, + "lr": 7.783605542499572e-05, + "bank_marketing_test_accuracy": 0.8987566828727722, + "bank_marketing_test_precision": 0.8896718621253967, + "bank_marketing_test_recall": 0.8987566828727722, + "bank_marketing_test_f1-score": 0.8928369283676147, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9094236493110657, + "bank_marketing_test_score": 0.8987566828727722 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.24413006007671356, + "lr": 0.00011474337225081399, + "bank_marketing_val_accuracy": 0.8989830613136292, + "bank_marketing_val_precision": 0.8930743932723999, + "bank_marketing_val_recall": 0.8989830613136292, + "bank_marketing_val_f1-score": 0.895570695400238, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9109374284744263, + "bank_marketing_val_score": 0.8989830613136292 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23934519290924072, + "lr": 0.0001147434813901782, + "bank_marketing_test_accuracy": 0.8970914483070374, + "bank_marketing_test_precision": 0.8912099599838257, + "bank_marketing_test_recall": 0.8970914483070374, + "bank_marketing_test_f1-score": 0.8936870098114014, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9154404401779175, + "bank_marketing_test_score": 0.8970914483070374 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.24874883890151978, + "lr": 0.00013551567099057138, + "bank_marketing_val_accuracy": 0.8946892619132996, + "bank_marketing_val_precision": 0.8882268071174622, + "bank_marketing_val_recall": 0.8946892619132996, + "bank_marketing_val_f1-score": 0.8909763097763062, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9057518243789673, + "bank_marketing_val_score": 0.8946892619132996 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23492081463336945, + "lr": 0.00013551574375014752, + "bank_marketing_test_accuracy": 0.8960923552513123, + "bank_marketing_test_precision": 0.8901025652885437, + "bank_marketing_test_recall": 0.8960923552513123, + "bank_marketing_test_f1-score": 0.8926306962966919, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9122893810272217, + "bank_marketing_test_score": 0.8960923552513123 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.23882153630256653, + "lr": 0.00020610730280168355, + "bank_marketing_val_accuracy": 0.8978531360626221, + "bank_marketing_val_precision": 0.8937793970108032, + "bank_marketing_val_recall": 0.8978531360626221, + "bank_marketing_val_f1-score": 0.8956195712089539, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9065815210342407, + "bank_marketing_val_score": 0.8978531360626221 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22839707136154175, + "lr": 0.00020610728824976832, + "bank_marketing_test_accuracy": 0.8966474533081055, + "bank_marketing_test_precision": 0.8924087285995483, + "bank_marketing_test_recall": 0.8966474533081055, + "bank_marketing_test_f1-score": 0.8943046927452087, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9155615568161011, + "bank_marketing_test_score": 0.8966474533081055 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.23461836576461792, + "lr": 0.00020610730280168355, + "bank_marketing_val_accuracy": 0.8964971899986267, + "bank_marketing_val_precision": 0.8901751041412354, + "bank_marketing_val_recall": 0.8964971899986267, + "bank_marketing_val_f1-score": 0.8928479552268982, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9086210131645203, + "bank_marketing_val_score": 0.8964971899986267 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22198297083377838, + "lr": 0.00020610728824976832, + "bank_marketing_test_accuracy": 0.8980906009674072, + "bank_marketing_test_precision": 0.891491174697876, + "bank_marketing_test_recall": 0.8980906009674072, + "bank_marketing_test_f1-score": 0.8941675424575806, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9196552038192749, + "bank_marketing_test_score": 0.8980906009674072 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.23374220728874207, + "std": 0.008749177934428204, + "n": 5, + "ci95": 0.0108617976691073 + }, + "lr": { + "mean": 0.0001480619714129716, + "std": 5.687206365098086e-05, + "n": 5, + "ci95": 7.060467315114873e-05 + }, + "bank_marketing_test_accuracy": { + "mean": 0.8973357081413269, + "std": 0.001080314347479614, + "n": 5, + "ci95": 0.0013411723877717802 + }, + "bank_marketing_test_precision": { + "mean": 0.890976858139038, + "std": 0.00109960374595414, + "n": 5, + "ci95": 0.0013651194997129643 + }, + "bank_marketing_test_recall": { + "mean": 0.8973357081413269, + "std": 0.001080314347479614, + "n": 5, + "ci95": 0.0013411723877717802 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8935253739356994, + "std": 0.0007616183019760432, + "n": 5, + "ci95": 0.0009455224204093743 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9144740462303161, + "std": 0.0038483847777140904, + "n": 5, + "ci95": 0.004777634781425297 + }, + "bank_marketing_test_score": { + "mean": 0.8973357081413269, + "std": 0.001080314347479614, + "n": 5, + "ci95": 0.0013411723877717802 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/downstream_results/fttransformer_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/downstream_results/fttransformer_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..4be30be0430c4e67344b47cdbc166879a13d3039 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/downstream_results/fttransformer_bank_marketing_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "bank_marketing", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpvaj2mmh9", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpvaj2mmh9/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.22880519926548004, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8985310792922974, + "bank_marketing_val_precision": 0.8886829614639282, + "bank_marketing_val_recall": 0.8985310792922974, + "bank_marketing_val_f1-score": 0.8921210765838623, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9074322581291199, + "bank_marketing_val_score": 0.8985310792922974 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22725100815296173, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8983126282691956, + "bank_marketing_test_precision": 0.8883972764015198, + "bank_marketing_test_recall": 0.8983126282691956, + "bank_marketing_test_f1-score": 0.8916651606559753, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9123130440711975, + "bank_marketing_test_score": 0.8983126282691956 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.22681915760040283, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8998870253562927, + "bank_marketing_val_precision": 0.8893855214118958, + "bank_marketing_val_recall": 0.8998870253562927, + "bank_marketing_val_f1-score": 0.8927624821662903, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9089639186859131, + "bank_marketing_val_score": 0.8998870253562927 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2252751886844635, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.899533748626709, + "bank_marketing_test_precision": 0.8877384662628174, + "bank_marketing_test_recall": 0.899533748626709, + "bank_marketing_test_f1-score": 0.8906639814376831, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9136983752250671, + "bank_marketing_test_score": 0.899533748626709 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.22098636627197266, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8980790972709656, + "bank_marketing_val_precision": 0.8886933922767639, + "bank_marketing_val_recall": 0.8980790972709656, + "bank_marketing_val_f1-score": 0.8920936584472656, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9152908325195312, + "bank_marketing_val_score": 0.8980790972709656 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21986855566501617, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8979795575141907, + "bank_marketing_test_precision": 0.888217568397522, + "bank_marketing_test_recall": 0.8979795575141907, + "bank_marketing_test_f1-score": 0.8915020227432251, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9189613461494446, + "bank_marketing_test_score": 0.8979795575141907 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.22112132608890533, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.900564968585968, + "bank_marketing_val_precision": 0.8928942084312439, + "bank_marketing_val_recall": 0.900564968585968, + "bank_marketing_val_f1-score": 0.895845890045166, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9157688021659851, + "bank_marketing_val_score": 0.900564968585968 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21707281470298767, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.9005328416824341, + "bank_marketing_test_precision": 0.8920986652374268, + "bank_marketing_test_recall": 0.9005328416824341, + "bank_marketing_test_f1-score": 0.8950780630111694, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9220709204673767, + "bank_marketing_test_score": 0.9005328416824341 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.22017355263233185, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8951412439346313, + "bank_marketing_val_precision": 0.8832139372825623, + "bank_marketing_val_recall": 0.8951412439346313, + "bank_marketing_val_f1-score": 0.8871246576309204, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9161455035209656, + "bank_marketing_val_score": 0.8951412439346313 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21808013319969177, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8986456394195557, + "bank_marketing_test_precision": 0.8868905305862427, + "bank_marketing_test_recall": 0.8986456394195557, + "bank_marketing_test_f1-score": 0.8899965286254883, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9209418296813965, + "bank_marketing_test_score": 0.8986456394195557 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.22150954008102416, + "std": 0.004507847560562779, + "n": 5, + "ci95": 0.005596334706297386 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_accuracy": { + "mean": 0.899000883102417, + "std": 0.0010336635701109838, + "n": 5, + "ci95": 0.0012832570831931968 + }, + "bank_marketing_test_precision": { + "mean": 0.8886685013771057, + "std": 0.002004312419202772, + "n": 5, + "ci95": 0.002488283599467368 + }, + "bank_marketing_test_recall": { + "mean": 0.899000883102417, + "std": 0.0010336635701109838, + "n": 5, + "ci95": 0.0012832570831931968 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8917811512947083, + "std": 0.001961755591530528, + "n": 5, + "ci95": 0.0024354507899075075 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9175971031188965, + "std": 0.004364193960559817, + "n": 5, + "ci95": 0.0054179937982297725 + }, + "bank_marketing_test_score": { + "mean": 0.899000883102417, + "std": 0.0010336635701109838, + "n": 5, + "ci95": 0.0012832570831931968 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/downstream_results/mlp_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/downstream_results/mlp_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..1c071410e97ffed3ab43ca8595baef284818c190 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/downstream_results/mlp_bank_marketing_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "bank_marketing", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp4tt4_0l3", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp4tt4_0l3/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.2532872259616852, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8987570405006409, + "bank_marketing_val_precision": 0.8883705735206604, + "bank_marketing_val_recall": 0.8987570405006409, + "bank_marketing_val_f1-score": 0.8918443322181702, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9066269993782043, + "bank_marketing_val_score": 0.8987570405006409 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2405346930027008, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.897535502910614, + "bank_marketing_test_precision": 0.8857455253601074, + "bank_marketing_test_recall": 0.897535502910614, + "bank_marketing_test_f1-score": 0.8890309929847717, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9123899936676025, + "bank_marketing_test_score": 0.897535502910614 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.22823964059352875, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.9032768607139587, + "bank_marketing_val_precision": 0.8909090757369995, + "bank_marketing_val_recall": 0.9032768607139587, + "bank_marketing_val_f1-score": 0.893208384513855, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9073143601417542, + "bank_marketing_val_score": 0.9032768607139587 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2288522869348526, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.9016429781913757, + "bank_marketing_test_precision": 0.8890806436538696, + "bank_marketing_test_recall": 0.9016429781913757, + "bank_marketing_test_f1-score": 0.8908568024635315, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.911186158657074, + "bank_marketing_test_score": 0.9016429781913757 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.2282499372959137, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8942372798919678, + "bank_marketing_val_precision": 0.88081955909729, + "bank_marketing_val_recall": 0.8942372798919678, + "bank_marketing_val_f1-score": 0.8848630785942078, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9072683453559875, + "bank_marketing_val_score": 0.8942372798919678 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22918492555618286, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8979795575141907, + "bank_marketing_test_precision": 0.8846979737281799, + "bank_marketing_test_recall": 0.8979795575141907, + "bank_marketing_test_f1-score": 0.8874108195304871, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9130426049232483, + "bank_marketing_test_score": 0.8979795575141907 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.22697462141513824, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8964971899986267, + "bank_marketing_val_precision": 0.8873831033706665, + "bank_marketing_val_recall": 0.8964971899986267, + "bank_marketing_val_f1-score": 0.8908141851425171, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9086675047874451, + "bank_marketing_val_score": 0.8964971899986267 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2227703481912613, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8978685736656189, + "bank_marketing_test_precision": 0.8881193399429321, + "bank_marketing_test_recall": 0.8978685736656189, + "bank_marketing_test_f1-score": 0.8914112448692322, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9155478477478027, + "bank_marketing_test_score": 0.8978685736656189 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.22666680812835693, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8983050584793091, + "bank_marketing_val_precision": 0.8840697407722473, + "bank_marketing_val_recall": 0.8983050584793091, + "bank_marketing_val_f1-score": 0.8870346546173096, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9088915586471558, + "bank_marketing_val_score": 0.8983050584793091 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23038692772388458, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8990896940231323, + "bank_marketing_test_precision": 0.8847148418426514, + "bank_marketing_test_recall": 0.8990896940231323, + "bank_marketing_test_f1-score": 0.885529100894928, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.914588987827301, + "bank_marketing_test_score": 0.8990896940231323 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.23034583628177643, + "std": 0.0064183508972354385, + "n": 5, + "ci95": 0.007968157618646122 + }, + "lr": { + "mean": 0.0009999994654208422, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_accuracy": { + "mean": 0.8988232612609863, + "std": 0.0016810246852982625, + "n": 5, + "ci95": 0.002086933211934705 + }, + "bank_marketing_test_precision": { + "mean": 0.8864716649055481, + "std": 0.002017503992837695, + "n": 5, + "ci95": 0.0025046604756531676 + }, + "bank_marketing_test_recall": { + "mean": 0.8988232612609863, + "std": 0.0016810246852982625, + "n": 5, + "ci95": 0.002086933211934705 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8888477921485901, + "std": 0.00243513740809411, + "n": 5, + "ci95": 0.003023137718929157 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9133511185646057, + "std": 0.0017366586129402418, + "n": 5, + "ci95": 0.002156000782638355 + }, + "bank_marketing_test_score": { + "mean": 0.8988232612609863, + "std": 0.0016810246852982625, + "n": 5, + "ci95": 0.002086933211934705 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/downstream_results/resnet_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/downstream_results/resnet_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..14a03d99be716fd1b2b0491243e9673607e92956 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/downstream_results/resnet_bank_marketing_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "bank_marketing", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp2k4hw49c", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp2k4hw49c/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.21838222444057465, + "lr": 0.0004999999655410647, + "bank_marketing_val_accuracy": 0.9003390073776245, + "bank_marketing_val_precision": 0.893959105014801, + "bank_marketing_val_recall": 0.9003390073776245, + "bank_marketing_val_f1-score": 0.8965775966644287, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9194280505180359, + "bank_marketing_val_score": 0.9003390073776245 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21638818085193634, + "lr": 0.0004999997327104211, + "bank_marketing_test_accuracy": 0.9017540216445923, + "bank_marketing_test_precision": 0.8951842784881592, + "bank_marketing_test_recall": 0.9017540216445923, + "bank_marketing_test_f1-score": 0.8977600932121277, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9235914945602417, + "bank_marketing_test_score": 0.9017540216445923 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.2211872786283493, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9021468758583069, + "bank_marketing_val_precision": 0.8943238854408264, + "bank_marketing_val_recall": 0.9021468758583069, + "bank_marketing_val_f1-score": 0.8972454071044922, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.918609619140625, + "bank_marketing_val_score": 0.9021468758583069 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21812550723552704, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.9006438851356506, + "bank_marketing_test_precision": 0.8925492167472839, + "bank_marketing_test_recall": 0.9006438851356506, + "bank_marketing_test_f1-score": 0.8954744338989258, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9244089126586914, + "bank_marketing_test_score": 0.9006438851356506 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.21614830195903778, + "lr": 0.0004999999655410647, + "bank_marketing_val_accuracy": 0.9030508399009705, + "bank_marketing_val_precision": 0.8909779787063599, + "bank_marketing_val_recall": 0.9030508399009705, + "bank_marketing_val_f1-score": 0.8935317993164062, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9209871888160706, + "bank_marketing_val_score": 0.9030508399009705 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2170896679162979, + "lr": 0.0004999997327104211, + "bank_marketing_test_accuracy": 0.900865912437439, + "bank_marketing_test_precision": 0.8892863988876343, + "bank_marketing_test_recall": 0.900865912437439, + "bank_marketing_test_f1-score": 0.8919957280158997, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9231061935424805, + "bank_marketing_test_score": 0.900865912437439 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.22052526473999023, + "lr": 0.0007938927737995982, + "bank_marketing_val_accuracy": 0.9021468758583069, + "bank_marketing_val_precision": 0.8938168287277222, + "bank_marketing_val_recall": 0.9021468758583069, + "bank_marketing_val_f1-score": 0.896826982498169, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9170241355895996, + "bank_marketing_val_score": 0.9021468758583069 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21697711944580078, + "lr": 0.0007938928320072591, + "bank_marketing_test_accuracy": 0.8928729891777039, + "bank_marketing_test_precision": 0.8815474510192871, + "bank_marketing_test_recall": 0.8928729891777039, + "bank_marketing_test_f1-score": 0.8853742480278015, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.922257661819458, + "bank_marketing_test_score": 0.8928729891777039 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.218247190117836, + "lr": 0.0007938927737995982, + "bank_marketing_val_accuracy": 0.9055367112159729, + "bank_marketing_val_precision": 0.8966887593269348, + "bank_marketing_val_recall": 0.9055367112159729, + "bank_marketing_val_f1-score": 0.899516224861145, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9223173260688782, + "bank_marketing_val_score": 0.9055367112159729 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21388763189315796, + "lr": 0.0007938928320072591, + "bank_marketing_test_accuracy": 0.8997557759284973, + "bank_marketing_test_precision": 0.889710009098053, + "bank_marketing_test_recall": 0.8997557759284973, + "bank_marketing_test_f1-score": 0.892848789691925, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9283337593078613, + "bank_marketing_test_score": 0.8997557759284973 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.216493621468544, + "std": 0.0015854330803462345, + "n": 5, + "ci95": 0.0019682595857225278 + }, + "lr": { + "mean": 0.0005175570258870721, + "std": 0.00032450144361028673, + "n": 5, + "ci95": 0.0004028571655810677 + }, + "bank_marketing_test_accuracy": { + "mean": 0.8991785168647766, + "std": 0.003595856653770663, + "n": 5, + "ci95": 0.004464129968905785 + }, + "bank_marketing_test_precision": { + "mean": 0.8896554708480835, + "std": 0.0051190979021443215, + "n": 5, + "ci95": 0.0063551805756110885 + }, + "bank_marketing_test_recall": { + "mean": 0.8991785168647766, + "std": 0.003595856653770663, + "n": 5, + "ci95": 0.004464129968905785 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8926906585693359, + "std": 0.0046767074374359984, + "n": 5, + "ci95": 0.00580596832339528 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9243396043777465, + "std": 0.0023650334377752287, + "n": 5, + "ci95": 0.0029361060975457987 + }, + "bank_marketing_test_score": { + "mean": 0.8991785168647766, + "std": 0.003595856653770663, + "n": 5, + "ci95": 0.004464129968905785 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..f39cdd6f79046bc2824578624a4ea5433b5b7201 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:11ecb3b8e7cfe6c27e022a9b896ac65be22981761d918083623ba7dffb241c60 +size 185184384 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..760921ce4598605ffd1d4615d4396d99486a8a23 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0869f3893202fcf2097054451800aac4c4ad1b5798bc154bf4eeefb9fad05524 +size 15469742 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..670d6f96d6d9f35aed5092d8e979d5803bfa5078 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.008_sx0.002_sy0.0002/seed_42/wjepa_deterministic/bank_marketing/20260916_002713/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: bank_marketing +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.001 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.3 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.1 +mask_min_trgt_share: 0.1 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 256 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.1 +recon_y_weight: 1.0 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.002 +sw_s_y_weight: 0.0002 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 15 +vtjepa_kl_x_end_weight: 0.0001 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 15 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 15 +wjepa_joint_end_weight: 0.008 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..376e722894505827adabde2dc97841a35cf5083f --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-16T05:50:50.323396+00:00", + "spec": { + "dataset": "bank_marketing", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.004, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0004, + "lambda_joint": 0.016, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..d7fe11239b622e79959ba440d7993c9f3aedbd34 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "bank_marketing", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.004, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0004, + "lambda_joint": 0.016, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "bank_marketing", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.016", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.004", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "0.0004", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..25825a91e6ad34e487d1ec8f8a6d3c6c96a7650f --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "bank_marketing", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.016, + "s_x": 0.004, + "z": 0.0, + "s_y": 0.0004 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 15 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "bank_marketing", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.016", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.004", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "0.0004", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/downstream_results/autoint_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/downstream_results/autoint_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..96d430156349beba7051d799ec3d05baeeadf75e --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/downstream_results/autoint_bank_marketing_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "bank_marketing", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpkm5gfl50", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpkm5gfl50/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.2282763123512268, + "lr": 4.758648356073536e-05, + "bank_marketing_val_accuracy": 0.8978531360626221, + "bank_marketing_val_precision": 0.8906064629554749, + "bank_marketing_val_recall": 0.8978531360626221, + "bank_marketing_val_f1-score": 0.8935333490371704, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9142635464668274, + "bank_marketing_val_score": 0.8978531360626221 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2182924747467041, + "lr": 4.758644718094729e-05, + "bank_marketing_test_accuracy": 0.9021980166435242, + "bank_marketing_test_precision": 0.8936693668365479, + "bank_marketing_test_recall": 0.9021980166435242, + "bank_marketing_test_f1-score": 0.8965560793876648, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9240198135375977, + "bank_marketing_test_score": 0.9021980166435242 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.22421376407146454, + "lr": 9.866358823273913e-07, + "bank_marketing_val_accuracy": 0.8994350433349609, + "bank_marketing_val_precision": 0.891029953956604, + "bank_marketing_val_recall": 0.8994350433349609, + "bank_marketing_val_f1-score": 0.8941837549209595, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9173430800437927, + "bank_marketing_val_score": 0.8994350433349609 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22159446775913239, + "lr": 9.86635200206365e-07, + "bank_marketing_test_accuracy": 0.8999778032302856, + "bank_marketing_test_precision": 0.8916221261024475, + "bank_marketing_test_recall": 0.8999778032302856, + "bank_marketing_test_f1-score": 0.8946207761764526, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9211511015892029, + "bank_marketing_test_score": 0.8999778032302856 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.21876733005046844, + "lr": 9.866358823273913e-07, + "bank_marketing_val_accuracy": 0.9041807651519775, + "bank_marketing_val_precision": 0.8971521854400635, + "bank_marketing_val_recall": 0.9041807651519775, + "bank_marketing_val_f1-score": 0.8998328447341919, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9215047359466553, + "bank_marketing_val_score": 0.9041807651519775 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21396416425704956, + "lr": 9.86635200206365e-07, + "bank_marketing_test_accuracy": 0.9027531147003174, + "bank_marketing_test_precision": 0.8954030275344849, + "bank_marketing_test_recall": 0.9027531147003174, + "bank_marketing_test_f1-score": 0.8981091380119324, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9276056289672852, + "bank_marketing_test_score": 0.9027531147003174 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.2197575569152832, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8998870253562927, + "bank_marketing_val_precision": 0.8895054459571838, + "bank_marketing_val_recall": 0.8998870253562927, + "bank_marketing_val_f1-score": 0.8928784728050232, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9193478226661682, + "bank_marketing_val_score": 0.8998870253562927 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2180752456188202, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8992007374763489, + "bank_marketing_test_precision": 0.8892562389373779, + "bank_marketing_test_recall": 0.8992007374763489, + "bank_marketing_test_f1-score": 0.8924477696418762, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9226531386375427, + "bank_marketing_test_score": 0.8992007374763489 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.21649889647960663, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.900564968585968, + "bank_marketing_val_precision": 0.8942745923995972, + "bank_marketing_val_recall": 0.900564968585968, + "bank_marketing_val_f1-score": 0.896861732006073, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9228026270866394, + "bank_marketing_val_score": 0.900564968585968 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21616466343402863, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8986456394195557, + "bank_marketing_test_precision": 0.8907520771026611, + "bank_marketing_test_recall": 0.8986456394195557, + "bank_marketing_test_f1-score": 0.8937296271324158, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9251865148544312, + "bank_marketing_test_score": 0.8986456394195557 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.21761820316314698, + "std": 0.002825729471453632, + "n": 5, + "ci95": 0.003508044071864993 + }, + "lr": { + "mean": 9.911943516272004e-06, + "std": 2.106646465352472e-05, + "n": 5, + "ci95": 2.6153277300439367e-05 + }, + "bank_marketing_test_accuracy": { + "mean": 0.9005550622940064, + "std": 0.0018264756244839005, + "n": 5, + "ci95": 0.0022675054535848897 + }, + "bank_marketing_test_precision": { + "mean": 0.8921405673027039, + "std": 0.002423680631363993, + "n": 5, + "ci95": 0.0030089145322806164 + }, + "bank_marketing_test_recall": { + "mean": 0.9005550622940064, + "std": 0.0018264756244839005, + "n": 5, + "ci95": 0.0022675054535848897 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8950926780700683, + "std": 0.0022537032771955995, + "n": 5, + "ci95": 0.002797893606298278 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9241232395172119, + "std": 0.0024628788093150884, + "n": 5, + "ci95": 0.0030575776959622094 + }, + "bank_marketing_test_score": { + "mean": 0.9005550622940064, + "std": 0.0018264756244839005, + "n": 5, + "ci95": 0.0022675054535848897 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/downstream_results/dcnv2_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/downstream_results/dcnv2_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..72311aee38249cbe97529bd72e7204c674550c72 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/downstream_results/dcnv2_bank_marketing_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "bank_marketing", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp45bjrp88", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp45bjrp88/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.2421044409275055, + "lr": 4.758648356073536e-05, + "bank_marketing_val_accuracy": 0.8989830613136292, + "bank_marketing_val_precision": 0.8903881311416626, + "bank_marketing_val_recall": 0.8989830613136292, + "bank_marketing_val_f1-score": 0.8935999274253845, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9103938937187195, + "bank_marketing_val_score": 0.8989830613136292 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23599492013454437, + "lr": 4.758644718094729e-05, + "bank_marketing_test_accuracy": 0.8989787101745605, + "bank_marketing_test_precision": 0.8895127177238464, + "bank_marketing_test_recall": 0.8989787101745605, + "bank_marketing_test_f1-score": 0.8926991820335388, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9164345860481262, + "bank_marketing_test_score": 0.8989787101745605 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.24268102645874023, + "lr": 6.184667290654033e-05, + "bank_marketing_val_accuracy": 0.8989830613136292, + "bank_marketing_val_precision": 0.8936686515808105, + "bank_marketing_val_recall": 0.8989830613136292, + "bank_marketing_val_f1-score": 0.8959634900093079, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9056128859519958, + "bank_marketing_val_score": 0.8989830613136292 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23693950474262238, + "lr": 6.184665107866749e-05, + "bank_marketing_test_accuracy": 0.8992007374763489, + "bank_marketing_test_precision": 0.8929637670516968, + "bank_marketing_test_recall": 0.8992007374763489, + "bank_marketing_test_f1-score": 0.8955118060112, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9119980931282043, + "bank_marketing_test_score": 0.8992007374763489 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.24836444854736328, + "lr": 7.783600449329242e-05, + "bank_marketing_val_accuracy": 0.8946892619132996, + "bank_marketing_val_precision": 0.8890167474746704, + "bank_marketing_val_recall": 0.8946892619132996, + "bank_marketing_val_f1-score": 0.8914905786514282, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9049744606018066, + "bank_marketing_val_score": 0.8946892619132996 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23731590807437897, + "lr": 7.783605542499572e-05, + "bank_marketing_test_accuracy": 0.8955373167991638, + "bank_marketing_test_precision": 0.8890393972396851, + "bank_marketing_test_recall": 0.8955373167991638, + "bank_marketing_test_f1-score": 0.8917389512062073, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9146032333374023, + "bank_marketing_test_score": 0.8955373167991638 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.24802711606025696, + "lr": 0.00011474337225081399, + "bank_marketing_val_accuracy": 0.8940113186836243, + "bank_marketing_val_precision": 0.8885356783866882, + "bank_marketing_val_recall": 0.8940113186836243, + "bank_marketing_val_f1-score": 0.8909449577331543, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9056969881057739, + "bank_marketing_val_score": 0.8940113186836243 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2370055764913559, + "lr": 0.0001147434813901782, + "bank_marketing_test_accuracy": 0.898201584815979, + "bank_marketing_test_precision": 0.8913757801055908, + "bank_marketing_test_recall": 0.898201584815979, + "bank_marketing_test_f1-score": 0.894112229347229, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9138755202293396, + "bank_marketing_test_score": 0.898201584815979 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.2573549449443817, + "lr": 9.54915230977349e-05, + "bank_marketing_val_accuracy": 0.8980790972709656, + "bank_marketing_val_precision": 0.8899284601211548, + "bank_marketing_val_recall": 0.8980790972709656, + "bank_marketing_val_f1-score": 0.893081545829773, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9054147601127625, + "bank_marketing_val_score": 0.8980790972709656 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.24103742837905884, + "lr": 9.549142851028591e-05, + "bank_marketing_test_accuracy": 0.897535502910614, + "bank_marketing_test_precision": 0.890245258808136, + "bank_marketing_test_recall": 0.897535502910614, + "bank_marketing_test_f1-score": 0.8931218981742859, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9131208062171936, + "bank_marketing_test_score": 0.897535502910614 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.23765866756439208, + "std": 0.0019522606981997688, + "n": 5, + "ci95": 0.0024236632127178108 + }, + "lr": { + "mean": 7.950081271701492e-05, + "std": 2.6602222915662248e-05, + "n": 5, + "ci95": 3.302572710533141e-05 + }, + "bank_marketing_test_accuracy": { + "mean": 0.8978907704353333, + "std": 0.0014714950845715262, + "n": 5, + "ci95": 0.0018268095585081293 + }, + "bank_marketing_test_precision": { + "mean": 0.890627384185791, + "std": 0.0015755145997027437, + "n": 5, + "ci95": 0.001955946139734577 + }, + "bank_marketing_test_recall": { + "mean": 0.8978907704353333, + "std": 0.0014714950845715262, + "n": 5, + "ci95": 0.0018268095585081293 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8934368133544922, + "std": 0.0014394259879888089, + "n": 5, + "ci95": 0.001786996899407683 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9140064477920532, + "std": 0.001664857340686009, + "n": 5, + "ci95": 0.0020668620204077827 + }, + "bank_marketing_test_score": { + "mean": 0.8978907704353333, + "std": 0.0014714950845715262, + "n": 5, + "ci95": 0.0018268095585081293 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/downstream_results/fttransformer_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/downstream_results/fttransformer_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..24dcddb34a2a502db8230162685e1f8c21cf84c5 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/downstream_results/fttransformer_bank_marketing_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "bank_marketing", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpjqadora5", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpjqadora5/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.2274584174156189, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9014689326286316, + "bank_marketing_val_precision": 0.8922443389892578, + "bank_marketing_val_recall": 0.9014689326286316, + "bank_marketing_val_f1-score": 0.8954098224639893, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9075677394866943, + "bank_marketing_val_score": 0.9014689326286316 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22433911263942719, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8996447324752808, + "bank_marketing_test_precision": 0.8881774544715881, + "bank_marketing_test_recall": 0.8996447324752808, + "bank_marketing_test_f1-score": 0.8911690711975098, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9144464135169983, + "bank_marketing_test_score": 0.8996447324752808 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.22684279084205627, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8976271152496338, + "bank_marketing_val_precision": 0.8869897723197937, + "bank_marketing_val_recall": 0.8976271152496338, + "bank_marketing_val_f1-score": 0.8905784487724304, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9093073606491089, + "bank_marketing_val_score": 0.8976271152496338 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2258182317018509, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8993117213249207, + "bank_marketing_test_precision": 0.8880416750907898, + "bank_marketing_test_recall": 0.8993117213249207, + "bank_marketing_test_f1-score": 0.8911294341087341, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9133086800575256, + "bank_marketing_test_score": 0.8993117213249207 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.22205166518688202, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8998870253562927, + "bank_marketing_val_precision": 0.8896262049674988, + "bank_marketing_val_recall": 0.8998870253562927, + "bank_marketing_val_f1-score": 0.8929939270019531, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9125630259513855, + "bank_marketing_val_score": 0.8998870253562927 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2213170826435089, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8996447324752808, + "bank_marketing_test_precision": 0.8883211612701416, + "bank_marketing_test_recall": 0.8996447324752808, + "bank_marketing_test_f1-score": 0.891344428062439, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9171424508094788, + "bank_marketing_test_score": 0.8996447324752808 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.22902105748653412, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8998870253562927, + "bank_marketing_val_precision": 0.889148473739624, + "bank_marketing_val_recall": 0.8998870253562927, + "bank_marketing_val_f1-score": 0.8925287127494812, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9084526896476746, + "bank_marketing_val_score": 0.8998870253562927 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22227229177951813, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8989787101745605, + "bank_marketing_test_precision": 0.8875656723976135, + "bank_marketing_test_recall": 0.8989787101745605, + "bank_marketing_test_f1-score": 0.890681803226471, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9170330762863159, + "bank_marketing_test_score": 0.8989787101745605 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.23018290102481842, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.8980790972709656, + "bank_marketing_val_precision": 0.8855907320976257, + "bank_marketing_val_recall": 0.8980790972709656, + "bank_marketing_val_f1-score": 0.8891089558601379, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9050434231758118, + "bank_marketing_val_score": 0.8980790972709656 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2254604995250702, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.8990896940231323, + "bank_marketing_test_precision": 0.8864827156066895, + "bank_marketing_test_recall": 0.8990896940231323, + "bank_marketing_test_f1-score": 0.8892030119895935, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9126611351966858, + "bank_marketing_test_score": 0.8990896940231323 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.22384144365787506, + "std": 0.001975559631807319, + "n": 5, + "ci95": 0.0024525880219567844 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_accuracy": { + "mean": 0.899333918094635, + "std": 0.00030802675151808315, + "n": 5, + "ci95": 0.00038240441293304966 + }, + "bank_marketing_test_precision": { + "mean": 0.8877177357673645, + "std": 0.0007464580322044123, + "n": 5, + "ci95": 0.000926701476990157 + }, + "bank_marketing_test_recall": { + "mean": 0.899333918094635, + "std": 0.00030802675151808315, + "n": 5, + "ci95": 0.00038240441293304966 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8907055497169495, + "std": 0.0008747684263296383, + "n": 5, + "ci95": 0.00108599433287636 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9149183511734009, + "std": 0.0020813139907728336, + "n": 5, + "ci95": 0.002583878350981809 + }, + "bank_marketing_test_score": { + "mean": 0.899333918094635, + "std": 0.00030802675151808315, + "n": 5, + "ci95": 0.00038240441293304966 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/downstream_results/mlp_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/downstream_results/mlp_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..e5df936e1460bce04ab73afa1c9b0c080e301fb1 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/downstream_results/mlp_bank_marketing_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "bank_marketing", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpybzzoq_0", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpybzzoq_0/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.22878247499465942, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8960452079772949, + "bank_marketing_val_precision": 0.8852113485336304, + "bank_marketing_val_recall": 0.8960452079772949, + "bank_marketing_val_f1-score": 0.8889473080635071, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9066123366355896, + "bank_marketing_val_score": 0.8960452079772949 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.22409798204898834, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8997557759284973, + "bank_marketing_test_precision": 0.8881022930145264, + "bank_marketing_test_recall": 0.8997557759284973, + "bank_marketing_test_f1-score": 0.8910243511199951, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9143516421318054, + "bank_marketing_test_score": 0.8997557759284973 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.2348724901676178, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8992090225219727, + "bank_marketing_val_precision": 0.8852483034133911, + "bank_marketing_val_recall": 0.8992090225219727, + "bank_marketing_val_f1-score": 0.8880388140678406, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9056040644645691, + "bank_marketing_val_score": 0.8992090225219727 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23286563158035278, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.9004218578338623, + "bank_marketing_test_precision": 0.8867059350013733, + "bank_marketing_test_recall": 0.9004218578338623, + "bank_marketing_test_f1-score": 0.8875302076339722, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9117294549942017, + "bank_marketing_test_score": 0.9004218578338623 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.2271963208913803, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8962711691856384, + "bank_marketing_val_precision": 0.8812652826309204, + "bank_marketing_val_recall": 0.8962711691856384, + "bank_marketing_val_f1-score": 0.8845623731613159, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9070462584495544, + "bank_marketing_val_score": 0.8962711691856384 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2275678962469101, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.899533748626709, + "bank_marketing_test_precision": 0.8854601979255676, + "bank_marketing_test_recall": 0.899533748626709, + "bank_marketing_test_f1-score": 0.886457085609436, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9106369018554688, + "bank_marketing_test_score": 0.899533748626709 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.23374688625335693, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8940113186836243, + "bank_marketing_val_precision": 0.8864773511886597, + "bank_marketing_val_recall": 0.8940113186836243, + "bank_marketing_val_f1-score": 0.8895831108093262, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9032974243164062, + "bank_marketing_val_score": 0.8940113186836243 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.2276753932237625, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8976465463638306, + "bank_marketing_test_precision": 0.8892873525619507, + "bank_marketing_test_recall": 0.8976465463638306, + "bank_marketing_test_f1-score": 0.8923989534378052, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9116812348365784, + "bank_marketing_test_score": 0.8976465463638306 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.23501098155975342, + "lr": 0.0009999999310821295, + "bank_marketing_val_accuracy": 0.8969491720199585, + "bank_marketing_val_precision": 0.886930525302887, + "bank_marketing_val_recall": 0.8969491720199585, + "bank_marketing_val_f1-score": 0.8904969692230225, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9005349278450012, + "bank_marketing_val_score": 0.8969491720199585 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23474524915218353, + "lr": 0.0009999994654208422, + "bank_marketing_test_accuracy": 0.8955373167991638, + "bank_marketing_test_precision": 0.8837233781814575, + "bank_marketing_test_recall": 0.8955373167991638, + "bank_marketing_test_f1-score": 0.8872881531715393, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9056728482246399, + "bank_marketing_test_score": 0.8955373167991638 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.22939043045043944, + "std": 0.00433082266768808, + "n": 5, + "ci95": 0.005376564508090068 + }, + "lr": { + "mean": 0.0009999994654208422, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_accuracy": { + "mean": 0.8985790491104126, + "std": 0.001988327059849344, + "n": 5, + "ci95": 0.002468438336259076 + }, + "bank_marketing_test_precision": { + "mean": 0.8866558313369751, + "std": 0.00218222859040285, + "n": 5, + "ci95": 0.0027091602884684127 + }, + "bank_marketing_test_recall": { + "mean": 0.8985790491104126, + "std": 0.001988327059849344, + "n": 5, + "ci95": 0.002468438336259076 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8889397501945495, + "std": 0.0026071945887865254, + "n": 5, + "ci95": 0.003236740676624655 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9108144164085388, + "std": 0.003184645601319443, + "n": 5, + "ci95": 0.003953625863891521 + }, + "bank_marketing_test_score": { + "mean": 0.8985790491104126, + "std": 0.001988327059849344, + "n": 5, + "ci95": 0.002468438336259076 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/downstream_results/resnet_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/downstream_results/resnet_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..e5fbe8e3f3c710bcb878efc9dacdb8b5fca09b3a --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/downstream_results/resnet_bank_marketing_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "bank_marketing", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpwyac3ff4", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpwyac3ff4/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.22161750495433807, + "lr": 0.0004999999655410647, + "bank_marketing_val_accuracy": 0.8985310792922974, + "bank_marketing_val_precision": 0.8907105326652527, + "bank_marketing_val_recall": 0.8985310792922974, + "bank_marketing_val_f1-score": 0.8937684893608093, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9177457094192505, + "bank_marketing_val_score": 0.8985310792922974 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21751359105110168, + "lr": 0.0004999997327104211, + "bank_marketing_test_accuracy": 0.8976465463638306, + "bank_marketing_test_precision": 0.8898050785064697, + "bank_marketing_test_recall": 0.8976465463638306, + "bank_marketing_test_f1-score": 0.8928095698356628, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9234350323677063, + "bank_marketing_test_score": 0.8976465463638306 + } + } + }, + { + "seed": 1, + "val_metrics": { + "bank_marketing_val_loss": 0.21663014590740204, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.902824878692627, + "bank_marketing_val_precision": 0.8963163495063782, + "bank_marketing_val_recall": 0.902824878692627, + "bank_marketing_val_f1-score": 0.8989129066467285, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9214954376220703, + "bank_marketing_val_score": 0.902824878692627 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21422654390335083, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.897535502910614, + "bank_marketing_test_precision": 0.891129195690155, + "bank_marketing_test_recall": 0.897535502910614, + "bank_marketing_test_f1-score": 0.8937614560127258, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9259956479072571, + "bank_marketing_test_score": 0.897535502910614 + } + } + }, + { + "seed": 2, + "val_metrics": { + "bank_marketing_val_loss": 0.2144278883934021, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9012429118156433, + "bank_marketing_val_precision": 0.8905249238014221, + "bank_marketing_val_recall": 0.9012429118156433, + "bank_marketing_val_f1-score": 0.8937513828277588, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9210366010665894, + "bank_marketing_val_score": 0.9012429118156433 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21546624600887299, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.9020870327949524, + "bank_marketing_test_precision": 0.891541600227356, + "bank_marketing_test_recall": 0.9020870327949524, + "bank_marketing_test_f1-score": 0.8943271636962891, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9230127930641174, + "bank_marketing_test_score": 0.9020870327949524 + } + } + }, + { + "seed": 3, + "val_metrics": { + "bank_marketing_val_loss": 0.21811479330062866, + "lr": 0.00020610730280168355, + "bank_marketing_val_accuracy": 0.9021468758583069, + "bank_marketing_val_precision": 0.8927248120307922, + "bank_marketing_val_recall": 0.9021468758583069, + "bank_marketing_val_f1-score": 0.8958552479743958, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.919053316116333, + "bank_marketing_val_score": 0.9021468758583069 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21096403896808624, + "lr": 0.00020610728824976832, + "bank_marketing_test_accuracy": 0.8974245190620422, + "bank_marketing_test_precision": 0.8871931433677673, + "bank_marketing_test_recall": 0.8974245190620422, + "bank_marketing_test_f1-score": 0.8905525803565979, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9277135729789734, + "bank_marketing_test_score": 0.8974245190620422 + } + } + }, + { + "seed": 4, + "val_metrics": { + "bank_marketing_val_loss": 0.21823810040950775, + "lr": 0.0, + "bank_marketing_val_accuracy": 0.9019209146499634, + "bank_marketing_val_precision": 0.8931270837783813, + "bank_marketing_val_recall": 0.9019209146499634, + "bank_marketing_val_f1-score": 0.89621502161026, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9197729825973511, + "bank_marketing_val_score": 0.9019209146499634 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.21206411719322205, + "lr": 0.0, + "bank_marketing_test_accuracy": 0.9003108143806458, + "bank_marketing_test_precision": 0.8914399147033691, + "bank_marketing_test_recall": 0.9003108143806458, + "bank_marketing_test_f1-score": 0.8944819569587708, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9269863367080688, + "bank_marketing_test_score": 0.9003108143806458 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.21404690742492677, + "std": 0.0026220326880768605, + "n": 5, + "ci95": 0.0032551616566862842 + }, + "lr": { + "mean": 0.00014122140419203788, + "std": 0.00021952364449547967, + "n": 5, + "ci95": 0.0002725309083853685 + }, + "bank_marketing_test_accuracy": { + "mean": 0.899000883102417, + "std": 0.0021039683085908433, + "n": 5, + "ci95": 0.002612002892317584 + }, + "bank_marketing_test_precision": { + "mean": 0.8902217864990234, + "std": 0.0018299904922861267, + "n": 5, + "ci95": 0.0022718690387339823 + }, + "bank_marketing_test_recall": { + "mean": 0.899000883102417, + "std": 0.0021039683085908433, + "n": 5, + "ci95": 0.002612002892317584 + }, + "bank_marketing_test_f1-score": { + "mean": 0.8931865453720093, + "std": 0.0016116378193488884, + "n": 5, + "ci95": 0.002000791850485205 + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9254286766052247, + "std": 0.0021082945431126292, + "n": 5, + "ci95": 0.0026173737608033913 + }, + "bank_marketing_test_score": { + "mean": 0.899000883102417, + "std": 0.0021039683085908433, + "n": 5, + "ci95": 0.002612002892317584 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/downstream_results/xgboost_bank_marketing_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/downstream_results/xgboost_bank_marketing_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..74f83254a173ce7ab5b840057a8a34b68e657d8e --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/downstream_results/xgboost_bank_marketing_downstream_results.json @@ -0,0 +1,107 @@ +{ + "dataset": "bank_marketing", + "model": "xgboost", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpl8bjxmwc", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "bank_marketing", + "model_name": "xgboost", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 1, + "split_config_file": "/tmp/tmpl8bjxmwc/split_data/split_config.json", + "sigma_aggregation": "mean", + "n_estimators": 100, + "max_depth": 3, + "learning_rate": 0.1, + "subsample": 1.0, + "colsample_bytree": 1.0, + "reg_alpha": 0.0, + "reg_lambda": 1.0, + "n_jobs": -1, + "random_state": 42, + "encoder_type": "linear_flatten" + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "bank_marketing_val_loss": 0.24640412628650665, + "bank_marketing_val_accuracy": 0.9021468758583069, + "bank_marketing_val_precision": 0.8922626376152039, + "bank_marketing_val_recall": 0.9021468758583069, + "bank_marketing_val_f1-score": 0.8954094052314758, + "bank_marketing_val_support": 4425.0, + "bank_marketing_val_roc_auc": 0.9228941202163696, + "bank_marketing_val_score": 0.9021468758583069 + }, + "test_metrics": { + "0": { + "bank_marketing_test_loss": 0.23810569941997528, + "bank_marketing_test_accuracy": 0.9023090600967407, + "bank_marketing_test_precision": 0.8922731876373291, + "bank_marketing_test_recall": 0.9023090600967407, + "bank_marketing_test_f1-score": 0.8951188921928406, + "bank_marketing_test_support": 9008.0, + "bank_marketing_test_roc_auc": 0.9287988543510437, + "bank_marketing_test_score": 0.9023090600967407 + } + } + } + ], + "summary": { + "0": { + "bank_marketing_test_loss": { + "mean": 0.23810569941997528, + "std": 0.0, + "n": 1, + "ci95": null + }, + "bank_marketing_test_accuracy": { + "mean": 0.9023090600967407, + "std": 0.0, + "n": 1, + "ci95": null + }, + "bank_marketing_test_precision": { + "mean": 0.8922731876373291, + "std": 0.0, + "n": 1, + "ci95": null + }, + "bank_marketing_test_recall": { + "mean": 0.9023090600967407, + "std": 0.0, + "n": 1, + "ci95": null + }, + "bank_marketing_test_f1-score": { + "mean": 0.8951188921928406, + "std": 0.0, + "n": 1, + "ci95": null + }, + "bank_marketing_test_support": { + "mean": 9008.0, + "std": 0.0, + "n": 1, + "ci95": null + }, + "bank_marketing_test_roc_auc": { + "mean": 0.9287988543510437, + "std": 0.0, + "n": 1, + "ci95": null + }, + "bank_marketing_test_score": { + "mean": 0.9023090600967407, + "std": 0.0, + "n": 1, + "ci95": null + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..25c9937fce5d15911c874c145be8b271ca5cd6cb --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c08565ef37094054d3145b73608814d01dc476ec3d88dcdb4ac4dc8786c1f071 +size 185184384 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..88e78aeed7c3691d091aa9f224477a48e321bd02 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e0e4c2164f00c59b67d834391607bc2ad3e15bf25fbbed4ae0b2a261cc1f2061 +size 15468526 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fede1e0409ba3531e0f82dc97e7d1a6dec1f1561 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: bank_marketing +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.001 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.3 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.1 +mask_min_trgt_share: 0.1 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 256 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.1 +recon_y_weight: 1.0 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.004 +sw_s_y_weight: 0.0004 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 15 +vtjepa_kl_x_end_weight: 0.0001 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 15 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 15 +wjepa_joint_end_weight: 0.016 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..aa9b12846199b72a9b18cf408dea633cb5a25f6e --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-15T21:47:41.262447+00:00", + "spec": { + "dataset": "covertype", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.0001875, + "sw_z_weight": 0.0, + "sw_s_y_weight": 1.875e-05, + "lambda_joint": 0.00046875, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..9293c4f323dd9d70c32ad8560a9c3e10b3d2aeb7 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "covertype", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.0001875, + "sw_z_weight": 0.0, + "sw_s_y_weight": 1.875e-05, + "lambda_joint": 0.00046875, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "covertype", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.00046875", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.0001875", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "1.875e-05", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..aa3d0d5e3dd1b92d56fffce0d3e6441f78c52853 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "covertype", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.00046875, + "s_x": 0.0001875, + "z": 0.0, + "s_y": 1.875e-05 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 60 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "covertype", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "64", + "--model_num_layers", + "8", + "--model_num_heads", + "8", + "--model_dropout_prob", + "0.0015", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.15", + "--mask_max_ctx_share", + "0.35", + "--mask_min_trgt_share", + "0.15", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.0005", + "--exp_warmup", + "12", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "1e-06", + "--vtjepa_kl_x_annealing_epochs", + "60", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-05", + "--vtjepa_kl_z_annealing_epochs", + "60", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-06", + "--vtjepa_kl_y_annealing_epochs", + "20", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.001", + "--recon_y_weight", + "0.5", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.00046875", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "60", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.0001875", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "1.875e-05", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/downstream_results/autoint_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/downstream_results/autoint_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..56a3860403bfe1654f35f5db73dedbbd88b2e2da --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/downstream_results/autoint_covertype_downstream_results.json @@ -0,0 +1,210 @@ +{ + "dataset": "covertype", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmppsxlt28g", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmppsxlt28g/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.6174989938735962, + "lr": 0.0, + "covertype_val_accuracy": 0.7497497200965881, + "covertype_val_precision": 0.7482361793518066, + "covertype_val_recall": 0.7497497200965881, + "covertype_val_f1-score": 0.7464077472686768, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7497497200965881 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5891000628471375, + "lr": 0.0, + "covertype_test_accuracy": 0.75, + "covertype_test_precision": 0.7465114593505859, + "covertype_test_recall": 0.75, + "covertype_test_f1-score": 0.7465739846229553, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.75 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.6075288653373718, + "lr": 0.0, + "covertype_val_accuracy": 0.7657657861709595, + "covertype_val_precision": 0.7631065249443054, + "covertype_val_recall": 0.7657657861709595, + "covertype_val_f1-score": 0.7613555192947388, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7657657861709595 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5870580673217773, + "lr": 0.0, + "covertype_test_accuracy": 0.7484999895095825, + "covertype_test_precision": 0.7495322823524475, + "covertype_test_recall": 0.7484999895095825, + "covertype_test_f1-score": 0.7475013136863708, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7484999895095825 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.6121671199798584, + "lr": 0.0, + "covertype_val_accuracy": 0.7487487196922302, + "covertype_val_precision": 0.7524433732032776, + "covertype_val_recall": 0.7487487196922302, + "covertype_val_f1-score": 0.7478883266448975, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7487487196922302 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.588174045085907, + "lr": 0.0, + "covertype_test_accuracy": 0.7480000257492065, + "covertype_test_precision": 0.7490994334220886, + "covertype_test_recall": 0.7480000257492065, + "covertype_test_f1-score": 0.7464427351951599, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7480000257492065 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.617547333240509, + "lr": 0.0, + "covertype_val_accuracy": 0.7497497200965881, + "covertype_val_precision": 0.7511450052261353, + "covertype_val_recall": 0.7497497200965881, + "covertype_val_f1-score": 0.746807873249054, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7497497200965881 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5924734473228455, + "lr": 0.0, + "covertype_test_accuracy": 0.7515000104904175, + "covertype_test_precision": 0.7538991570472717, + "covertype_test_recall": 0.7515000104904175, + "covertype_test_f1-score": 0.7505937814712524, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7515000104904175 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.6094942092895508, + "lr": 0.0, + "covertype_val_accuracy": 0.7577577829360962, + "covertype_val_precision": 0.7577868103981018, + "covertype_val_recall": 0.7577577829360962, + "covertype_val_f1-score": 0.7525967359542847, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7577577829360962 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5996159315109253, + "lr": 0.0, + "covertype_test_accuracy": 0.7605000138282776, + "covertype_test_precision": 0.7603112459182739, + "covertype_test_recall": 0.7605000138282776, + "covertype_test_f1-score": 0.757434606552124, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7605000138282776 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5912843108177185, + "std": 0.005077789263295361, + "n": 5, + "ci95": 0.0063038973487152174 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.7517000079154968, + "std": 0.005106371406802072, + "n": 5, + "ci95": 0.006339381077820512 + }, + "covertype_test_precision": { + "mean": 0.7518707156181336, + "std": 0.005413571586034099, + "n": 5, + "ci95": 0.006720759330239132 + }, + "covertype_test_recall": { + "mean": 0.7517000079154968, + "std": 0.005106371406802072, + "n": 5, + "ci95": 0.006339381077820512 + }, + "covertype_test_f1-score": { + "mean": 0.7497092843055725, + "std": 0.004632454006651714, + "n": 5, + "ci95": 0.005751029240552847 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.7517000079154968, + "std": 0.005106371406802072, + "n": 5, + "ci95": 0.006339381077820512 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/downstream_results/dcnv2_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/downstream_results/dcnv2_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..1b511a34439fcb9644b349736e3fe0d6d55f67d4 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/downstream_results/dcnv2_covertype_downstream_results.json @@ -0,0 +1,209 @@ +{ + "dataset": "covertype", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpz8ncm5ed", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpz8ncm5ed/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.5569818615913391, + "lr": 0.0, + "covertype_val_accuracy": 0.7907907962799072, + "covertype_val_precision": 0.7933158874511719, + "covertype_val_recall": 0.7907907962799072, + "covertype_val_f1-score": 0.7864508032798767, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7907907962799072 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5536759495735168, + "lr": 0.0, + "covertype_test_accuracy": 0.7804999947547913, + "covertype_test_precision": 0.7802493572235107, + "covertype_test_recall": 0.7804999947547913, + "covertype_test_f1-score": 0.7773028016090393, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7804999947547913 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.5788687467575073, + "lr": 0.0, + "covertype_val_accuracy": 0.7677677869796753, + "covertype_val_precision": 0.7668536305427551, + "covertype_val_recall": 0.7677677869796753, + "covertype_val_f1-score": 0.7628348469734192, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7677677869796753 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5547317862510681, + "lr": 0.0, + "covertype_test_accuracy": 0.7730000019073486, + "covertype_test_precision": 0.7746561169624329, + "covertype_test_recall": 0.7730000019073486, + "covertype_test_f1-score": 0.7703948020935059, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7730000019073486 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.5738442540168762, + "lr": 0.0, + "covertype_val_accuracy": 0.7547547817230225, + "covertype_val_precision": 0.7464178800582886, + "covertype_val_recall": 0.7547547817230225, + "covertype_val_f1-score": 0.748012900352478, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7547547817230225 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5687882900238037, + "lr": 0.0, + "covertype_test_accuracy": 0.7705000042915344, + "covertype_test_precision": 0.7723273038864136, + "covertype_test_recall": 0.7705000042915344, + "covertype_test_f1-score": 0.7678394317626953, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7705000042915344 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.572620153427124, + "lr": 0.0, + "covertype_val_accuracy": 0.7677677869796753, + "covertype_val_precision": 0.7605739235877991, + "covertype_val_recall": 0.7677677869796753, + "covertype_val_f1-score": 0.7618176341056824, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7677677869796753 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5768039226531982, + "lr": 0.0, + "covertype_test_accuracy": 0.7695000171661377, + "covertype_test_precision": 0.7714239954948425, + "covertype_test_recall": 0.7695000171661377, + "covertype_test_f1-score": 0.7670117616653442, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7695000171661377 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.5737892985343933, + "lr": 0.0, + "covertype_val_accuracy": 0.759759783744812, + "covertype_val_precision": 0.7503368258476257, + "covertype_val_recall": 0.759759783744812, + "covertype_val_f1-score": 0.7509111166000366, + "covertype_val_support": 999.0, + "covertype_val_score": 0.759759783744812 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5549345016479492, + "lr": 0.0, + "covertype_test_accuracy": 0.7735000252723694, + "covertype_test_precision": 0.7730095982551575, + "covertype_test_recall": 0.7735000252723694, + "covertype_test_f1-score": 0.7708216905593872, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7735000252723694 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5617868900299072, + "std": 0.010452845804349329, + "n": 5, + "ci95": 0.012976841600906322 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.7734000086784363, + "std": 0.004306964425801064, + "n": 5, + "ci95": 0.005346945337230865 + }, + "covertype_test_precision": { + "mean": 0.7743332743644714, + "std": 0.0035122714616240397, + "n": 5, + "ci95": 0.004360361883259987 + }, + "covertype_test_recall": { + "mean": 0.7734000086784363, + "std": 0.004306964425801064, + "n": 5, + "ci95": 0.005346945337230865 + }, + "covertype_test_f1-score": { + "mean": 0.7706740975379944, + "std": 0.004046218648581519, + "n": 5, + "ci95": 0.005023238596270874 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.7734000086784363, + "std": 0.004306964425801064, + "n": 5, + "ci95": 0.005346945337230865 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/downstream_results/fttransformer_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/downstream_results/fttransformer_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..f6f3790a4811e0a441323a30c552bcdf9a5894eb --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/downstream_results/fttransformer_covertype_downstream_results.json @@ -0,0 +1,209 @@ +{ + "dataset": "covertype", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp8bx0f_7f", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp8bx0f_7f/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.6029737591743469, + "lr": 0.0, + "covertype_val_accuracy": 0.7497497200965881, + "covertype_val_precision": 0.7537214159965515, + "covertype_val_recall": 0.7497497200965881, + "covertype_val_f1-score": 0.7467451095581055, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7497497200965881 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5759369730949402, + "lr": 0.0, + "covertype_test_accuracy": 0.7509999871253967, + "covertype_test_precision": 0.754407525062561, + "covertype_test_recall": 0.7509999871253967, + "covertype_test_f1-score": 0.7480729222297668, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7509999871253967 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.59492027759552, + "lr": 0.0, + "covertype_val_accuracy": 0.7497497200965881, + "covertype_val_precision": 0.7314669489860535, + "covertype_val_recall": 0.7497497200965881, + "covertype_val_f1-score": 0.7380244135856628, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7497497200965881 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5965020060539246, + "lr": 0.0, + "covertype_test_accuracy": 0.7369999885559082, + "covertype_test_precision": 0.7223562002182007, + "covertype_test_recall": 0.7369999885559082, + "covertype_test_f1-score": 0.7259088754653931, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7369999885559082 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.5964241623878479, + "lr": 0.0, + "covertype_val_accuracy": 0.7487487196922302, + "covertype_val_precision": 0.7512862682342529, + "covertype_val_recall": 0.7487487196922302, + "covertype_val_f1-score": 0.7403479218482971, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7487487196922302 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5760630965232849, + "lr": 0.0, + "covertype_test_accuracy": 0.7459999918937683, + "covertype_test_precision": 0.7482266426086426, + "covertype_test_recall": 0.7459999918937683, + "covertype_test_f1-score": 0.7409065961837769, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7459999918937683 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.6123899817466736, + "lr": 0.0, + "covertype_val_accuracy": 0.7407407164573669, + "covertype_val_precision": 0.7363642454147339, + "covertype_val_recall": 0.7407407164573669, + "covertype_val_f1-score": 0.732924222946167, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7407407164573669 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5820234417915344, + "lr": 0.0, + "covertype_test_accuracy": 0.7459999918937683, + "covertype_test_precision": 0.7424235343933105, + "covertype_test_recall": 0.7459999918937683, + "covertype_test_f1-score": 0.7375316619873047, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7459999918937683 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.6064786314964294, + "lr": 0.0, + "covertype_val_accuracy": 0.7547547817230225, + "covertype_val_precision": 0.7537180185317993, + "covertype_val_recall": 0.7547547817230225, + "covertype_val_f1-score": 0.7471916675567627, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7547547817230225 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5870689153671265, + "lr": 0.0, + "covertype_test_accuracy": 0.7544999718666077, + "covertype_test_precision": 0.7569332122802734, + "covertype_test_recall": 0.7544999718666077, + "covertype_test_f1-score": 0.7506306767463684, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7544999718666077 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5835188865661621, + "std": 0.008609219886896228, + "n": 5, + "ci95": 0.010688044659870442 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.7468999862670899, + "std": 0.006599237125918761, + "n": 5, + "ci95": 0.008192721529885691 + }, + "covertype_test_precision": { + "mean": 0.7448694229125976, + "std": 0.013789569403670538, + "n": 5, + "ci95": 0.01711926696763091 + }, + "covertype_test_recall": { + "mean": 0.7468999862670899, + "std": 0.006599237125918761, + "n": 5, + "ci95": 0.008192721529885691 + }, + "covertype_test_f1-score": { + "mean": 0.740610146522522, + "std": 0.009769805107499205, + "n": 5, + "ci95": 0.012128870522416997 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.7468999862670899, + "std": 0.006599237125918761, + "n": 5, + "ci95": 0.008192721529885691 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/downstream_results/mlp_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/downstream_results/mlp_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..d4f33c2a90512f77f2f17d1b07cdeba253934306 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/downstream_results/mlp_covertype_downstream_results.json @@ -0,0 +1,205 @@ +{ + "dataset": "covertype", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpvqvetdym", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpvqvetdym/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.6495128870010376, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.7307307124137878, + "covertype_val_precision": 0.7233118414878845, + "covertype_val_recall": 0.7307307124137878, + "covertype_val_f1-score": 0.7223528027534485, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7307307124137878 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.6471050381660461, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.7174999713897705, + "covertype_test_precision": 0.7051432132720947, + "covertype_test_recall": 0.7174999713897705, + "covertype_test_f1-score": 0.708031952381134, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7174999713897705 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.7548342347145081, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.6656656861305237, + "covertype_val_precision": 0.651267409324646, + "covertype_val_recall": 0.6656656861305237, + "covertype_val_f1-score": 0.625358521938324, + "covertype_val_support": 999.0, + "covertype_val_score": 0.6656656861305237 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.7665451169013977, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.6610000133514404, + "covertype_test_precision": 0.6602426171302795, + "covertype_test_recall": 0.6610000133514404, + "covertype_test_f1-score": 0.6204760074615479, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.6610000133514404 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.6595560312271118, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.7387387156486511, + "covertype_val_precision": 0.7388865351676941, + "covertype_val_recall": 0.7387387156486511, + "covertype_val_f1-score": 0.7338352203369141, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7387387156486511 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.6364644765853882, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.7325000166893005, + "covertype_test_precision": 0.7301033139228821, + "covertype_test_recall": 0.7325000166893005, + "covertype_test_f1-score": 0.7284690141677856, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7325000166893005 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 1.1139482259750366, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.5985985994338989, + "covertype_val_precision": 0.6513458490371704, + "covertype_val_recall": 0.5985985994338989, + "covertype_val_f1-score": 0.5387859344482422, + "covertype_val_support": 999.0, + "covertype_val_score": 0.5985985994338989 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 1.1363834142684937, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.5835000276565552, + "covertype_test_precision": 0.6360043287277222, + "covertype_test_recall": 0.5835000276565552, + "covertype_test_f1-score": 0.5232917666435242, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.5835000276565552 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.7670564651489258, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.7107107043266296, + "covertype_val_precision": 0.7094417810440063, + "covertype_val_recall": 0.7107107043266296, + "covertype_val_f1-score": 0.6902990937232971, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7107107043266296 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.6963205337524414, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.7095000147819519, + "covertype_test_precision": 0.7044821977615356, + "covertype_test_recall": 0.7095000147819519, + "covertype_test_f1-score": 0.688781201839447, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7095000147819519 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.7765637159347534, + "std": 0.20760352961728, + "n": 5, + "ci95": 0.2577325036701052 + }, + "lr": { + "mean": 0.0010000001639127731, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.6808000087738038, + "std": 0.060627540283863704, + "n": 5, + "ci95": 0.07526696572802266 + }, + "covertype_test_precision": { + "mean": 0.6871951341629028, + "std": 0.0381084791660073, + "n": 5, + "ci95": 0.04731034084353824 + }, + "covertype_test_recall": { + "mean": 0.6808000087738038, + "std": 0.060627540283863704, + "n": 5, + "ci95": 0.07526696572802266 + }, + "covertype_test_f1-score": { + "mean": 0.6538099884986878, + "std": 0.08349122374787564, + "n": 5, + "ci95": 0.10365142717318153 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.6808000087738038, + "std": 0.060627540283863704, + "n": 5, + "ci95": 0.07526696572802266 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/downstream_results/resnet_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/downstream_results/resnet_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..fd64b9483a1ce6ec97d9bcd771f2da7c3f464302 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/downstream_results/resnet_covertype_downstream_results.json @@ -0,0 +1,208 @@ +{ + "dataset": "covertype", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpojlty7gh", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpojlty7gh/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.594020664691925, + "lr": 0.0, + "covertype_val_accuracy": 0.7557557821273804, + "covertype_val_precision": 0.7510254383087158, + "covertype_val_recall": 0.7557557821273804, + "covertype_val_f1-score": 0.7518117427825928, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7557557821273804 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5618748664855957, + "lr": 0.0, + "covertype_test_accuracy": 0.7804999947547913, + "covertype_test_precision": 0.7801459431648254, + "covertype_test_recall": 0.7804999947547913, + "covertype_test_f1-score": 0.7795448303222656, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7804999947547913 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.6044595837593079, + "lr": 0.0, + "covertype_val_accuracy": 0.7577577829360962, + "covertype_val_precision": 0.7454033493995667, + "covertype_val_recall": 0.7577577829360962, + "covertype_val_f1-score": 0.7495087385177612, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7577577829360962 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5723586678504944, + "lr": 0.0, + "covertype_test_accuracy": 0.7774999737739563, + "covertype_test_precision": 0.778622567653656, + "covertype_test_recall": 0.7774999737739563, + "covertype_test_f1-score": 0.7760329842567444, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7774999737739563 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.6160089373588562, + "lr": 0.0, + "covertype_val_accuracy": 0.7647647857666016, + "covertype_val_precision": 0.7711515426635742, + "covertype_val_recall": 0.7647647857666016, + "covertype_val_f1-score": 0.7610491514205933, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7647647857666016 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5762115120887756, + "lr": 0.0, + "covertype_test_accuracy": 0.7649999856948853, + "covertype_test_precision": 0.7742794156074524, + "covertype_test_recall": 0.7649999856948853, + "covertype_test_f1-score": 0.7635968923568726, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7649999856948853 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.5927913784980774, + "lr": 0.0, + "covertype_val_accuracy": 0.7677677869796753, + "covertype_val_precision": 0.763477087020874, + "covertype_val_recall": 0.7677677869796753, + "covertype_val_f1-score": 0.7646081447601318, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7677677869796753 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5668526291847229, + "lr": 0.0, + "covertype_test_accuracy": 0.7634999752044678, + "covertype_test_precision": 0.7612871527671814, + "covertype_test_recall": 0.7634999752044678, + "covertype_test_f1-score": 0.7618939280509949, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7634999752044678 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.5726922154426575, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.7737737894058228, + "covertype_val_precision": 0.7696520686149597, + "covertype_val_recall": 0.7737737894058228, + "covertype_val_f1-score": 0.770448625087738, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7737737894058228 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5555123090744019, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.7664999961853027, + "covertype_test_precision": 0.7682499885559082, + "covertype_test_recall": 0.7664999961853027, + "covertype_test_f1-score": 0.7661905288696289, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7664999961853027 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5665619969367981, + "std": 0.008228970123539854, + "n": 5, + "ci95": 0.010215977909798935 + }, + "lr": { + "mean": 0.00020000003278255462, + "std": 0.0004472136688039785, + "n": 5, + "ci95": 0.0005552000910043715 + }, + "covertype_test_accuracy": { + "mean": 0.7705999851226807, + "std": 0.007813451187219813, + "n": 5, + "ci95": 0.009700125717991165 + }, + "covertype_test_precision": { + "mean": 0.7725170135498047, + "std": 0.007792651899296463, + "n": 5, + "ci95": 0.009674304131234317 + }, + "covertype_test_recall": { + "mean": 0.7705999851226807, + "std": 0.007813451187219813, + "n": 5, + "ci95": 0.009700125717991165 + }, + "covertype_test_f1-score": { + "mean": 0.7694518327713012, + "std": 0.007861591896602849, + "n": 5, + "ci95": 0.009759890720930267 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.7705999851226807, + "std": 0.007813451187219813, + "n": 5, + "ci95": 0.009700125717991165 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..aee27cc29f063e2743b2b29708c347f8ae2343c1 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9390445ccb397d5894fa3c1f585c085353dcbe70e6092d60a84efd38a5427c6a +size 138240128 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..22a31a507516e00c753e0e43b194c67c6da37f29 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d5d8173c592d2fcd5cf843823feec4c397522c64b0b6bcb08e28820289d4f4a +size 10878670 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..aaa7643e19a6d39de4c268b2d0086d418633d49c --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00046875_sx0.0001875_sy1.875e-05/seed_42/wjepa_deterministic/covertype/20260915_211751/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 +data_set: covertype +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.0005 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 12 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.35 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.15 +mask_min_trgt_share: 0.15 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 64 +model_dim_hidden: 64 +model_dropout_prob: 0.0015 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 8 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.001 +recon_y_weight: 0.5 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.0001875 +sw_s_y_weight: 1.875e-05 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 60 +vtjepa_kl_x_end_weight: 1.0e-06 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 20 +vtjepa_kl_y_end_weight: 1.0e-06 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 60 +vtjepa_kl_z_end_weight: 1.0e-05 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 60 +wjepa_joint_end_weight: 0.00046875 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..d11b7b85d44cd609d924dddc5e5c185a32ffa73e --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-15T21:18:10.451371+00:00", + "spec": { + "dataset": "covertype", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.000375, + "sw_z_weight": 0.0, + "sw_s_y_weight": 3.75e-05, + "lambda_joint": 0.0009375, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..3864117623f7f300d8a9ee278996f6abc50554a9 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "covertype", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.000375, + "sw_z_weight": 0.0, + "sw_s_y_weight": 3.75e-05, + "lambda_joint": 0.0009375, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "covertype", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.0009375", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.000375", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "3.75e-05", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..ab088f699faf8f2b79714a4c7788ec41364a374c --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "covertype", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.0009375, + "s_x": 0.000375, + "z": 0.0, + "s_y": 3.75e-05 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 60 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "covertype", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "64", + "--model_num_layers", + "8", + "--model_num_heads", + "8", + "--model_dropout_prob", + "0.0015", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.15", + "--mask_max_ctx_share", + "0.35", + "--mask_min_trgt_share", + "0.15", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.0005", + "--exp_warmup", + "12", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "1e-06", + "--vtjepa_kl_x_annealing_epochs", + "60", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-05", + "--vtjepa_kl_z_annealing_epochs", + "60", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-06", + "--vtjepa_kl_y_annealing_epochs", + "20", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.001", + "--recon_y_weight", + "0.5", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.0009375", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "60", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.000375", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "3.75e-05", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/downstream_results/autoint_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/downstream_results/autoint_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..a6631c49e02dd5baad4b2c46e2897954a9cace0d --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/downstream_results/autoint_covertype_downstream_results.json @@ -0,0 +1,210 @@ +{ + "dataset": "covertype", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpvuv0yg6f", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpvuv0yg6f/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.6242567300796509, + "lr": 9.866357686405536e-07, + "covertype_val_accuracy": 0.7417417168617249, + "covertype_val_precision": 0.741938054561615, + "covertype_val_recall": 0.7417417168617249, + "covertype_val_f1-score": 0.7394734621047974, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7417417168617249 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.6189635992050171, + "lr": 9.866358823273913e-07, + "covertype_test_accuracy": 0.7390000224113464, + "covertype_test_precision": 0.7389342188835144, + "covertype_test_recall": 0.7390000224113464, + "covertype_test_f1-score": 0.7369086146354675, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7390000224113464 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.5961858630180359, + "lr": 0.0, + "covertype_val_accuracy": 0.7547547817230225, + "covertype_val_precision": 0.7559151649475098, + "covertype_val_recall": 0.7547547817230225, + "covertype_val_f1-score": 0.7511102557182312, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7547547817230225 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5894886255264282, + "lr": 0.0, + "covertype_test_accuracy": 0.7559999823570251, + "covertype_test_precision": 0.7539926767349243, + "covertype_test_recall": 0.7559999823570251, + "covertype_test_f1-score": 0.7516686320304871, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7559999823570251 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.5975145101547241, + "lr": 0.0, + "covertype_val_accuracy": 0.7507507801055908, + "covertype_val_precision": 0.7483989000320435, + "covertype_val_recall": 0.7507507801055908, + "covertype_val_f1-score": 0.7474316954612732, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7507507801055908 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5969483852386475, + "lr": 0.0, + "covertype_test_accuracy": 0.7505000233650208, + "covertype_test_precision": 0.7511382102966309, + "covertype_test_recall": 0.7505000233650208, + "covertype_test_f1-score": 0.7491548657417297, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7505000233650208 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.6191619038581848, + "lr": 0.0, + "covertype_val_accuracy": 0.7507507801055908, + "covertype_val_precision": 0.752962052822113, + "covertype_val_recall": 0.7507507801055908, + "covertype_val_f1-score": 0.7465463876724243, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7507507801055908 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.577721357345581, + "lr": 0.0, + "covertype_test_accuracy": 0.7609999775886536, + "covertype_test_precision": 0.7606657147407532, + "covertype_test_recall": 0.7609999775886536, + "covertype_test_f1-score": 0.7595718502998352, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7609999775886536 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.6144819259643555, + "lr": 0.0, + "covertype_val_accuracy": 0.7467467188835144, + "covertype_val_precision": 0.7443207502365112, + "covertype_val_recall": 0.7467467188835144, + "covertype_val_f1-score": 0.7437683939933777, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7467467188835144 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5981170535087585, + "lr": 0.0, + "covertype_test_accuracy": 0.7490000128746033, + "covertype_test_precision": 0.7485879063606262, + "covertype_test_recall": 0.7490000128746033, + "covertype_test_f1-score": 0.7470973134040833, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7490000128746033 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5962478041648864, + "std": 0.015074050375320855, + "n": 5, + "ci95": 0.01871390506145497 + }, + "lr": { + "mean": 1.9732717646547825e-07, + "std": 4.4123698038490607e-07, + "n": 5, + "ci95": 5.477802418681676e-07 + }, + "covertype_test_accuracy": { + "mean": 0.7511000037193298, + "std": 0.00826436263256585, + "n": 5, + "ci95": 0.010259916468932552 + }, + "covertype_test_precision": { + "mean": 0.7506637454032898, + "std": 0.007956714012116894, + "n": 5, + "ci95": 0.009877981492464967 + }, + "covertype_test_recall": { + "mean": 0.7511000037193298, + "std": 0.00826436263256585, + "n": 5, + "ci95": 0.010259916468932552 + }, + "covertype_test_f1-score": { + "mean": 0.7488802552223206, + "std": 0.008195421302947647, + "n": 5, + "ci95": 0.010174328225218191 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.7511000037193298, + "std": 0.00826436263256585, + "n": 5, + "ci95": 0.010259916468932552 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/downstream_results/dcnv2_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/downstream_results/dcnv2_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..4609783980ab6dc1e392f0f76e9fb71dfd9c8311 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/downstream_results/dcnv2_covertype_downstream_results.json @@ -0,0 +1,209 @@ +{ + "dataset": "covertype", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp91y4hjrz", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp91y4hjrz/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.5674107670783997, + "lr": 0.0, + "covertype_val_accuracy": 0.7747747898101807, + "covertype_val_precision": 0.7689729332923889, + "covertype_val_recall": 0.7747747898101807, + "covertype_val_f1-score": 0.7688935399055481, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7747747898101807 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5620741248130798, + "lr": 0.0, + "covertype_test_accuracy": 0.7754999995231628, + "covertype_test_precision": 0.7762084603309631, + "covertype_test_recall": 0.7754999995231628, + "covertype_test_f1-score": 0.7733892798423767, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7754999995231628 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.5693020224571228, + "lr": 0.0, + "covertype_val_accuracy": 0.7727727890014648, + "covertype_val_precision": 0.7732557654380798, + "covertype_val_recall": 0.7727727890014648, + "covertype_val_f1-score": 0.7680808305740356, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7727727890014648 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5725733637809753, + "lr": 0.0, + "covertype_test_accuracy": 0.7645000219345093, + "covertype_test_precision": 0.7636183500289917, + "covertype_test_recall": 0.7645000219345093, + "covertype_test_f1-score": 0.7625875473022461, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7645000219345093 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.5696365237236023, + "lr": 0.0, + "covertype_val_accuracy": 0.7647647857666016, + "covertype_val_precision": 0.7583976984024048, + "covertype_val_recall": 0.7647647857666016, + "covertype_val_f1-score": 0.7586537599563599, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7647647857666016 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5696688294410706, + "lr": 0.0, + "covertype_test_accuracy": 0.7705000042915344, + "covertype_test_precision": 0.7708288431167603, + "covertype_test_recall": 0.7705000042915344, + "covertype_test_f1-score": 0.7680527567863464, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7705000042915344 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.5802058577537537, + "lr": 0.0, + "covertype_val_accuracy": 0.7877877950668335, + "covertype_val_precision": 0.7824431657791138, + "covertype_val_recall": 0.7877877950668335, + "covertype_val_f1-score": 0.7826768755912781, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7877877950668335 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5635465979576111, + "lr": 0.0, + "covertype_test_accuracy": 0.781499981880188, + "covertype_test_precision": 0.7814568877220154, + "covertype_test_recall": 0.781499981880188, + "covertype_test_f1-score": 0.7793304920196533, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.781499981880188 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.553926944732666, + "lr": 0.0, + "covertype_val_accuracy": 0.7787787914276123, + "covertype_val_precision": 0.773069441318512, + "covertype_val_recall": 0.7787787914276123, + "covertype_val_f1-score": 0.7727245688438416, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7787787914276123 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5603642463684082, + "lr": 0.0, + "covertype_test_accuracy": 0.7770000100135803, + "covertype_test_precision": 0.7784990668296814, + "covertype_test_recall": 0.7770000100135803, + "covertype_test_f1-score": 0.7751080989837646, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7770000100135803 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.565645432472229, + "std": 0.0052257589584391965, + "n": 5, + "ci95": 0.006487596537582709 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.773800003528595, + "std": 0.006515353925832216, + "n": 5, + "ci95": 0.008088583477830306 + }, + "covertype_test_precision": { + "mean": 0.7741223216056824, + "std": 0.0070442202033421135, + "n": 5, + "ci95": 0.008745152419893078 + }, + "covertype_test_recall": { + "mean": 0.773800003528595, + "std": 0.006515353925832216, + "n": 5, + "ci95": 0.008088583477830306 + }, + "covertype_test_f1-score": { + "mean": 0.7716936349868775, + "std": 0.006500618974228052, + "n": 5, + "ci95": 0.008070290552004816 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.773800003528595, + "std": 0.006515353925832216, + "n": 5, + "ci95": 0.008088583477830306 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/downstream_results/fttransformer_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/downstream_results/fttransformer_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..12fd4696fdf5f029d2d4fc24e764b6866d036ee8 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/downstream_results/fttransformer_covertype_downstream_results.json @@ -0,0 +1,209 @@ +{ + "dataset": "covertype", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp770j91lf", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp770j91lf/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.606400728225708, + "lr": 0.0, + "covertype_val_accuracy": 0.7357357144355774, + "covertype_val_precision": 0.7300575971603394, + "covertype_val_recall": 0.7357357144355774, + "covertype_val_f1-score": 0.7245064973831177, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7357357144355774 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5987634658813477, + "lr": 0.0, + "covertype_test_accuracy": 0.7434999942779541, + "covertype_test_precision": 0.7441664934158325, + "covertype_test_recall": 0.7434999942779541, + "covertype_test_f1-score": 0.7355547547340393, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7434999942779541 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.6106162071228027, + "lr": 0.0, + "covertype_val_accuracy": 0.7567567825317383, + "covertype_val_precision": 0.7557698488235474, + "covertype_val_recall": 0.7567567825317383, + "covertype_val_f1-score": 0.7495996356010437, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7567567825317383 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5847806930541992, + "lr": 0.0, + "covertype_test_accuracy": 0.7555000185966492, + "covertype_test_precision": 0.7555152177810669, + "covertype_test_recall": 0.7555000185966492, + "covertype_test_f1-score": 0.7481939196586609, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7555000185966492 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.6126089692115784, + "lr": 0.0, + "covertype_val_accuracy": 0.7437437176704407, + "covertype_val_precision": 0.7385696172714233, + "covertype_val_recall": 0.7437437176704407, + "covertype_val_f1-score": 0.7369713187217712, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7437437176704407 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5841242074966431, + "lr": 0.0, + "covertype_test_accuracy": 0.7490000128746033, + "covertype_test_precision": 0.7502676844596863, + "covertype_test_recall": 0.7490000128746033, + "covertype_test_f1-score": 0.7450605630874634, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7490000128746033 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.6182376146316528, + "lr": 3.511175964376889e-05, + "covertype_val_accuracy": 0.7367367148399353, + "covertype_val_precision": 0.7360720634460449, + "covertype_val_recall": 0.7367367148399353, + "covertype_val_f1-score": 0.7272462844848633, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7367367148399353 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5932967066764832, + "lr": 3.511175964376889e-05, + "covertype_test_accuracy": 0.7400000095367432, + "covertype_test_precision": 0.7455840706825256, + "covertype_test_recall": 0.7400000095367432, + "covertype_test_f1-score": 0.7308873534202576, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7400000095367432 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.6032359600067139, + "lr": 0.0, + "covertype_val_accuracy": 0.7347347140312195, + "covertype_val_precision": 0.7296208739280701, + "covertype_val_recall": 0.7347347140312195, + "covertype_val_f1-score": 0.72708660364151, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7347347140312195 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5970275402069092, + "lr": 0.0, + "covertype_test_accuracy": 0.7384999990463257, + "covertype_test_precision": 0.7426486015319824, + "covertype_test_recall": 0.7384999990463257, + "covertype_test_f1-score": 0.73340904712677, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7384999990463257 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5915985226631164, + "std": 0.0068198727909423535, + "n": 5, + "ci95": 0.008466632972770503 + }, + "lr": { + "mean": 7.022351928753778e-06, + "std": 1.5702456274620208e-05, + "n": 5, + "ci95": 1.9494048954220486e-05 + }, + "covertype_test_accuracy": { + "mean": 0.745300006866455, + "std": 0.006987496102395445, + "n": 5, + "ci95": 0.008674731437251924 + }, + "covertype_test_precision": { + "mean": 0.7476364135742187, + "std": 0.005247138944746299, + "n": 5, + "ci95": 0.006514139041024343 + }, + "covertype_test_recall": { + "mean": 0.745300006866455, + "std": 0.006987496102395445, + "n": 5, + "ci95": 0.008674731437251924 + }, + "covertype_test_f1-score": { + "mean": 0.7386211276054382, + "std": 0.007574370694508879, + "n": 5, + "ci95": 0.009403315668187742 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.745300006866455, + "std": 0.006987496102395445, + "n": 5, + "ci95": 0.008674731437251924 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/downstream_results/mlp_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/downstream_results/mlp_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..f4f66e140333eff83447e7cac705eace07b0e280 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/downstream_results/mlp_covertype_downstream_results.json @@ -0,0 +1,205 @@ +{ + "dataset": "covertype", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpkbvogru0", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpkbvogru0/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.6145040392875671, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.7197197079658508, + "covertype_val_precision": 0.7057517766952515, + "covertype_val_recall": 0.7197197079658508, + "covertype_val_f1-score": 0.7098142504692078, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7197197079658508 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.6191563010215759, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.722000002861023, + "covertype_test_precision": 0.7141382694244385, + "covertype_test_recall": 0.722000002861023, + "covertype_test_f1-score": 0.7140401601791382, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.722000002861023 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.9386884570121765, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.6296296119689941, + "covertype_val_precision": 0.7361564636230469, + "covertype_val_recall": 0.6296296119689941, + "covertype_val_f1-score": 0.6404695510864258, + "covertype_val_support": 999.0, + "covertype_val_score": 0.6296296119689941 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.9165419936180115, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.6140000224113464, + "covertype_test_precision": 0.6985296010971069, + "covertype_test_recall": 0.6140000224113464, + "covertype_test_f1-score": 0.6103162169456482, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.6140000224113464 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.6616693139076233, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.7217217087745667, + "covertype_val_precision": 0.720229983329773, + "covertype_val_recall": 0.7217217087745667, + "covertype_val_f1-score": 0.7064301371574402, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7217217087745667 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.6576308608055115, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.7139999866485596, + "covertype_test_precision": 0.7054161429405212, + "covertype_test_recall": 0.7139999866485596, + "covertype_test_f1-score": 0.700498104095459, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7139999866485596 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.787821888923645, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.6266266107559204, + "covertype_val_precision": 0.6616939902305603, + "covertype_val_recall": 0.6266266107559204, + "covertype_val_f1-score": 0.6064246892929077, + "covertype_val_support": 999.0, + "covertype_val_score": 0.6266266107559204 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.7834094166755676, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.6359999775886536, + "covertype_test_precision": 0.6862476468086243, + "covertype_test_recall": 0.6359999775886536, + "covertype_test_f1-score": 0.6164293885231018, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.6359999775886536 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.6974802017211914, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.6826826930046082, + "covertype_val_precision": 0.6926795840263367, + "covertype_val_recall": 0.6826826930046082, + "covertype_val_f1-score": 0.6763802170753479, + "covertype_val_support": 999.0, + "covertype_val_score": 0.6826826930046082 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.6866143345832825, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.6930000185966492, + "covertype_test_precision": 0.7043637037277222, + "covertype_test_recall": 0.6930000185966492, + "covertype_test_f1-score": 0.6859701871871948, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.6930000185966492 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.7326705813407898, + "std": 0.1193950951921182, + "n": 5, + "ci95": 0.14822482482125313 + }, + "lr": { + "mean": 0.0010000001639127731, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.6758000016212463, + "std": 0.04819958220237879, + "n": 5, + "ci95": 0.05983809148030076 + }, + "covertype_test_precision": { + "mean": 0.7017390727996826, + "std": 0.010300905265005815, + "n": 5, + "ci95": 0.012788212748178327 + }, + "covertype_test_recall": { + "mean": 0.6758000016212463, + "std": 0.04819958220237879, + "n": 5, + "ci95": 0.05983809148030076 + }, + "covertype_test_f1-score": { + "mean": 0.6654508113861084, + "std": 0.04861379878511744, + "n": 5, + "ci95": 0.060352326845796306 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.6758000016212463, + "std": 0.04819958220237879, + "n": 5, + "ci95": 0.05983809148030076 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/downstream_results/resnet_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/downstream_results/resnet_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..b742a1dd364977204afeece9d25afbd1740a6e3f --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/downstream_results/resnet_covertype_downstream_results.json @@ -0,0 +1,208 @@ +{ + "dataset": "covertype", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpmur6hxc7", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpmur6hxc7/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.6101607084274292, + "lr": 0.0, + "covertype_val_accuracy": 0.7537537813186646, + "covertype_val_precision": 0.7506377100944519, + "covertype_val_recall": 0.7537537813186646, + "covertype_val_f1-score": 0.7510454058647156, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7537537813186646 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5641831159591675, + "lr": 0.0, + "covertype_test_accuracy": 0.7754999995231628, + "covertype_test_precision": 0.7762905955314636, + "covertype_test_recall": 0.7754999995231628, + "covertype_test_f1-score": 0.7751973867416382, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7754999995231628 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.596743106842041, + "lr": 0.0, + "covertype_val_accuracy": 0.7667667865753174, + "covertype_val_precision": 0.7584322690963745, + "covertype_val_recall": 0.7667667865753174, + "covertype_val_f1-score": 0.7604787945747375, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7667667865753174 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5700294375419617, + "lr": 0.0, + "covertype_test_accuracy": 0.7745000123977661, + "covertype_test_precision": 0.7750622630119324, + "covertype_test_recall": 0.7745000123977661, + "covertype_test_f1-score": 0.7729076743125916, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7745000123977661 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.6056398153305054, + "lr": 0.0, + "covertype_val_accuracy": 0.7557557821273804, + "covertype_val_precision": 0.7565683126449585, + "covertype_val_recall": 0.7557557821273804, + "covertype_val_f1-score": 0.7529407143592834, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7557557821273804 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5701727271080017, + "lr": 0.0, + "covertype_test_accuracy": 0.7555000185966492, + "covertype_test_precision": 0.7631667256355286, + "covertype_test_recall": 0.7555000185966492, + "covertype_test_f1-score": 0.7551863193511963, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7555000185966492 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.5998233556747437, + "lr": 0.0, + "covertype_val_accuracy": 0.7467467188835144, + "covertype_val_precision": 0.7450889945030212, + "covertype_val_recall": 0.7467467188835144, + "covertype_val_f1-score": 0.7430373430252075, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7467467188835144 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5897141098976135, + "lr": 0.0, + "covertype_test_accuracy": 0.7680000066757202, + "covertype_test_precision": 0.767589271068573, + "covertype_test_recall": 0.7680000066757202, + "covertype_test_f1-score": 0.7667495012283325, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7680000066757202 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.5699985027313232, + "lr": 0.0005000000237487257, + "covertype_val_accuracy": 0.7787787914276123, + "covertype_val_precision": 0.7761978507041931, + "covertype_val_recall": 0.7787787914276123, + "covertype_val_f1-score": 0.7747555375099182, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7787787914276123 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5447656512260437, + "lr": 0.0005000000819563866, + "covertype_test_accuracy": 0.781000018119812, + "covertype_test_precision": 0.7806215882301331, + "covertype_test_recall": 0.781000018119812, + "covertype_test_f1-score": 0.7776663899421692, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.781000018119812 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5677730083465576, + "std": 0.016081727745686645, + "n": 5, + "ci95": 0.01996490118871188 + }, + "lr": { + "mean": 0.00010000001639127731, + "std": 0.00022360683440198926, + "n": 5, + "ci95": 0.00027760004550218576 + }, + "covertype_test_accuracy": { + "mean": 0.7709000110626221, + "std": 0.009768570000280311, + "n": 5, + "ci95": 0.01212733718010623 + }, + "covertype_test_precision": { + "mean": 0.7725460886955261, + "std": 0.00703755372566494, + "n": 5, + "ci95": 0.008736876221576189 + }, + "covertype_test_recall": { + "mean": 0.7709000110626221, + "std": 0.009768570000280311, + "n": 5, + "ci95": 0.01212733718010623 + }, + "covertype_test_f1-score": { + "mean": 0.7695414543151855, + "std": 0.008988888828710942, + "n": 5, + "ci95": 0.011159390340360937 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.7709000110626221, + "std": 0.009768570000280311, + "n": 5, + "ci95": 0.01212733718010623 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..8cd15191273d134651a9209fee82e52a1e2ef512 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9197f75b0fa6e928a76b67b0c2c01602dce738d4ffbc51c7ab436ac09ae4daf9 +size 138240128 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..42b3dc7289690338ad00ba06634853872b85a44e --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9a53da33394d1f2439644ffd293dd379c7c4311585c445947d3156b1ff300848 +size 10878670 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5058fe000cafbcf2edca3e3e37b3a9da26c415cb --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0009375_sx0.000375_sy3.75e-05/seed_42/wjepa_deterministic/covertype/20260915_205359/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 +data_set: covertype +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.0005 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 12 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.35 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.15 +mask_min_trgt_share: 0.15 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 64 +model_dim_hidden: 64 +model_dropout_prob: 0.0015 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 8 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.001 +recon_y_weight: 0.5 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.000375 +sw_s_y_weight: 3.75e-05 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 60 +vtjepa_kl_x_end_weight: 1.0e-06 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 20 +vtjepa_kl_y_end_weight: 1.0e-06 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 60 +vtjepa_kl_z_end_weight: 1.0e-05 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 60 +wjepa_joint_end_weight: 0.0009375 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..ccab2cb2c1c37ae4a23036cc2428dc07f0db6ddc --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-15T21:44:12.755820+00:00", + "spec": { + "dataset": "covertype", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.000375, + "sw_z_weight": 0.0, + "sw_s_y_weight": 7.5e-05, + "lambda_joint": 0.001875, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..b0d2ebb88af237766167f503469b6204912231c4 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "covertype", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.000375, + "sw_z_weight": 0.0, + "sw_s_y_weight": 7.5e-05, + "lambda_joint": 0.001875, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "covertype", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.001875", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.000375", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "7.5e-05", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..8a78f16362c77c4fe9687d0381ef892bafb0114a --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "covertype", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.001875, + "s_x": 0.000375, + "z": 0.0, + "s_y": 7.5e-05 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 60 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "covertype", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "64", + "--model_num_layers", + "8", + "--model_num_heads", + "8", + "--model_dropout_prob", + "0.0015", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.15", + "--mask_max_ctx_share", + "0.35", + "--mask_min_trgt_share", + "0.15", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.0005", + "--exp_warmup", + "12", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "1e-06", + "--vtjepa_kl_x_annealing_epochs", + "60", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-05", + "--vtjepa_kl_z_annealing_epochs", + "60", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-06", + "--vtjepa_kl_y_annealing_epochs", + "20", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.001", + "--recon_y_weight", + "0.5", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.001875", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "60", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.000375", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "7.5e-05", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/downstream_results/autoint_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/downstream_results/autoint_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..8558bf24f27e05a8a578cee84e9f48b946d11f78 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/downstream_results/autoint_covertype_downstream_results.json @@ -0,0 +1,210 @@ +{ + "dataset": "covertype", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpi559j6tm", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpi559j6tm/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.619693398475647, + "lr": 0.0, + "covertype_val_accuracy": 0.7577577829360962, + "covertype_val_precision": 0.7532408833503723, + "covertype_val_recall": 0.7577577829360962, + "covertype_val_f1-score": 0.7539095878601074, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7577577829360962 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5849414467811584, + "lr": 0.0, + "covertype_test_accuracy": 0.7549999952316284, + "covertype_test_precision": 0.7529736757278442, + "covertype_test_recall": 0.7549999952316284, + "covertype_test_f1-score": 0.7516788244247437, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7549999952316284 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.6235203146934509, + "lr": 9.866357686405536e-07, + "covertype_val_accuracy": 0.7527527809143066, + "covertype_val_precision": 0.7508582472801208, + "covertype_val_recall": 0.7527527809143066, + "covertype_val_f1-score": 0.7490429878234863, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7527527809143066 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5912920236587524, + "lr": 9.866358823273913e-07, + "covertype_test_accuracy": 0.7555000185966492, + "covertype_test_precision": 0.7541188597679138, + "covertype_test_recall": 0.7555000185966492, + "covertype_test_f1-score": 0.7526675462722778, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7555000185966492 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.6250841617584229, + "lr": 9.866357686405536e-07, + "covertype_val_accuracy": 0.7417417168617249, + "covertype_val_precision": 0.7378722429275513, + "covertype_val_recall": 0.7417417168617249, + "covertype_val_f1-score": 0.7378944158554077, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7417417168617249 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.599564790725708, + "lr": 9.866358823273913e-07, + "covertype_test_accuracy": 0.7490000128746033, + "covertype_test_precision": 0.7485790252685547, + "covertype_test_recall": 0.7490000128746033, + "covertype_test_f1-score": 0.7473034262657166, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7490000128746033 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.6190212965011597, + "lr": 0.0, + "covertype_val_accuracy": 0.7547547817230225, + "covertype_val_precision": 0.7500556111335754, + "covertype_val_recall": 0.7547547817230225, + "covertype_val_f1-score": 0.749700129032135, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7547547817230225 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5703136920928955, + "lr": 0.0, + "covertype_test_accuracy": 0.7574999928474426, + "covertype_test_precision": 0.7581517100334167, + "covertype_test_recall": 0.7574999928474426, + "covertype_test_f1-score": 0.7553126215934753, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7574999928474426 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.626970112323761, + "lr": 0.0, + "covertype_val_accuracy": 0.7567567825317383, + "covertype_val_precision": 0.7605102062225342, + "covertype_val_recall": 0.7567567825317383, + "covertype_val_f1-score": 0.7538949847221375, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7567567825317383 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5864923596382141, + "lr": 0.0, + "covertype_test_accuracy": 0.7419999837875366, + "covertype_test_precision": 0.7421045303344727, + "covertype_test_recall": 0.7419999837875366, + "covertype_test_f1-score": 0.7391722202301025, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7419999837875366 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5865208625793457, + "std": 0.010701352500866273, + "n": 5, + "ci95": 0.013285353952262645 + }, + "lr": { + "mean": 3.946543529309565e-07, + "std": 5.404027287947248e-07, + "n": 5, + "ci95": 6.708910418776824e-07 + }, + "covertype_test_accuracy": { + "mean": 0.751800000667572, + "std": 0.006330487169756797, + "n": 5, + "ci95": 0.007859077881386331 + }, + "covertype_test_precision": { + "mean": 0.7511855602264405, + "std": 0.006115308143607676, + "n": 5, + "ci95": 0.007591940664360462 + }, + "covertype_test_recall": { + "mean": 0.751800000667572, + "std": 0.006330487169756797, + "n": 5, + "ci95": 0.007859077881386331 + }, + "covertype_test_f1-score": { + "mean": 0.7492269277572632, + "std": 0.006318273576434806, + "n": 5, + "ci95": 0.00784391513347213 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.751800000667572, + "std": 0.006330487169756797, + "n": 5, + "ci95": 0.007859077881386331 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/downstream_results/dcnv2_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/downstream_results/dcnv2_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..a6061d2183081b6c85b6e2d8c3f9269c89bc8dc3 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/downstream_results/dcnv2_covertype_downstream_results.json @@ -0,0 +1,209 @@ +{ + "dataset": "covertype", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp224a80bm", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp224a80bm/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.5670343041419983, + "lr": 0.0, + "covertype_val_accuracy": 0.7677677869796753, + "covertype_val_precision": 0.763542890548706, + "covertype_val_recall": 0.7677677869796753, + "covertype_val_f1-score": 0.7617312669754028, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7677677869796753 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5539454817771912, + "lr": 0.0, + "covertype_test_accuracy": 0.7714999914169312, + "covertype_test_precision": 0.7744823694229126, + "covertype_test_recall": 0.7714999914169312, + "covertype_test_f1-score": 0.7681478261947632, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7714999914169312 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.5659714937210083, + "lr": 0.0, + "covertype_val_accuracy": 0.7727727890014648, + "covertype_val_precision": 0.7697402834892273, + "covertype_val_recall": 0.7727727890014648, + "covertype_val_f1-score": 0.7677854299545288, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7727727890014648 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5640353560447693, + "lr": 0.0, + "covertype_test_accuracy": 0.7760000228881836, + "covertype_test_precision": 0.7760930061340332, + "covertype_test_recall": 0.7760000228881836, + "covertype_test_f1-score": 0.7736854553222656, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7760000228881836 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.5665016174316406, + "lr": 0.0, + "covertype_val_accuracy": 0.7727727890014648, + "covertype_val_precision": 0.7647953629493713, + "covertype_val_recall": 0.7727727890014648, + "covertype_val_f1-score": 0.766312301158905, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7727727890014648 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5597935914993286, + "lr": 0.0, + "covertype_test_accuracy": 0.7799999713897705, + "covertype_test_precision": 0.780884861946106, + "covertype_test_recall": 0.7799999713897705, + "covertype_test_f1-score": 0.7773051261901855, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7799999713897705 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.5682074427604675, + "lr": 0.0, + "covertype_val_accuracy": 0.7717717885971069, + "covertype_val_precision": 0.7655641436576843, + "covertype_val_recall": 0.7717717885971069, + "covertype_val_f1-score": 0.7650895714759827, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7717717885971069 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5768951773643494, + "lr": 0.0, + "covertype_test_accuracy": 0.7699999809265137, + "covertype_test_precision": 0.771893322467804, + "covertype_test_recall": 0.7699999809265137, + "covertype_test_f1-score": 0.7663094401359558, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7699999809265137 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.5622308254241943, + "lr": 0.0, + "covertype_val_accuracy": 0.7687687873840332, + "covertype_val_precision": 0.7598406076431274, + "covertype_val_recall": 0.7687687873840332, + "covertype_val_f1-score": 0.7607771158218384, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7687687873840332 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5500445365905762, + "lr": 0.0, + "covertype_test_accuracy": 0.781499981880188, + "covertype_test_precision": 0.7791077494621277, + "covertype_test_recall": 0.781499981880188, + "covertype_test_f1-score": 0.777650773525238, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.781499981880188 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5609428286552429, + "std": 0.010405497757087281, + "n": 5, + "ci95": 0.012918060660200573 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.7757999897003174, + "std": 0.0050571695970152136, + "n": 5, + "ci95": 0.0062782987559310685 + }, + "covertype_test_precision": { + "mean": 0.7764922618865967, + "std": 0.0035876533831256828, + "n": 5, + "ci95": 0.004453945895997623 + }, + "covertype_test_recall": { + "mean": 0.7757999897003174, + "std": 0.0050571695970152136, + "n": 5, + "ci95": 0.0062782987559310685 + }, + "covertype_test_f1-score": { + "mean": 0.7726197242736816, + "std": 0.0052013889655972995, + "n": 5, + "ci95": 0.006457342045854444 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.7757999897003174, + "std": 0.0050571695970152136, + "n": 5, + "ci95": 0.0062782987559310685 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/downstream_results/fttransformer_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/downstream_results/fttransformer_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..d3c777868cd9fabfbeb480559e499f28596a60ee --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/downstream_results/fttransformer_covertype_downstream_results.json @@ -0,0 +1,209 @@ +{ + "dataset": "covertype", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmplssxy0xz", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmplssxy0xz/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.5927572846412659, + "lr": 0.0, + "covertype_val_accuracy": 0.7457457184791565, + "covertype_val_precision": 0.7419591546058655, + "covertype_val_recall": 0.7457457184791565, + "covertype_val_f1-score": 0.7374268174171448, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7457457184791565 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5826665759086609, + "lr": 0.0, + "covertype_test_accuracy": 0.7459999918937683, + "covertype_test_precision": 0.748648464679718, + "covertype_test_recall": 0.7459999918937683, + "covertype_test_f1-score": 0.7396411299705505, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7459999918937683 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.6183735728263855, + "lr": 0.0, + "covertype_val_accuracy": 0.7347347140312195, + "covertype_val_precision": 0.7315829396247864, + "covertype_val_recall": 0.7347347140312195, + "covertype_val_f1-score": 0.727230966091156, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7347347140312195 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5991592407226562, + "lr": 0.0, + "covertype_test_accuracy": 0.7404999732971191, + "covertype_test_precision": 0.7412769794464111, + "covertype_test_recall": 0.7404999732971191, + "covertype_test_f1-score": 0.7352632880210876, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7404999732971191 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.604399561882019, + "lr": 0.0, + "covertype_val_accuracy": 0.7587587833404541, + "covertype_val_precision": 0.7515845894813538, + "covertype_val_recall": 0.7587587833404541, + "covertype_val_f1-score": 0.7517743110656738, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7587587833404541 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5808154344558716, + "lr": 0.0, + "covertype_test_accuracy": 0.7419999837875366, + "covertype_test_precision": 0.7473835945129395, + "covertype_test_recall": 0.7419999837875366, + "covertype_test_f1-score": 0.7363055348396301, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7419999837875366 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.6076353192329407, + "lr": 0.0, + "covertype_val_accuracy": 0.7437437176704407, + "covertype_val_precision": 0.7436360120773315, + "covertype_val_recall": 0.7437437176704407, + "covertype_val_f1-score": 0.739002525806427, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7437437176704407 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5787150859832764, + "lr": 0.0, + "covertype_test_accuracy": 0.7544999718666077, + "covertype_test_precision": 0.7570748925209045, + "covertype_test_recall": 0.7544999718666077, + "covertype_test_f1-score": 0.7501082420349121, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7544999718666077 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.6089609861373901, + "lr": 0.0, + "covertype_val_accuracy": 0.7427427172660828, + "covertype_val_precision": 0.7406097054481506, + "covertype_val_recall": 0.7427427172660828, + "covertype_val_f1-score": 0.7342133522033691, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7427427172660828 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5874844789505005, + "lr": 0.0, + "covertype_test_accuracy": 0.7429999709129333, + "covertype_test_precision": 0.7427246570587158, + "covertype_test_recall": 0.7429999709129333, + "covertype_test_f1-score": 0.7368111610412598, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7429999709129333 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5857681632041931, + "std": 0.00815723545291209, + "n": 5, + "ci95": 0.010126921831152644 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.745199978351593, + "std": 0.005574494293189444, + "n": 5, + "ci95": 0.006920539229400664 + }, + "covertype_test_precision": { + "mean": 0.7474217176437378, + "std": 0.0062151092614209855, + "n": 5, + "ci95": 0.007715840253209063 + }, + "covertype_test_recall": { + "mean": 0.745199978351593, + "std": 0.005574494293189444, + "n": 5, + "ci95": 0.006920539229400664 + }, + "covertype_test_f1-score": { + "mean": 0.739625871181488, + "std": 0.006079878493180054, + "n": 5, + "ci95": 0.00754795599547886 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.745199978351593, + "std": 0.005574494293189444, + "n": 5, + "ci95": 0.006920539229400664 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/downstream_results/mlp_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/downstream_results/mlp_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..3909b2ce24d6a520f78307c96c6034fb41601105 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/downstream_results/mlp_covertype_downstream_results.json @@ -0,0 +1,205 @@ +{ + "dataset": "covertype", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpnb4m9u0r", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpnb4m9u0r/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.8640649914741516, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.6606606841087341, + "covertype_val_precision": 0.7004674077033997, + "covertype_val_recall": 0.6606606841087341, + "covertype_val_f1-score": 0.653672993183136, + "covertype_val_support": 999.0, + "covertype_val_score": 0.6606606841087341 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.868095338344574, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.6480000019073486, + "covertype_test_precision": 0.6894012689590454, + "covertype_test_recall": 0.6480000019073486, + "covertype_test_f1-score": 0.6423313021659851, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.6480000019073486 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.7250179648399353, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.6816816926002502, + "covertype_val_precision": 0.7066971063613892, + "covertype_val_recall": 0.6816816926002502, + "covertype_val_f1-score": 0.6789296865463257, + "covertype_val_support": 999.0, + "covertype_val_score": 0.6816816926002502 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.7129509449005127, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.6894999742507935, + "covertype_test_precision": 0.7080581188201904, + "covertype_test_recall": 0.6894999742507935, + "covertype_test_f1-score": 0.6824399828910828, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.6894999742507935 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.714799165725708, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.7277277112007141, + "covertype_val_precision": 0.713214635848999, + "covertype_val_recall": 0.7277277112007141, + "covertype_val_f1-score": 0.7086532115936279, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7277277112007141 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.7070547938346863, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.7149999737739563, + "covertype_test_precision": 0.689846932888031, + "covertype_test_recall": 0.7149999737739563, + "covertype_test_f1-score": 0.6948254704475403, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7149999737739563 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.6635671257972717, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.7297297120094299, + "covertype_val_precision": 0.7044103741645813, + "covertype_val_recall": 0.7297297120094299, + "covertype_val_f1-score": 0.711485743522644, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7297297120094299 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.6544139981269836, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.7174999713897705, + "covertype_test_precision": 0.6938683390617371, + "covertype_test_recall": 0.7174999713897705, + "covertype_test_f1-score": 0.700504720211029, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7174999713897705 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.9710546135902405, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.5775775909423828, + "covertype_val_precision": 0.6303550601005554, + "covertype_val_recall": 0.5775775909423828, + "covertype_val_f1-score": 0.5521331429481506, + "covertype_val_support": 999.0, + "covertype_val_score": 0.5775775909423828 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.9521642327308655, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.578000009059906, + "covertype_test_precision": 0.6271469593048096, + "covertype_test_recall": 0.578000009059906, + "covertype_test_f1-score": 0.5478879809379578, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.578000009059906 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.7789358615875244, + "std": 0.12548313878339545, + "n": 5, + "ci95": 0.15578291749976036 + }, + "lr": { + "mean": 0.0010000001639127731, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.669599986076355, + "std": 0.058328149530963755, + "n": 5, + "ci95": 0.07241235272238972 + }, + "covertype_test_precision": { + "mean": 0.6816643238067627, + "std": 0.03140274210741633, + "n": 5, + "ci95": 0.038985403381009656 + }, + "covertype_test_recall": { + "mean": 0.669599986076355, + "std": 0.058328149530963755, + "n": 5, + "ci95": 0.07241235272238972 + }, + "covertype_test_f1-score": { + "mean": 0.653597891330719, + "std": 0.0633114351986854, + "n": 5, + "ci95": 0.07859892717039152 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.669599986076355, + "std": 0.058328149530963755, + "n": 5, + "ci95": 0.07241235272238972 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/downstream_results/resnet_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/downstream_results/resnet_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..f8ff15897808263264ce4bd5fc5a8c7f6f59db1a --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/downstream_results/resnet_covertype_downstream_results.json @@ -0,0 +1,208 @@ +{ + "dataset": "covertype", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp0cefbomy", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp0cefbomy/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.603825569152832, + "lr": 0.0, + "covertype_val_accuracy": 0.7627627849578857, + "covertype_val_precision": 0.7594751715660095, + "covertype_val_recall": 0.7627627849578857, + "covertype_val_f1-score": 0.7603427767753601, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7627627849578857 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5540969371795654, + "lr": 0.0, + "covertype_test_accuracy": 0.781499981880188, + "covertype_test_precision": 0.7814279198646545, + "covertype_test_recall": 0.781499981880188, + "covertype_test_f1-score": 0.7809062600135803, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.781499981880188 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.586609423160553, + "lr": 0.0, + "covertype_val_accuracy": 0.7627627849578857, + "covertype_val_precision": 0.7546623349189758, + "covertype_val_recall": 0.7627627849578857, + "covertype_val_f1-score": 0.757086992263794, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7627627849578857 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5604619383811951, + "lr": 0.0, + "covertype_test_accuracy": 0.7770000100135803, + "covertype_test_precision": 0.7756558060646057, + "covertype_test_recall": 0.7770000100135803, + "covertype_test_f1-score": 0.7746079564094543, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7770000100135803 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.618913471698761, + "lr": 0.0, + "covertype_val_accuracy": 0.7427427172660828, + "covertype_val_precision": 0.7313351631164551, + "covertype_val_recall": 0.7427427172660828, + "covertype_val_f1-score": 0.7318178415298462, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7427427172660828 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5665544271469116, + "lr": 0.0, + "covertype_test_accuracy": 0.7804999947547913, + "covertype_test_precision": 0.7814571857452393, + "covertype_test_recall": 0.7804999947547913, + "covertype_test_f1-score": 0.7766857147216797, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7804999947547913 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.6156212687492371, + "lr": 0.0, + "covertype_val_accuracy": 0.7547547817230225, + "covertype_val_precision": 0.7553582787513733, + "covertype_val_recall": 0.7547547817230225, + "covertype_val_f1-score": 0.7512439489364624, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7547547817230225 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5878245234489441, + "lr": 0.0, + "covertype_test_accuracy": 0.7680000066757202, + "covertype_test_precision": 0.76950603723526, + "covertype_test_recall": 0.7680000066757202, + "covertype_test_f1-score": 0.7671436667442322, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7680000066757202 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.604916512966156, + "lr": 0.0, + "covertype_val_accuracy": 0.7447447180747986, + "covertype_val_precision": 0.741312563419342, + "covertype_val_recall": 0.7447447180747986, + "covertype_val_f1-score": 0.7414860129356384, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7447447180747986 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5912460684776306, + "lr": 0.0, + "covertype_test_accuracy": 0.7754999995231628, + "covertype_test_precision": 0.7787951231002808, + "covertype_test_recall": 0.7754999995231628, + "covertype_test_f1-score": 0.7757450938224792, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7754999995231628 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5720367789268493, + "std": 0.01661415600289845, + "n": 5, + "ci95": 0.020625892203695508 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.7764999985694885, + "std": 0.005350226030755888, + "n": 5, + "ci95": 0.006642118044186221 + }, + "covertype_test_precision": { + "mean": 0.7773684144020081, + "std": 0.004999578368166494, + "n": 5, + "ci95": 0.006206801264400062 + }, + "covertype_test_recall": { + "mean": 0.7764999985694885, + "std": 0.005350226030755888, + "n": 5, + "ci95": 0.006642118044186221 + }, + "covertype_test_f1-score": { + "mean": 0.7750177383422852, + "std": 0.0050038685892476835, + "n": 5, + "ci95": 0.006212127423461961 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.7764999985694885, + "std": 0.005350226030755888, + "n": 5, + "ci95": 0.006642118044186221 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..fa95a3909e659672a98a8b2025dcaea083181f99 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7b7ee838c50fc2deaa43c001fd55aa122cbda57b6ff36a30f021503de9764693 +size 138240128 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..b0d52c652f034f6865dc3bd195a136a5f65b7c2f --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:deae1840489f1d12a822e820110e9b8cd5abde43e1cc07a9f47b58b6c06d0c38 +size 10878670 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..48a768810869cde71029cca5a26292169458029e --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.000375_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_211828/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 +data_set: covertype +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.0005 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 12 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.35 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.15 +mask_min_trgt_share: 0.15 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 64 +model_dim_hidden: 64 +model_dropout_prob: 0.0015 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 8 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.001 +recon_y_weight: 0.5 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.000375 +sw_s_y_weight: 7.5e-05 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 60 +vtjepa_kl_x_end_weight: 1.0e-06 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 20 +vtjepa_kl_y_end_weight: 1.0e-06 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 60 +vtjepa_kl_z_end_weight: 1.0e-05 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 60 +wjepa_joint_end_weight: 0.001875 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..af89053a2445fea96344749666d8350854c7f274 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-15T23:35:38.879689+00:00", + "spec": { + "dataset": "covertype", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.00075, + "sw_z_weight": 0.0, + "sw_s_y_weight": 7.5e-05, + "lambda_joint": 0.001875, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..4035da4df8f5acf60d0c098f2fd091e023f8b1b6 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "covertype", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.00075, + "sw_z_weight": 0.0, + "sw_s_y_weight": 7.5e-05, + "lambda_joint": 0.001875, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "covertype", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.001875", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.00075", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "7.5e-05", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..790d04be92d6cb4409dd84cada341e0fff729bf9 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "covertype", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.001875, + "s_x": 0.00075, + "z": 0.0, + "s_y": 7.5e-05 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 60 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "covertype", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "64", + "--model_num_layers", + "8", + "--model_num_heads", + "8", + "--model_dropout_prob", + "0.0015", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.15", + "--mask_max_ctx_share", + "0.35", + "--mask_min_trgt_share", + "0.15", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.0005", + "--exp_warmup", + "12", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "1e-06", + "--vtjepa_kl_x_annealing_epochs", + "60", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-05", + "--vtjepa_kl_z_annealing_epochs", + "60", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-06", + "--vtjepa_kl_y_annealing_epochs", + "20", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.001", + "--recon_y_weight", + "0.5", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.001875", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "60", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.00075", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "7.5e-05", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/downstream_results/autoint_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/downstream_results/autoint_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..284497d449c4ff6554e017725f3a7357c3e84360 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/downstream_results/autoint_covertype_downstream_results.json @@ -0,0 +1,210 @@ +{ + "dataset": "covertype", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp469suu8_", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp469suu8_/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.6210604906082153, + "lr": 0.0, + "covertype_val_accuracy": 0.7527527809143066, + "covertype_val_precision": 0.7481434345245361, + "covertype_val_recall": 0.7527527809143066, + "covertype_val_f1-score": 0.748886227607727, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7527527809143066 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5905119776725769, + "lr": 0.0, + "covertype_test_accuracy": 0.753000020980835, + "covertype_test_precision": 0.7534472942352295, + "covertype_test_recall": 0.753000020980835, + "covertype_test_f1-score": 0.7517149448394775, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.753000020980835 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.6148595213890076, + "lr": 0.0, + "covertype_val_accuracy": 0.7467467188835144, + "covertype_val_precision": 0.7431632876396179, + "covertype_val_recall": 0.7467467188835144, + "covertype_val_f1-score": 0.7426261901855469, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7467467188835144 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5982586741447449, + "lr": 0.0, + "covertype_test_accuracy": 0.7484999895095825, + "covertype_test_precision": 0.7465230822563171, + "covertype_test_recall": 0.7484999895095825, + "covertype_test_f1-score": 0.7457489371299744, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7484999895095825 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.5999589562416077, + "lr": 9.866357686405536e-07, + "covertype_val_accuracy": 0.7577577829360962, + "covertype_val_precision": 0.7574031949043274, + "covertype_val_recall": 0.7577577829360962, + "covertype_val_f1-score": 0.7554540634155273, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7577577829360962 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5876102447509766, + "lr": 9.866358823273913e-07, + "covertype_test_accuracy": 0.7574999928474426, + "covertype_test_precision": 0.7553665637969971, + "covertype_test_recall": 0.7574999928474426, + "covertype_test_f1-score": 0.7547937035560608, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7574999928474426 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.6177465319633484, + "lr": 0.0, + "covertype_val_accuracy": 0.7267267107963562, + "covertype_val_precision": 0.7270073890686035, + "covertype_val_recall": 0.7267267107963562, + "covertype_val_f1-score": 0.7243560552597046, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7267267107963562 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5888822078704834, + "lr": 0.0, + "covertype_test_accuracy": 0.7565000057220459, + "covertype_test_precision": 0.7572526335716248, + "covertype_test_recall": 0.7565000057220459, + "covertype_test_f1-score": 0.7551760673522949, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7565000057220459 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.6296648979187012, + "lr": 3.511175964376889e-05, + "covertype_val_accuracy": 0.7437437176704407, + "covertype_val_precision": 0.7408604025840759, + "covertype_val_recall": 0.7437437176704407, + "covertype_val_f1-score": 0.7389692664146423, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7437437176704407 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5885452628135681, + "lr": 3.511175964376889e-05, + "covertype_test_accuracy": 0.7455000281333923, + "covertype_test_precision": 0.7455663681030273, + "covertype_test_recall": 0.7455000281333923, + "covertype_test_f1-score": 0.7421423196792603, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7455000281333923 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5907616734504699, + "std": 0.004319830355962276, + "n": 5, + "ci95": 0.005362917938460753 + }, + "lr": { + "mean": 7.219679105219257e-06, + "std": 1.5597998940043528e-05, + "n": 5, + "ci95": 1.9364368835501962e-05 + }, + "covertype_test_accuracy": { + "mean": 0.7522000074386597, + "std": 0.005142949241549142, + "n": 5, + "ci95": 0.006384791177280638 + }, + "covertype_test_precision": { + "mean": 0.7516311883926392, + "std": 0.0052850427904732, + "n": 5, + "ci95": 0.006561195336627453 + }, + "covertype_test_recall": { + "mean": 0.7522000074386597, + "std": 0.005142949241549142, + "n": 5, + "ci95": 0.006384791177280638 + }, + "covertype_test_f1-score": { + "mean": 0.7499151945114135, + "std": 0.0057552371479095725, + "n": 5, + "ci95": 0.007144925146891458 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.7522000074386597, + "std": 0.005142949241549142, + "n": 5, + "ci95": 0.006384791177280638 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/downstream_results/dcnv2_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/downstream_results/dcnv2_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..0a048dde1a94c9657cd421c5e28a98dfa5c9bbbd --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/downstream_results/dcnv2_covertype_downstream_results.json @@ -0,0 +1,209 @@ +{ + "dataset": "covertype", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp1u1_d6lo", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp1u1_d6lo/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.5759305357933044, + "lr": 0.0, + "covertype_val_accuracy": 0.759759783744812, + "covertype_val_precision": 0.7565332651138306, + "covertype_val_recall": 0.759759783744812, + "covertype_val_f1-score": 0.75389164686203, + "covertype_val_support": 999.0, + "covertype_val_score": 0.759759783744812 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5478729605674744, + "lr": 0.0, + "covertype_test_accuracy": 0.7850000262260437, + "covertype_test_precision": 0.7864188551902771, + "covertype_test_recall": 0.7850000262260437, + "covertype_test_f1-score": 0.7826421856880188, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7850000262260437 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.5856520533561707, + "lr": 0.0, + "covertype_val_accuracy": 0.7667667865753174, + "covertype_val_precision": 0.7599090933799744, + "covertype_val_recall": 0.7667667865753174, + "covertype_val_f1-score": 0.7613160610198975, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7667667865753174 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5665362477302551, + "lr": 0.0, + "covertype_test_accuracy": 0.7705000042915344, + "covertype_test_precision": 0.7719066143035889, + "covertype_test_recall": 0.7705000042915344, + "covertype_test_f1-score": 0.7678247094154358, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7705000042915344 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.5612449049949646, + "lr": 0.0, + "covertype_val_accuracy": 0.7727727890014648, + "covertype_val_precision": 0.767389714717865, + "covertype_val_recall": 0.7727727890014648, + "covertype_val_f1-score": 0.7663441300392151, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7727727890014648 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5588735938072205, + "lr": 0.0, + "covertype_test_accuracy": 0.7770000100135803, + "covertype_test_precision": 0.7781260013580322, + "covertype_test_recall": 0.7770000100135803, + "covertype_test_f1-score": 0.7740790247917175, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7770000100135803 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.5603235960006714, + "lr": 0.0, + "covertype_val_accuracy": 0.7777777910232544, + "covertype_val_precision": 0.778019368648529, + "covertype_val_recall": 0.7777777910232544, + "covertype_val_f1-score": 0.773967981338501, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7777777910232544 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5782790184020996, + "lr": 0.0, + "covertype_test_accuracy": 0.765999972820282, + "covertype_test_precision": 0.7669026255607605, + "covertype_test_recall": 0.765999972820282, + "covertype_test_f1-score": 0.7647589445114136, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.765999972820282 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.5580820441246033, + "lr": 0.0, + "covertype_val_accuracy": 0.7727727890014648, + "covertype_val_precision": 0.7668650150299072, + "covertype_val_recall": 0.7727727890014648, + "covertype_val_f1-score": 0.766056478023529, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7727727890014648 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5508959889411926, + "lr": 0.0, + "covertype_test_accuracy": 0.7764999866485596, + "covertype_test_precision": 0.7792395353317261, + "covertype_test_recall": 0.7764999866485596, + "covertype_test_f1-score": 0.7748356461524963, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7764999866485596 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5604915618896484, + "std": 0.012317201146636748, + "n": 5, + "ci95": 0.01529137339612334 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.775, + "std": 0.007202447069116718, + "n": 5, + "ci95": 0.008941585526493632 + }, + "covertype_test_precision": { + "mean": 0.776518726348877, + "std": 0.007445462669271063, + "n": 5, + "ci95": 0.009243280874227542 + }, + "covertype_test_recall": { + "mean": 0.775, + "std": 0.007202447069116718, + "n": 5, + "ci95": 0.008941585526493632 + }, + "covertype_test_f1-score": { + "mean": 0.7728281021118164, + "std": 0.006929222898734383, + "n": 5, + "ci95": 0.008602387297900676 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.775, + "std": 0.007202447069116718, + "n": 5, + "ci95": 0.008941585526493632 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/downstream_results/fttransformer_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/downstream_results/fttransformer_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..3088cd2cf53c7ef2eeb8c734cb9b21b0ab99ad2b --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/downstream_results/fttransformer_covertype_downstream_results.json @@ -0,0 +1,209 @@ +{ + "dataset": "covertype", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpasor2jmy", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpasor2jmy/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.6155408620834351, + "lr": 0.0, + "covertype_val_accuracy": 0.7477477192878723, + "covertype_val_precision": 0.7297002077102661, + "covertype_val_recall": 0.7477477192878723, + "covertype_val_f1-score": 0.7361925840377808, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7477477192878723 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5945818424224854, + "lr": 0.0, + "covertype_test_accuracy": 0.7390000224113464, + "covertype_test_precision": 0.7274198532104492, + "covertype_test_recall": 0.7390000224113464, + "covertype_test_f1-score": 0.7285676598548889, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7390000224113464 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.6016526222229004, + "lr": 0.0, + "covertype_val_accuracy": 0.7457457184791565, + "covertype_val_precision": 0.7388072609901428, + "covertype_val_recall": 0.7457457184791565, + "covertype_val_f1-score": 0.7360843420028687, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7457457184791565 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5934622883796692, + "lr": 0.0, + "covertype_test_accuracy": 0.7404999732971191, + "covertype_test_precision": 0.7396467924118042, + "covertype_test_recall": 0.7404999732971191, + "covertype_test_f1-score": 0.7323582172393799, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7404999732971191 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.5965853929519653, + "lr": 0.0, + "covertype_val_accuracy": 0.7427427172660828, + "covertype_val_precision": 0.7358342409133911, + "covertype_val_recall": 0.7427427172660828, + "covertype_val_f1-score": 0.7351154685020447, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7427427172660828 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5890328884124756, + "lr": 0.0, + "covertype_test_accuracy": 0.7415000200271606, + "covertype_test_precision": 0.7395013570785522, + "covertype_test_recall": 0.7415000200271606, + "covertype_test_f1-score": 0.7356262803077698, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7415000200271606 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.6058993935585022, + "lr": 0.0, + "covertype_val_accuracy": 0.7527527809143066, + "covertype_val_precision": 0.7523147463798523, + "covertype_val_recall": 0.7527527809143066, + "covertype_val_f1-score": 0.7450913190841675, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7527527809143066 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5745090246200562, + "lr": 0.0, + "covertype_test_accuracy": 0.75, + "covertype_test_precision": 0.7542448043823242, + "covertype_test_recall": 0.75, + "covertype_test_f1-score": 0.7460992336273193, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.75 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.5903725028038025, + "lr": 0.0, + "covertype_val_accuracy": 0.7527527809143066, + "covertype_val_precision": 0.7500065565109253, + "covertype_val_recall": 0.7527527809143066, + "covertype_val_f1-score": 0.7474658489227295, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7527527809143066 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5884156227111816, + "lr": 0.0, + "covertype_test_accuracy": 0.7494999766349792, + "covertype_test_precision": 0.7507426738739014, + "covertype_test_recall": 0.7494999766349792, + "covertype_test_f1-score": 0.7457274794578552, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7494999766349792 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5880003333091736, + "std": 0.008006298471743188, + "n": 5, + "ci95": 0.00993953886071479 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.7440999984741211, + "std": 0.005236879012916466, + "n": 5, + "ci95": 0.00650140169535945 + }, + "covertype_test_precision": { + "mean": 0.7423110961914062, + "std": 0.010609514119790522, + "n": 5, + "ci95": 0.013171339821908995 + }, + "covertype_test_recall": { + "mean": 0.7440999984741211, + "std": 0.005236879012916466, + "n": 5, + "ci95": 0.00650140169535945 + }, + "covertype_test_f1-score": { + "mean": 0.7376757740974427, + "std": 0.00792494811419001, + "n": 5, + "ci95": 0.009838545243865928 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.7440999984741211, + "std": 0.005236879012916466, + "n": 5, + "ci95": 0.00650140169535945 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/downstream_results/mlp_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/downstream_results/mlp_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..a8bcfaeb50531398e5c6c780a304b78a45af2146 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/downstream_results/mlp_covertype_downstream_results.json @@ -0,0 +1,205 @@ +{ + "dataset": "covertype", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp48gj1mv2", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp48gj1mv2/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.6716936826705933, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.7277277112007141, + "covertype_val_precision": 0.7335260510444641, + "covertype_val_recall": 0.7277277112007141, + "covertype_val_f1-score": 0.7182832956314087, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7277277112007141 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.7074241042137146, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.7114999890327454, + "covertype_test_precision": 0.6978535652160645, + "covertype_test_recall": 0.7114999890327454, + "covertype_test_f1-score": 0.7001162767410278, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7114999890327454 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.747097373008728, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.7197197079658508, + "covertype_val_precision": 0.7187588810920715, + "covertype_val_recall": 0.7197197079658508, + "covertype_val_f1-score": 0.7083337306976318, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7197197079658508 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.731846809387207, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.6945000290870667, + "covertype_test_precision": 0.6960458159446716, + "covertype_test_recall": 0.6945000290870667, + "covertype_test_f1-score": 0.6832040548324585, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.6945000290870667 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.6517016291618347, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.7137137055397034, + "covertype_val_precision": 0.7004646062850952, + "covertype_val_recall": 0.7137137055397034, + "covertype_val_f1-score": 0.6883257031440735, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7137137055397034 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.6506989598274231, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.7074999809265137, + "covertype_test_precision": 0.6984860301017761, + "covertype_test_recall": 0.7074999809265137, + "covertype_test_f1-score": 0.6821770071983337, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7074999809265137 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 1.1511790752410889, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.5125125050544739, + "covertype_val_precision": 0.5424755215644836, + "covertype_val_recall": 0.5125125050544739, + "covertype_val_f1-score": 0.4801228940486908, + "covertype_val_support": 999.0, + "covertype_val_score": 0.5125125050544739 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 1.154642939567566, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.5264999866485596, + "covertype_test_precision": 0.5562233924865723, + "covertype_test_recall": 0.5264999866485596, + "covertype_test_f1-score": 0.4904860258102417, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.5264999866485596 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.7696166634559631, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.661661684513092, + "covertype_val_precision": 0.6684898138046265, + "covertype_val_recall": 0.661661684513092, + "covertype_val_f1-score": 0.648414134979248, + "covertype_val_support": 999.0, + "covertype_val_score": 0.661661684513092 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.7460566759109497, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.6744999885559082, + "covertype_test_precision": 0.6900953650474548, + "covertype_test_recall": 0.6744999885559082, + "covertype_test_f1-score": 0.6627953052520752, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.6744999885559082 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.7981338977813721, + "std": 0.20258951082138912, + "n": 5, + "ci95": 0.2515077751209507 + }, + "lr": { + "mean": 0.0010000001639127731, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.6628999948501587, + "std": 0.07760348324608568, + "n": 5, + "ci95": 0.09634200375786836 + }, + "covertype_test_precision": { + "mean": 0.6677408337593078, + "std": 0.06242811964084814, + "n": 5, + "ci95": 0.07750232187340142 + }, + "covertype_test_recall": { + "mean": 0.6628999948501587, + "std": 0.07760348324608568, + "n": 5, + "ci95": 0.09634200375786836 + }, + "covertype_test_f1-score": { + "mean": 0.6437557339668274, + "std": 0.08669340192710935, + "n": 5, + "ci95": 0.10762681911788086 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.6628999948501587, + "std": 0.07760348324608568, + "n": 5, + "ci95": 0.09634200375786836 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/downstream_results/resnet_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/downstream_results/resnet_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..4036c4ffef6b72bc71e4716a87e7ca2f18be97f1 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/downstream_results/resnet_covertype_downstream_results.json @@ -0,0 +1,208 @@ +{ + "dataset": "covertype", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpfck10p_2", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpfck10p_2/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.5991008281707764, + "lr": 0.0, + "covertype_val_accuracy": 0.7477477192878723, + "covertype_val_precision": 0.7439250349998474, + "covertype_val_recall": 0.7477477192878723, + "covertype_val_f1-score": 0.7452293634414673, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7477477192878723 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5643165111541748, + "lr": 0.0, + "covertype_test_accuracy": 0.7739999890327454, + "covertype_test_precision": 0.774341881275177, + "covertype_test_recall": 0.7739999890327454, + "covertype_test_f1-score": 0.7733880281448364, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7739999890327454 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.6245357394218445, + "lr": 0.0, + "covertype_val_accuracy": 0.7557557821273804, + "covertype_val_precision": 0.74927818775177, + "covertype_val_recall": 0.7557557821273804, + "covertype_val_f1-score": 0.7491279244422913, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7557557821273804 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5583758354187012, + "lr": 0.0, + "covertype_test_accuracy": 0.7789999842643738, + "covertype_test_precision": 0.7850139141082764, + "covertype_test_recall": 0.7789999842643738, + "covertype_test_f1-score": 0.777698278427124, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7789999842643738 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.6009353399276733, + "lr": 0.0, + "covertype_val_accuracy": 0.7577577829360962, + "covertype_val_precision": 0.7495850324630737, + "covertype_val_recall": 0.7577577829360962, + "covertype_val_f1-score": 0.7501642107963562, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7577577829360962 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5516045093536377, + "lr": 0.0, + "covertype_test_accuracy": 0.7705000042915344, + "covertype_test_precision": 0.7724647521972656, + "covertype_test_recall": 0.7705000042915344, + "covertype_test_f1-score": 0.7685503959655762, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7705000042915344 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.5863093137741089, + "lr": 0.0, + "covertype_val_accuracy": 0.7657657861709595, + "covertype_val_precision": 0.7628076076507568, + "covertype_val_recall": 0.7657657861709595, + "covertype_val_f1-score": 0.7634060382843018, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7657657861709595 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5828087329864502, + "lr": 0.0, + "covertype_test_accuracy": 0.7820000052452087, + "covertype_test_precision": 0.7827760577201843, + "covertype_test_recall": 0.7820000052452087, + "covertype_test_f1-score": 0.7818713784217834, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7820000052452087 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.5974613428115845, + "lr": 0.0, + "covertype_val_accuracy": 0.7627627849578857, + "covertype_val_precision": 0.753847062587738, + "covertype_val_recall": 0.7627627849578857, + "covertype_val_f1-score": 0.7572939395904541, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7627627849578857 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5822221636772156, + "lr": 0.0, + "covertype_test_accuracy": 0.7689999938011169, + "covertype_test_precision": 0.7705439925193787, + "covertype_test_recall": 0.7689999938011169, + "covertype_test_f1-score": 0.7680985331535339, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7689999938011169 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5678655505180359, + "std": 0.014111014106981162, + "n": 5, + "ci95": 0.017518329297295875 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.7748999953269958, + "std": 0.005527205455122589, + "n": 5, + "ci95": 0.006861831794834935 + }, + "covertype_test_precision": { + "mean": 0.7770281195640564, + "std": 0.006459408040788618, + "n": 5, + "ci95": 0.008019128622949427 + }, + "covertype_test_recall": { + "mean": 0.7748999953269958, + "std": 0.005527205455122589, + "n": 5, + "ci95": 0.006861831794834935 + }, + "covertype_test_f1-score": { + "mean": 0.7739213228225708, + "std": 0.005926738473133153, + "n": 5, + "ci95": 0.007357838029510075 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.7748999953269958, + "std": 0.005527205455122589, + "n": 5, + "ci95": 0.006861831794834935 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..669b3427458e7453ff2448b46ea43432c69b74e5 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ac9ce37e63b3db73ff5fe104ad6d972869b6710618209ecd18aaaeae2608036d +size 138240128 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..f06a02090a4501544e9d71ff7aa526d7c00dfd8d --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4b103836cf14f37c49dee6d5942311e355c585fb9600554b0c1a561498b4bd3f +size 10878670 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..77b4a6e571dbb89f0a8278003a31e68111099421 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.001875_sx0.00075_sy7.5e-05/seed_42/wjepa_deterministic/covertype/20260915_231131/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 +data_set: covertype +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.0005 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 12 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.35 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.15 +mask_min_trgt_share: 0.15 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 64 +model_dim_hidden: 64 +model_dropout_prob: 0.0015 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 8 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.001 +recon_y_weight: 0.5 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.00075 +sw_s_y_weight: 7.5e-05 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 60 +vtjepa_kl_x_end_weight: 1.0e-06 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 20 +vtjepa_kl_y_end_weight: 1.0e-06 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 60 +vtjepa_kl_z_end_weight: 1.0e-05 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 60 +wjepa_joint_end_weight: 0.001875 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..5bd4a135e29a0b5c7ad0cefcdddee3bfc8071101 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-15T22:54:27.861490+00:00", + "spec": { + "dataset": "covertype", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.0015, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.00015, + "lambda_joint": 0.00375, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..2b92d8160d80feae7cf306afb0c9c1a06613d375 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "covertype", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.0015, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.00015, + "lambda_joint": 0.00375, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "covertype", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.00375", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.0015", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "0.00015", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..564c9bc35e252c8f939be3b15764b878aeb05b12 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "covertype", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.00375, + "s_x": 0.0015, + "z": 0.0, + "s_y": 0.00015 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 60 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "covertype", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "64", + "--model_num_layers", + "8", + "--model_num_heads", + "8", + "--model_dropout_prob", + "0.0015", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.15", + "--mask_max_ctx_share", + "0.35", + "--mask_min_trgt_share", + "0.15", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.0005", + "--exp_warmup", + "12", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "1e-06", + "--vtjepa_kl_x_annealing_epochs", + "60", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-05", + "--vtjepa_kl_z_annealing_epochs", + "60", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-06", + "--vtjepa_kl_y_annealing_epochs", + "20", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.001", + "--recon_y_weight", + "0.5", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.00375", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "60", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.0015", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "0.00015", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/downstream_results/autoint_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/downstream_results/autoint_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..737685323b794ecefd5fad48da97f647892bd7f3 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/downstream_results/autoint_covertype_downstream_results.json @@ -0,0 +1,210 @@ +{ + "dataset": "covertype", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpsz85vgj_", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpsz85vgj_/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.6166302561759949, + "lr": 0.0, + "covertype_val_accuracy": 0.7437437176704407, + "covertype_val_precision": 0.7459841370582581, + "covertype_val_recall": 0.7437437176704407, + "covertype_val_f1-score": 0.7415215373039246, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7437437176704407 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5933197140693665, + "lr": 0.0, + "covertype_test_accuracy": 0.7515000104904175, + "covertype_test_precision": 0.7529565095901489, + "covertype_test_recall": 0.7515000104904175, + "covertype_test_f1-score": 0.7492609620094299, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7515000104904175 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.6002148389816284, + "lr": 0.0, + "covertype_val_accuracy": 0.759759783744812, + "covertype_val_precision": 0.7612782716751099, + "covertype_val_recall": 0.759759783744812, + "covertype_val_f1-score": 0.7569073438644409, + "covertype_val_support": 999.0, + "covertype_val_score": 0.759759783744812 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5780664086341858, + "lr": 0.0, + "covertype_test_accuracy": 0.7540000081062317, + "covertype_test_precision": 0.7527186870574951, + "covertype_test_recall": 0.7540000081062317, + "covertype_test_f1-score": 0.7505075931549072, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7540000081062317 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.6231839656829834, + "lr": 0.0, + "covertype_val_accuracy": 0.7347347140312195, + "covertype_val_precision": 0.7358850836753845, + "covertype_val_recall": 0.7347347140312195, + "covertype_val_f1-score": 0.7316188216209412, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7347347140312195 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5840855240821838, + "lr": 0.0, + "covertype_test_accuracy": 0.7559999823570251, + "covertype_test_precision": 0.7568144798278809, + "covertype_test_recall": 0.7559999823570251, + "covertype_test_f1-score": 0.754235565662384, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7559999823570251 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.5912851095199585, + "lr": 0.0, + "covertype_val_accuracy": 0.7487487196922302, + "covertype_val_precision": 0.7454690933227539, + "covertype_val_recall": 0.7487487196922302, + "covertype_val_f1-score": 0.7446185946464539, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7487487196922302 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5973242521286011, + "lr": 0.0, + "covertype_test_accuracy": 0.7559999823570251, + "covertype_test_precision": 0.7578176856040955, + "covertype_test_recall": 0.7559999823570251, + "covertype_test_f1-score": 0.7546535134315491, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7559999823570251 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.6150115728378296, + "lr": 1.5708421415183693e-05, + "covertype_val_accuracy": 0.7497497200965881, + "covertype_val_precision": 0.7459551692008972, + "covertype_val_recall": 0.7497497200965881, + "covertype_val_f1-score": 0.7455016374588013, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7497497200965881 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5829192996025085, + "lr": 1.5708415958215483e-05, + "covertype_test_accuracy": 0.7570000290870667, + "covertype_test_precision": 0.7567775249481201, + "covertype_test_recall": 0.7570000290870667, + "covertype_test_f1-score": 0.7548285126686096, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7570000290870667 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5871430397033691, + "std": 0.007927515475398644, + "n": 5, + "ci95": 0.009841732532797609 + }, + "lr": { + "mean": 3.1416831916430967e-06, + "std": 7.025017180282464e-06, + "n": 5, + "ci95": 8.721312540001236e-06 + }, + "covertype_test_accuracy": { + "mean": 0.7549000024795532, + "std": 0.0021908878686589766, + "n": 5, + "ci95": 0.0027199104788386917 + }, + "covertype_test_precision": { + "mean": 0.7554169774055481, + "std": 0.002392810855646987, + "n": 5, + "ci95": 0.0029705907879880897 + }, + "covertype_test_recall": { + "mean": 0.7549000024795532, + "std": 0.0021908878686589766, + "n": 5, + "ci95": 0.0027199104788386917 + }, + "covertype_test_f1-score": { + "mean": 0.752697229385376, + "std": 0.002614303659898467, + "n": 5, + "ci95": 0.003245566339173974 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.7549000024795532, + "std": 0.0021908878686589766, + "n": 5, + "ci95": 0.0027199104788386917 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/downstream_results/dcnv2_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/downstream_results/dcnv2_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..8e6ca304cbad8f56bc1156db2da3f23885cf3a9c --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/downstream_results/dcnv2_covertype_downstream_results.json @@ -0,0 +1,209 @@ +{ + "dataset": "covertype", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpc6g8lzqd", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpc6g8lzqd/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.5730653405189514, + "lr": 0.0, + "covertype_val_accuracy": 0.7657657861709595, + "covertype_val_precision": 0.7608907222747803, + "covertype_val_recall": 0.7657657861709595, + "covertype_val_f1-score": 0.7612565755844116, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7657657861709595 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5517092347145081, + "lr": 0.0, + "covertype_test_accuracy": 0.7839999794960022, + "covertype_test_precision": 0.7848765850067139, + "covertype_test_recall": 0.7839999794960022, + "covertype_test_f1-score": 0.7813963294029236, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7839999794960022 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.5603543519973755, + "lr": 0.0, + "covertype_val_accuracy": 0.7797797918319702, + "covertype_val_precision": 0.7789238691329956, + "covertype_val_recall": 0.7797797918319702, + "covertype_val_f1-score": 0.7759314775466919, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7797797918319702 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5692827701568604, + "lr": 0.0, + "covertype_test_accuracy": 0.781000018119812, + "covertype_test_precision": 0.7803807854652405, + "covertype_test_recall": 0.781000018119812, + "covertype_test_f1-score": 0.7794114947319031, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.781000018119812 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.5779673457145691, + "lr": 0.0, + "covertype_val_accuracy": 0.7767767906188965, + "covertype_val_precision": 0.7709806561470032, + "covertype_val_recall": 0.7767767906188965, + "covertype_val_f1-score": 0.7719776630401611, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7767767906188965 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5792982578277588, + "lr": 0.0, + "covertype_test_accuracy": 0.7760000228881836, + "covertype_test_precision": 0.7756079435348511, + "covertype_test_recall": 0.7760000228881836, + "covertype_test_f1-score": 0.773810625076294, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7760000228881836 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.5650310516357422, + "lr": 0.0, + "covertype_val_accuracy": 0.7637637853622437, + "covertype_val_precision": 0.76324862241745, + "covertype_val_recall": 0.7637637853622437, + "covertype_val_f1-score": 0.7601956725120544, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7637637853622437 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5651774406433105, + "lr": 0.0, + "covertype_test_accuracy": 0.7795000076293945, + "covertype_test_precision": 0.7797838449478149, + "covertype_test_recall": 0.7795000076293945, + "covertype_test_f1-score": 0.777758002281189, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7795000076293945 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.5586784482002258, + "lr": 0.0, + "covertype_val_accuracy": 0.7767767906188965, + "covertype_val_precision": 0.7723803520202637, + "covertype_val_recall": 0.7767767906188965, + "covertype_val_f1-score": 0.7716455459594727, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7767767906188965 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5723748803138733, + "lr": 0.0, + "covertype_test_accuracy": 0.7770000100135803, + "covertype_test_precision": 0.7761189341545105, + "covertype_test_recall": 0.7770000100135803, + "covertype_test_f1-score": 0.7743250131607056, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7770000100135803 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5675685167312622, + "std": 0.010257444769916893, + "n": 5, + "ci95": 0.012734258067202239 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.7795000076293945, + "std": 0.0032015488259360147, + "n": 5, + "ci95": 0.003974610624644667 + }, + "covertype_test_precision": { + "mean": 0.7793536186218262, + "std": 0.003749552958842632, + "n": 5, + "ci95": 0.004654938543230457 + }, + "covertype_test_recall": { + "mean": 0.7795000076293945, + "std": 0.0032015488259360147, + "n": 5, + "ci95": 0.003974610624644667 + }, + "covertype_test_f1-score": { + "mean": 0.777340292930603, + "std": 0.0032583069653029104, + "n": 5, + "ci95": 0.004045073864791183 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.7795000076293945, + "std": 0.0032015488259360147, + "n": 5, + "ci95": 0.003974610624644667 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/downstream_results/fttransformer_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/downstream_results/fttransformer_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..00844e7539960775a2234ff84e57bd7090ff0026 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/downstream_results/fttransformer_covertype_downstream_results.json @@ -0,0 +1,209 @@ +{ + "dataset": "covertype", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpv4wq1cgs", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpv4wq1cgs/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.5849657654762268, + "lr": 0.0, + "covertype_val_accuracy": 0.7477477192878723, + "covertype_val_precision": 0.746700644493103, + "covertype_val_recall": 0.7477477192878723, + "covertype_val_f1-score": 0.7420284152030945, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7477477192878723 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5693378448486328, + "lr": 0.0, + "covertype_test_accuracy": 0.75, + "covertype_test_precision": 0.7527173757553101, + "covertype_test_recall": 0.75, + "covertype_test_f1-score": 0.7458217740058899, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.75 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.6161307692527771, + "lr": 0.0, + "covertype_val_accuracy": 0.7477477192878723, + "covertype_val_precision": 0.7467359304428101, + "covertype_val_recall": 0.7477477192878723, + "covertype_val_f1-score": 0.7425765991210938, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7477477192878723 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5970577597618103, + "lr": 0.0, + "covertype_test_accuracy": 0.7459999918937683, + "covertype_test_precision": 0.7488184571266174, + "covertype_test_recall": 0.7459999918937683, + "covertype_test_f1-score": 0.7420924305915833, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7459999918937683 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.6073358654975891, + "lr": 0.0, + "covertype_val_accuracy": 0.7457457184791565, + "covertype_val_precision": 0.7434555888175964, + "covertype_val_recall": 0.7457457184791565, + "covertype_val_f1-score": 0.7382279634475708, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7457457184791565 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5947052240371704, + "lr": 0.0, + "covertype_test_accuracy": 0.7450000047683716, + "covertype_test_precision": 0.7475780844688416, + "covertype_test_recall": 0.7450000047683716, + "covertype_test_f1-score": 0.7369008660316467, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7450000047683716 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.6109738945960999, + "lr": 0.0, + "covertype_val_accuracy": 0.7427427172660828, + "covertype_val_precision": 0.740642249584198, + "covertype_val_recall": 0.7427427172660828, + "covertype_val_f1-score": 0.7365875840187073, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7427427172660828 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5998197197914124, + "lr": 0.0, + "covertype_test_accuracy": 0.737500011920929, + "covertype_test_precision": 0.7341545820236206, + "covertype_test_recall": 0.737500011920929, + "covertype_test_f1-score": 0.7333523631095886, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.737500011920929 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.6269732713699341, + "lr": 0.0, + "covertype_val_accuracy": 0.7267267107963562, + "covertype_val_precision": 0.7245329022407532, + "covertype_val_recall": 0.7267267107963562, + "covertype_val_f1-score": 0.7202314138412476, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7267267107963562 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5938042998313904, + "lr": 0.0, + "covertype_test_accuracy": 0.7369999885559082, + "covertype_test_precision": 0.7367392182350159, + "covertype_test_recall": 0.7369999885559082, + "covertype_test_f1-score": 0.7300902009010315, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7369999885559082 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5909449696540833, + "std": 0.012301529552741051, + "n": 5, + "ci95": 0.015271917661730552 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.7430999994277954, + "std": 0.005661271438633369, + "n": 5, + "ci95": 0.007028270013158717 + }, + "covertype_test_precision": { + "mean": 0.7440015435218811, + "std": 0.008088016639796877, + "n": 5, + "ci95": 0.010040989101405008 + }, + "covertype_test_recall": { + "mean": 0.7430999994277954, + "std": 0.005661271438633369, + "n": 5, + "ci95": 0.007028270013158717 + }, + "covertype_test_f1-score": { + "mean": 0.737651526927948, + "std": 0.006377588009348797, + "n": 5, + "ci95": 0.007917551922436546 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.7430999994277954, + "std": 0.005661271438633369, + "n": 5, + "ci95": 0.007028270013158717 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/downstream_results/mlp_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/downstream_results/mlp_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..cc1312bcda79681c1212e38fdce45b7a1d719c05 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/downstream_results/mlp_covertype_downstream_results.json @@ -0,0 +1,205 @@ +{ + "dataset": "covertype", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpv2mere6i", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpv2mere6i/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.6387186050415039, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.7297297120094299, + "covertype_val_precision": 0.6934195756912231, + "covertype_val_recall": 0.7297297120094299, + "covertype_val_f1-score": 0.7064539194107056, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7297297120094299 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.6415005922317505, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.7179999947547913, + "covertype_test_precision": 0.7217504978179932, + "covertype_test_recall": 0.7179999947547913, + "covertype_test_f1-score": 0.6974831223487854, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7179999947547913 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.9935271143913269, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.586586594581604, + "covertype_val_precision": 0.6438586115837097, + "covertype_val_recall": 0.586586594581604, + "covertype_val_f1-score": 0.5666596293449402, + "covertype_val_support": 999.0, + "covertype_val_score": 0.586586594581604 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.9696019887924194, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.5770000219345093, + "covertype_test_precision": 0.6321200728416443, + "covertype_test_recall": 0.5770000219345093, + "covertype_test_f1-score": 0.550093412399292, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.5770000219345093 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.6425489187240601, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.7267267107963562, + "covertype_val_precision": 0.7109326720237732, + "covertype_val_recall": 0.7267267107963562, + "covertype_val_f1-score": 0.7078247666358948, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7267267107963562 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.6463703513145447, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.7210000157356262, + "covertype_test_precision": 0.7132411003112793, + "covertype_test_recall": 0.7210000157356262, + "covertype_test_f1-score": 0.7004935145378113, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7210000157356262 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.6805802583694458, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.7167167067527771, + "covertype_val_precision": 0.7143189907073975, + "covertype_val_recall": 0.7167167067527771, + "covertype_val_f1-score": 0.6992518305778503, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7167167067527771 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.6696910262107849, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.7074999809265137, + "covertype_test_precision": 0.7030519247055054, + "covertype_test_recall": 0.7074999809265137, + "covertype_test_f1-score": 0.6925882697105408, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7074999809265137 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 1.6453590393066406, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.5875875949859619, + "covertype_val_precision": 0.5513067841529846, + "covertype_val_recall": 0.5875875949859619, + "covertype_val_f1-score": 0.5075923204421997, + "covertype_val_support": 999.0, + "covertype_val_score": 0.5875875949859619 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 1.5949995517730713, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.5835000276565552, + "covertype_test_precision": 0.5502362251281738, + "covertype_test_recall": 0.5835000276565552, + "covertype_test_f1-score": 0.5036627650260925, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.5835000276565552 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.9044327020645142, + "std": 0.40986666596813864, + "n": 5, + "ci95": 0.5088350963282197 + }, + "lr": { + "mean": 0.0010000001639127731, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.6614000082015992, + "std": 0.07428440534849604, + "n": 5, + "ci95": 0.09222148491120476 + }, + "covertype_test_precision": { + "mean": 0.6640799641609192, + "std": 0.07287549834936023, + "n": 5, + "ci95": 0.09047237626649612 + }, + "covertype_test_recall": { + "mean": 0.6614000082015992, + "std": 0.07428440534849604, + "n": 5, + "ci95": 0.09222148491120476 + }, + "covertype_test_f1-score": { + "mean": 0.6288642168045044, + "std": 0.09457841794603464, + "n": 5, + "ci95": 0.11741579006545065 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.6614000082015992, + "std": 0.07428440534849604, + "n": 5, + "ci95": 0.09222148491120476 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/downstream_results/resnet_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/downstream_results/resnet_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..c45f2439915638ed8aef790493351b6fe3276efb --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/downstream_results/resnet_covertype_downstream_results.json @@ -0,0 +1,208 @@ +{ + "dataset": "covertype", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpy4lw0gmr", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpy4lw0gmr/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.5973950624465942, + "lr": 0.0, + "covertype_val_accuracy": 0.7627627849578857, + "covertype_val_precision": 0.7563494443893433, + "covertype_val_recall": 0.7627627849578857, + "covertype_val_f1-score": 0.7588021755218506, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7627627849578857 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5636202096939087, + "lr": 0.0, + "covertype_test_accuracy": 0.781499981880188, + "covertype_test_precision": 0.7789103984832764, + "covertype_test_recall": 0.781499981880188, + "covertype_test_f1-score": 0.7787503600120544, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.781499981880188 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.6209166646003723, + "lr": 0.0, + "covertype_val_accuracy": 0.7697697877883911, + "covertype_val_precision": 0.7628315687179565, + "covertype_val_recall": 0.7697697877883911, + "covertype_val_f1-score": 0.7633577585220337, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7697697877883911 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5634870529174805, + "lr": 0.0, + "covertype_test_accuracy": 0.7850000262260437, + "covertype_test_precision": 0.7845066785812378, + "covertype_test_recall": 0.7850000262260437, + "covertype_test_f1-score": 0.7828440070152283, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7850000262260437 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.629848837852478, + "lr": 0.0, + "covertype_val_accuracy": 0.7547547817230225, + "covertype_val_precision": 0.75566565990448, + "covertype_val_recall": 0.7547547817230225, + "covertype_val_f1-score": 0.7510293126106262, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7547547817230225 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5653629899024963, + "lr": 0.0, + "covertype_test_accuracy": 0.7764999866485596, + "covertype_test_precision": 0.7795354127883911, + "covertype_test_recall": 0.7764999866485596, + "covertype_test_f1-score": 0.7737704515457153, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7764999866485596 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.5866146683692932, + "lr": 0.0, + "covertype_val_accuracy": 0.7697697877883911, + "covertype_val_precision": 0.7659818530082703, + "covertype_val_recall": 0.7697697877883911, + "covertype_val_f1-score": 0.7665532827377319, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7697697877883911 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5801534056663513, + "lr": 0.0, + "covertype_test_accuracy": 0.7835000157356262, + "covertype_test_precision": 0.7823936343193054, + "covertype_test_recall": 0.7835000157356262, + "covertype_test_f1-score": 0.782187819480896, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7835000157356262 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.5958480834960938, + "lr": 0.0, + "covertype_val_accuracy": 0.7587587833404541, + "covertype_val_precision": 0.7560653686523438, + "covertype_val_recall": 0.7587587833404541, + "covertype_val_f1-score": 0.756770670413971, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7587587833404541 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5633795261383057, + "lr": 0.0, + "covertype_test_accuracy": 0.7735000252723694, + "covertype_test_precision": 0.7790331244468689, + "covertype_test_recall": 0.7735000252723694, + "covertype_test_f1-score": 0.775626540184021, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7735000252723694 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5672006368637085, + "std": 0.007286326537228502, + "n": 5, + "ci95": 0.009045718945433188 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.7800000071525574, + "std": 0.004847681996879808, + "n": 5, + "ci95": 0.006018227244766136 + }, + "covertype_test_precision": { + "mean": 0.780875849723816, + "std": 0.0024769938672351234, + "n": 5, + "ci95": 0.0030751010455116403 + }, + "covertype_test_recall": { + "mean": 0.7800000071525574, + "std": 0.004847681996879808, + "n": 5, + "ci95": 0.006018227244766136 + }, + "covertype_test_f1-score": { + "mean": 0.778635835647583, + "std": 0.00397071255845111, + "n": 5, + "ci95": 0.004929500432533839 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.7800000071525574, + "std": 0.004847681996879808, + "n": 5, + "ci95": 0.006018227244766136 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..cd2b2a92aca1f9c641aabf5461c8f3d82443356d --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:415261ebf9be93e49c582fd844b48f2b53172994ae0693ad4fe69d841abb1cbf +size 138240128 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..a8f2c42297e98cf605a72b8a2b69bd54fde2fd2d --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:28bc0e2da890f7418b820a71e6156816f89c509a09a4f08ff70fe0e7af81f1b6 +size 10879822 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3f6a2ac891cda9eaf3a5ff322a87884ca0d6efea --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.00375_sx0.0015_sy0.00015/seed_42/wjepa_deterministic/covertype/20260915_222933/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 +data_set: covertype +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.0005 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 12 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.35 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.15 +mask_min_trgt_share: 0.15 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 64 +model_dim_hidden: 64 +model_dropout_prob: 0.0015 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 8 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.001 +recon_y_weight: 0.5 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.0015 +sw_s_y_weight: 0.00015 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 60 +vtjepa_kl_x_end_weight: 1.0e-06 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 20 +vtjepa_kl_y_end_weight: 1.0e-06 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 60 +vtjepa_kl_z_end_weight: 1.0e-05 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 60 +wjepa_joint_end_weight: 0.00375 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..a23f0f2670de7c59458590e9ef6d575beefa9f67 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-15T22:15:02.069480+00:00", + "spec": { + "dataset": "covertype", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.003, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0003, + "lambda_joint": 0.0075, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..110493d41a5e56940654415e3889765c9bac69c2 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "covertype", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.003, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0003, + "lambda_joint": 0.0075, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "covertype", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.0075", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.003", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "0.0003", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..48fb1d924bf4ea920ab10b64b96fd1339e87dbd7 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "covertype", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.0075, + "s_x": 0.003, + "z": 0.0, + "s_y": 0.0003 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 60 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "covertype", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "64", + "--model_num_layers", + "8", + "--model_num_heads", + "8", + "--model_dropout_prob", + "0.0015", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.15", + "--mask_max_ctx_share", + "0.35", + "--mask_min_trgt_share", + "0.15", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.0005", + "--exp_warmup", + "12", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "1e-06", + "--vtjepa_kl_x_annealing_epochs", + "60", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-05", + "--vtjepa_kl_z_annealing_epochs", + "60", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-06", + "--vtjepa_kl_y_annealing_epochs", + "20", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.001", + "--recon_y_weight", + "0.5", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.0075", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "60", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.003", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "0.0003", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/downstream_results/autoint_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/downstream_results/autoint_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..fd7379817d2773686a22280ac1a2224dee179b63 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/downstream_results/autoint_covertype_downstream_results.json @@ -0,0 +1,210 @@ +{ + "dataset": "covertype", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp2o4diecz", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp2o4diecz/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.6227709054946899, + "lr": 9.866357686405536e-07, + "covertype_val_accuracy": 0.7537537813186646, + "covertype_val_precision": 0.7508715987205505, + "covertype_val_recall": 0.7537537813186646, + "covertype_val_f1-score": 0.7492019534111023, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7537537813186646 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5965843796730042, + "lr": 9.866358823273913e-07, + "covertype_test_accuracy": 0.7649999856948853, + "covertype_test_precision": 0.7649998068809509, + "covertype_test_recall": 0.7649999856948853, + "covertype_test_f1-score": 0.7608588933944702, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7649999856948853 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.591583788394928, + "lr": 0.0, + "covertype_val_accuracy": 0.7557557821273804, + "covertype_val_precision": 0.7502531409263611, + "covertype_val_recall": 0.7557557821273804, + "covertype_val_f1-score": 0.750498354434967, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7557557821273804 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.6003732085227966, + "lr": 0.0, + "covertype_test_accuracy": 0.7515000104904175, + "covertype_test_precision": 0.7503300905227661, + "covertype_test_recall": 0.7515000104904175, + "covertype_test_f1-score": 0.7487834095954895, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7515000104904175 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.6020187139511108, + "lr": 0.0, + "covertype_val_accuracy": 0.7527527809143066, + "covertype_val_precision": 0.7502567768096924, + "covertype_val_recall": 0.7527527809143066, + "covertype_val_f1-score": 0.7494097948074341, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7527527809143066 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5802618265151978, + "lr": 0.0, + "covertype_test_accuracy": 0.7450000047683716, + "covertype_test_precision": 0.7470070123672485, + "covertype_test_recall": 0.7450000047683716, + "covertype_test_f1-score": 0.7440933585166931, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7450000047683716 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.6219367980957031, + "lr": 0.0, + "covertype_val_accuracy": 0.7427427172660828, + "covertype_val_precision": 0.7471427321434021, + "covertype_val_recall": 0.7427427172660828, + "covertype_val_f1-score": 0.741443395614624, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7427427172660828 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.6044445633888245, + "lr": 0.0, + "covertype_test_accuracy": 0.7609999775886536, + "covertype_test_precision": 0.7597412467002869, + "covertype_test_recall": 0.7609999775886536, + "covertype_test_f1-score": 0.7585837244987488, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7609999775886536 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.6137548685073853, + "lr": 3.9426490729965735e-06, + "covertype_val_accuracy": 0.739739716053009, + "covertype_val_precision": 0.7351955771446228, + "covertype_val_recall": 0.739739716053009, + "covertype_val_f1-score": 0.7362640500068665, + "covertype_val_support": 999.0, + "covertype_val_score": 0.739739716053009 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5827392339706421, + "lr": 3.9426490729965735e-06, + "covertype_test_accuracy": 0.7544999718666077, + "covertype_test_precision": 0.7573593258857727, + "covertype_test_recall": 0.7544999718666077, + "covertype_test_f1-score": 0.7537063956260681, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7544999718666077 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.592880642414093, + "std": 0.010789614969882892, + "n": 5, + "ci95": 0.013394928713162398 + }, + "lr": { + "mean": 9.85856991064793e-07, + "std": 1.7072171832356343e-06, + "n": 5, + "ci95": 2.119450279843993e-06 + }, + "covertype_test_accuracy": { + "mean": 0.7553999900817872, + "std": 0.007869233507282922, + "n": 5, + "ci95": 0.009769377512683173 + }, + "covertype_test_precision": { + "mean": 0.7558874964714051, + "std": 0.007242230879598955, + "n": 5, + "ci95": 0.008990975732431008 + }, + "covertype_test_recall": { + "mean": 0.7553999900817872, + "std": 0.007869233507282922, + "n": 5, + "ci95": 0.009769377512683173 + }, + "covertype_test_f1-score": { + "mean": 0.7532051563262939, + "std": 0.006898127119091529, + "n": 5, + "ci95": 0.008563782977657655 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.7553999900817872, + "std": 0.007869233507282922, + "n": 5, + "ci95": 0.009769377512683173 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/downstream_results/dcnv2_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/downstream_results/dcnv2_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..e5d50031c8e6ca858004b5b885f54de1bfef38ed --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/downstream_results/dcnv2_covertype_downstream_results.json @@ -0,0 +1,209 @@ +{ + "dataset": "covertype", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp0j1q8roj", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp0j1q8roj/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.5677909255027771, + "lr": 0.0, + "covertype_val_accuracy": 0.7617617845535278, + "covertype_val_precision": 0.7567856907844543, + "covertype_val_recall": 0.7617617845535278, + "covertype_val_f1-score": 0.756018340587616, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7617617845535278 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5584063529968262, + "lr": 0.0, + "covertype_test_accuracy": 0.7689999938011169, + "covertype_test_precision": 0.7696625590324402, + "covertype_test_recall": 0.7689999938011169, + "covertype_test_f1-score": 0.7662068605422974, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7689999938011169 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.5856038331985474, + "lr": 0.0, + "covertype_val_accuracy": 0.7627627849578857, + "covertype_val_precision": 0.7552188038825989, + "covertype_val_recall": 0.7627627849578857, + "covertype_val_f1-score": 0.755889892578125, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7627627849578857 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5876356363296509, + "lr": 0.0, + "covertype_test_accuracy": 0.7749999761581421, + "covertype_test_precision": 0.7760087847709656, + "covertype_test_recall": 0.7749999761581421, + "covertype_test_f1-score": 0.773255467414856, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7749999761581421 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.5700198411941528, + "lr": 0.0, + "covertype_val_accuracy": 0.770770788192749, + "covertype_val_precision": 0.7632765173912048, + "covertype_val_recall": 0.770770788192749, + "covertype_val_f1-score": 0.7628383040428162, + "covertype_val_support": 999.0, + "covertype_val_score": 0.770770788192749 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5609036684036255, + "lr": 0.0, + "covertype_test_accuracy": 0.7825000286102295, + "covertype_test_precision": 0.7830548286437988, + "covertype_test_recall": 0.7825000286102295, + "covertype_test_f1-score": 0.7802314758300781, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7825000286102295 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.5688031315803528, + "lr": 0.0, + "covertype_val_accuracy": 0.7657657861709595, + "covertype_val_precision": 0.7652583122253418, + "covertype_val_recall": 0.7657657861709595, + "covertype_val_f1-score": 0.762113630771637, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7657657861709595 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5489407777786255, + "lr": 0.0, + "covertype_test_accuracy": 0.777999997138977, + "covertype_test_precision": 0.7770349383354187, + "covertype_test_recall": 0.777999997138977, + "covertype_test_f1-score": 0.7743245959281921, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.777999997138977 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.5739994049072266, + "lr": 0.0, + "covertype_val_accuracy": 0.7647647857666016, + "covertype_val_precision": 0.7608082294464111, + "covertype_val_recall": 0.7647647857666016, + "covertype_val_f1-score": 0.7588103413581848, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7647647857666016 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5534422397613525, + "lr": 0.0, + "covertype_test_accuracy": 0.7774999737739563, + "covertype_test_precision": 0.7763988971710205, + "covertype_test_recall": 0.7774999737739563, + "covertype_test_f1-score": 0.7740729451179504, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7774999737739563 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5618657350540162, + "std": 0.01512440496126676, + "n": 5, + "ci95": 0.018776418514530813 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.7763999938964844, + "std": 0.004942176663547063, + "n": 5, + "ci95": 0.006135539060555209 + }, + "covertype_test_precision": { + "mean": 0.7764320015907288, + "std": 0.004749498179336353, + "n": 5, + "ci95": 0.005896335477501803 + }, + "covertype_test_recall": { + "mean": 0.7763999938964844, + "std": 0.004942176663547063, + "n": 5, + "ci95": 0.006135539060555209 + }, + "covertype_test_f1-score": { + "mean": 0.7736182689666748, + "std": 0.004987502706671069, + "n": 5, + "ci95": 0.0061918097540128066 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.7763999938964844, + "std": 0.004942176663547063, + "n": 5, + "ci95": 0.006135539060555209 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/downstream_results/fttransformer_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/downstream_results/fttransformer_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..12915aaf780c2cd09e4e0b7faf059606f8e170cf --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/downstream_results/fttransformer_covertype_downstream_results.json @@ -0,0 +1,209 @@ +{ + "dataset": "covertype", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmprt7900r3", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmprt7900r3/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.6131318211555481, + "lr": 0.0, + "covertype_val_accuracy": 0.7407407164573669, + "covertype_val_precision": 0.7346999049186707, + "covertype_val_recall": 0.7407407164573669, + "covertype_val_f1-score": 0.7344666719436646, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7407407164573669 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5921431183815002, + "lr": 0.0, + "covertype_test_accuracy": 0.7369999885559082, + "covertype_test_precision": 0.7380644679069519, + "covertype_test_recall": 0.7369999885559082, + "covertype_test_f1-score": 0.7324556708335876, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7369999885559082 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.616991400718689, + "lr": 0.0, + "covertype_val_accuracy": 0.728728711605072, + "covertype_val_precision": 0.7257769703865051, + "covertype_val_recall": 0.728728711605072, + "covertype_val_f1-score": 0.722662091255188, + "covertype_val_support": 999.0, + "covertype_val_score": 0.728728711605072 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5900807976722717, + "lr": 0.0, + "covertype_test_accuracy": 0.7409999966621399, + "covertype_test_precision": 0.748101532459259, + "covertype_test_recall": 0.7409999966621399, + "covertype_test_f1-score": 0.7350011467933655, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7409999966621399 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.6099487543106079, + "lr": 0.0, + "covertype_val_accuracy": 0.7477477192878723, + "covertype_val_precision": 0.7455762624740601, + "covertype_val_recall": 0.7477477192878723, + "covertype_val_f1-score": 0.7413235902786255, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7477477192878723 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.591465175151825, + "lr": 0.0, + "covertype_test_accuracy": 0.7400000095367432, + "covertype_test_precision": 0.7382622361183167, + "covertype_test_recall": 0.7400000095367432, + "covertype_test_f1-score": 0.7344686985015869, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7400000095367432 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.6159301400184631, + "lr": 0.0, + "covertype_val_accuracy": 0.7447447180747986, + "covertype_val_precision": 0.737123429775238, + "covertype_val_recall": 0.7447447180747986, + "covertype_val_f1-score": 0.7358610033988953, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7447447180747986 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5801007747650146, + "lr": 0.0, + "covertype_test_accuracy": 0.7540000081062317, + "covertype_test_precision": 0.7565879225730896, + "covertype_test_recall": 0.7540000081062317, + "covertype_test_f1-score": 0.7489565014839172, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7540000081062317 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.595604658126831, + "lr": 0.0, + "covertype_val_accuracy": 0.7527527809143066, + "covertype_val_precision": 0.7453463673591614, + "covertype_val_recall": 0.7527527809143066, + "covertype_val_f1-score": 0.7450771927833557, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7527527809143066 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5832383632659912, + "lr": 0.0, + "covertype_test_accuracy": 0.75, + "covertype_test_precision": 0.7495657205581665, + "covertype_test_recall": 0.75, + "covertype_test_f1-score": 0.745711088180542, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.75 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5874056458473206, + "std": 0.005403870074181478, + "n": 5, + "ci95": 0.00670871524339836 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.7444000005722046, + "std": 0.007231878452148741, + "n": 5, + "ci95": 0.008978123556696206 + }, + "covertype_test_precision": { + "mean": 0.7461163759231567, + "std": 0.007937456281785394, + "n": 5, + "ci95": 0.009854073695413102 + }, + "covertype_test_recall": { + "mean": 0.7444000005722046, + "std": 0.007231878452148741, + "n": 5, + "ci95": 0.008978123556696206 + }, + "covertype_test_f1-score": { + "mean": 0.7393186211585998, + "std": 0.007466839921091223, + "n": 5, + "ci95": 0.009269819982899505 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.7444000005722046, + "std": 0.007231878452148741, + "n": 5, + "ci95": 0.008978123556696206 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/downstream_results/mlp_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/downstream_results/mlp_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..42d3f1d51b0299b197d1129d276fc224999054fd --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/downstream_results/mlp_covertype_downstream_results.json @@ -0,0 +1,205 @@ +{ + "dataset": "covertype", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpdrv786q6", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpdrv786q6/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.767726719379425, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.6486486196517944, + "covertype_val_precision": 0.6495037078857422, + "covertype_val_recall": 0.6486486196517944, + "covertype_val_f1-score": 0.5912317633628845, + "covertype_val_support": 999.0, + "covertype_val_score": 0.6486486196517944 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.7830761671066284, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.6359999775886536, + "covertype_test_precision": 0.6079962849617004, + "covertype_test_recall": 0.6359999775886536, + "covertype_test_f1-score": 0.5740435123443604, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.6359999775886536 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.6654766798019409, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.7147147059440613, + "covertype_val_precision": 0.7096186280250549, + "covertype_val_recall": 0.7147147059440613, + "covertype_val_f1-score": 0.7017911076545715, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7147147059440613 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.6640293598175049, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.7020000219345093, + "covertype_test_precision": 0.7076472640037537, + "covertype_test_recall": 0.7020000219345093, + "covertype_test_f1-score": 0.6916620135307312, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7020000219345093 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.6501673460006714, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.7227227091789246, + "covertype_val_precision": 0.7046359777450562, + "covertype_val_recall": 0.7227227091789246, + "covertype_val_f1-score": 0.7084230780601501, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7227227091789246 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.6401070356369019, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.7200000286102295, + "covertype_test_precision": 0.7071603536605835, + "covertype_test_recall": 0.7200000286102295, + "covertype_test_f1-score": 0.7060540318489075, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7200000286102295 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 1.0306665897369385, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.7087087035179138, + "covertype_val_precision": 0.7124620079994202, + "covertype_val_recall": 0.7087087035179138, + "covertype_val_f1-score": 0.6889351010322571, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7087087035179138 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 1.03887939453125, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.6974999904632568, + "covertype_test_precision": 0.689578115940094, + "covertype_test_recall": 0.6974999904632568, + "covertype_test_f1-score": 0.6777116060256958, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.6974999904632568 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.656154990196228, + "lr": 0.0010000000474974513, + "covertype_val_accuracy": 0.7147147059440613, + "covertype_val_precision": 0.6897753477096558, + "covertype_val_recall": 0.7147147059440613, + "covertype_val_f1-score": 0.6910732388496399, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7147147059440613 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.6602410674095154, + "lr": 0.0010000001639127731, + "covertype_test_accuracy": 0.7110000252723694, + "covertype_test_precision": 0.7027683854103088, + "covertype_test_recall": 0.7110000252723694, + "covertype_test_f1-score": 0.693601667881012, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7110000252723694 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.7572666049003601, + "std": 0.16718657261499642, + "n": 5, + "ci95": 0.20755626852550532 + }, + "lr": { + "mean": 0.0010000001639127731, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.6933000087738037, + "std": 0.03317681906211926, + "n": 5, + "ci95": 0.04118785772310078 + }, + "covertype_test_precision": { + "mean": 0.683030080795288, + "std": 0.04257588439540114, + "n": 5, + "ci95": 0.05285646781355271 + }, + "covertype_test_recall": { + "mean": 0.6933000087738037, + "std": 0.03317681906211926, + "n": 5, + "ci95": 0.04118785772310078 + }, + "covertype_test_f1-score": { + "mean": 0.6686145663261414, + "std": 0.05381378547340907, + "n": 5, + "ci95": 0.06680792801353805 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.6933000087738037, + "std": 0.03317681906211926, + "n": 5, + "ci95": 0.04118785772310078 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/downstream_results/resnet_covertype_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/downstream_results/resnet_covertype_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..ae645f8b8476a605f31c2073e1c112b277c98ec6 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/downstream_results/resnet_covertype_downstream_results.json @@ -0,0 +1,208 @@ +{ + "dataset": "covertype", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpj4didhtk", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000", + "data_set": "covertype", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpj4didhtk/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "covertype_val_loss": 0.5862805843353271, + "lr": 0.0, + "covertype_val_accuracy": 0.7527527809143066, + "covertype_val_precision": 0.7509596347808838, + "covertype_val_recall": 0.7527527809143066, + "covertype_val_f1-score": 0.7511112093925476, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7527527809143066 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5744960904121399, + "lr": 0.0, + "covertype_test_accuracy": 0.781499981880188, + "covertype_test_precision": 0.7824848294258118, + "covertype_test_recall": 0.781499981880188, + "covertype_test_f1-score": 0.781628429889679, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.781499981880188 + } + } + }, + { + "seed": 1, + "val_metrics": { + "covertype_val_loss": 0.6381723284721375, + "lr": 0.0, + "covertype_val_accuracy": 0.7587587833404541, + "covertype_val_precision": 0.7568015456199646, + "covertype_val_recall": 0.7587587833404541, + "covertype_val_f1-score": 0.754927396774292, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7587587833404541 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5927181839942932, + "lr": 0.0, + "covertype_test_accuracy": 0.7695000171661377, + "covertype_test_precision": 0.770908772945404, + "covertype_test_recall": 0.7695000171661377, + "covertype_test_f1-score": 0.7675900459289551, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7695000171661377 + } + } + }, + { + "seed": 2, + "val_metrics": { + "covertype_val_loss": 0.6420478820800781, + "lr": 0.0, + "covertype_val_accuracy": 0.728728711605072, + "covertype_val_precision": 0.7312394976615906, + "covertype_val_recall": 0.728728711605072, + "covertype_val_f1-score": 0.7241413593292236, + "covertype_val_support": 999.0, + "covertype_val_score": 0.728728711605072 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5749556422233582, + "lr": 0.0, + "covertype_test_accuracy": 0.7684999704360962, + "covertype_test_precision": 0.7742021083831787, + "covertype_test_recall": 0.7684999704360962, + "covertype_test_f1-score": 0.7671692371368408, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7684999704360962 + } + } + }, + { + "seed": 3, + "val_metrics": { + "covertype_val_loss": 0.6067997813224792, + "lr": 0.0, + "covertype_val_accuracy": 0.759759783744812, + "covertype_val_precision": 0.75462406873703, + "covertype_val_recall": 0.759759783744812, + "covertype_val_f1-score": 0.7553516626358032, + "covertype_val_support": 999.0, + "covertype_val_score": 0.759759783744812 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5838314294815063, + "lr": 0.0, + "covertype_test_accuracy": 0.7749999761581421, + "covertype_test_precision": 0.7760921716690063, + "covertype_test_recall": 0.7749999761581421, + "covertype_test_f1-score": 0.7739266157150269, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7749999761581421 + } + } + }, + { + "seed": 4, + "val_metrics": { + "covertype_val_loss": 0.6377906799316406, + "lr": 0.0, + "covertype_val_accuracy": 0.7477477192878723, + "covertype_val_precision": 0.7473138570785522, + "covertype_val_recall": 0.7477477192878723, + "covertype_val_f1-score": 0.7456397414207458, + "covertype_val_support": 999.0, + "covertype_val_score": 0.7477477192878723 + }, + "test_metrics": { + "0": { + "covertype_test_loss": 0.5757411122322083, + "lr": 0.0, + "covertype_test_accuracy": 0.7714999914169312, + "covertype_test_precision": 0.7738337516784668, + "covertype_test_recall": 0.7714999914169312, + "covertype_test_f1-score": 0.770883321762085, + "covertype_test_support": 2000.0, + "covertype_test_score": 0.7714999914169312 + } + } + } + ], + "summary": { + "0": { + "covertype_test_loss": { + "mean": 0.5803484916687012, + "std": 0.007900973277347435, + "n": 5, + "ci95": 0.009808781324457092 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_accuracy": { + "mean": 0.773199987411499, + "std": 0.005263074040793314, + "n": 5, + "ci95": 0.0065339219040999 + }, + "covertype_test_precision": { + "mean": 0.7755043268203735, + "std": 0.00432084157807446, + "n": 5, + "ci95": 0.005364173335260701 + }, + "covertype_test_recall": { + "mean": 0.773199987411499, + "std": 0.005263074040793314, + "n": 5, + "ci95": 0.0065339219040999 + }, + "covertype_test_f1-score": { + "mean": 0.7722395300865174, + "std": 0.005919513746456569, + "n": 5, + "ci95": 0.007348868784632009 + }, + "covertype_test_support": { + "mean": 2000.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "covertype_test_score": { + "mean": 0.773199987411499, + "std": 0.005263074040793314, + "n": 5, + "ci95": 0.0065339219040999 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..44e589f70ecf23f0248c56e8d80d03779c711a84 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:de3222cc7d141b20b889dd3a91c69e093a39d129de376acbb3cd652d153d8bc7 +size 138240128 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..d6c0c85e3925b4efca63b45ae9bc72ea0d3f9c34 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3174dc03635649da55b3777aff69a585c37e4bd7d2a22a008b739807f67d7e1a +size 10878670 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..821c9e4a83f4577b79758303eadc4cd5d544fa91 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/covertype/deterministic_sw/sw_f0.0075_sx0.003_sy0.0003/seed_42/wjepa_deterministic/covertype/20260915_214757/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets/subset_views/size_10000 +data_set: covertype +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.0005 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 12 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.35 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.15 +mask_min_trgt_share: 0.15 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 64 +model_dim_hidden: 64 +model_dropout_prob: 0.0015 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 8 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.001 +recon_y_weight: 0.5 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.003 +sw_s_y_weight: 0.0003 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 60 +vtjepa_kl_x_end_weight: 1.0e-06 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 20 +vtjepa_kl_y_end_weight: 1.0e-06 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 60 +vtjepa_kl_z_end_weight: 1.0e-05 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 60 +wjepa_joint_end_weight: 0.0075 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..dec662a0facf30ae1ee0fcbbebfc6b84dc1a8e2a --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-16T00:17:13.957399+00:00", + "spec": { + "dataset": "credit_card_default", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.00025, + "sw_z_weight": 0.0, + "sw_s_y_weight": 2.5e-05, + "lambda_joint": 0.0003125, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..06fb71d0ead0edbe3027ba829bd62ae040140087 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "credit_card_default", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.00025, + "sw_z_weight": 0.0, + "sw_s_y_weight": 2.5e-05, + "lambda_joint": 0.0003125, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "credit_card_default", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.0003125", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.00025", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "2.5e-05", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..974666b0dddfe4a54457bdf4e883b94c0d3c2ae9 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "credit_card_default", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.0003125, + "s_x": 0.00025, + "z": 0.0, + "s_y": 2.5e-05 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 15 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "credit_card_default", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.0003125", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.00025", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "2.5e-05", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/downstream_results/autoint_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/downstream_results/autoint_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..e7f18da391db40b926373d6855cd6ba0d779ccf9 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/downstream_results/autoint_credit_card_default_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "credit_card_default", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpq2ugrlr4", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpq2ugrlr4/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.4481733441352844, + "lr": 0.00025912324781529605, + "credit_card_default_val_accuracy": 0.8180588483810425, + "credit_card_default_val_precision": 0.8024962544441223, + "credit_card_default_val_recall": 0.8180588483810425, + "credit_card_default_val_f1-score": 0.7902538776397705, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.752275824546814, + "credit_card_default_val_score": 0.8180588483810425 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4433985650539398, + "lr": 0.0002591230731923133, + "credit_card_default_test_accuracy": 0.8154801726341248, + "credit_card_default_test_precision": 0.798980712890625, + "credit_card_default_test_recall": 0.8154801726341248, + "credit_card_default_test_f1-score": 0.7905749082565308, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7656022310256958, + "credit_card_default_test_score": 0.8154801726341248 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.44931939244270325, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.816706120967865, + "credit_card_default_val_precision": 0.7991493940353394, + "credit_card_default_val_recall": 0.816706120967865, + "credit_card_default_val_f1-score": 0.7928685545921326, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7498471140861511, + "credit_card_default_val_score": 0.816706120967865 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4445647597312927, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.815144419670105, + "credit_card_default_test_precision": 0.7979071736335754, + "credit_card_default_test_recall": 0.815144419670105, + "credit_card_default_test_f1-score": 0.7944098114967346, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7626562714576721, + "credit_card_default_test_score": 0.815144419670105 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4451044499874115, + "lr": 1.5708415958215483e-05, + "credit_card_default_val_accuracy": 0.8150151968002319, + "credit_card_default_val_precision": 0.7968158721923828, + "credit_card_default_val_recall": 0.8150151968002319, + "credit_card_default_val_f1-score": 0.7904254198074341, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7556415796279907, + "credit_card_default_val_score": 0.8150151968002319 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4429336190223694, + "lr": 1.57084250531625e-05, + "credit_card_default_test_accuracy": 0.8169912695884705, + "credit_card_default_test_precision": 0.8004621863365173, + "credit_card_default_test_recall": 0.8169912695884705, + "credit_card_default_test_f1-score": 0.7954587340354919, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7641879320144653, + "credit_card_default_test_score": 0.8169912695884705 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.44777241349220276, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8136624693870544, + "credit_card_default_val_precision": 0.7947706580162048, + "credit_card_default_val_recall": 0.8136624693870544, + "credit_card_default_val_f1-score": 0.7905862331390381, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.753254771232605, + "credit_card_default_val_score": 0.8136624693870544 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4407835900783539, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8166554570198059, + "credit_card_default_test_precision": 0.7999688386917114, + "credit_card_default_test_recall": 0.8166554570198059, + "credit_card_default_test_f1-score": 0.7954699397087097, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7674849033355713, + "credit_card_default_test_score": 0.8166554570198059 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.44714176654815674, + "lr": 0.00011474337225081399, + "credit_card_default_val_accuracy": 0.8133243322372437, + "credit_card_default_val_precision": 0.794337272644043, + "credit_card_default_val_recall": 0.8133243322372437, + "credit_card_default_val_f1-score": 0.7913352251052856, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7535678148269653, + "credit_card_default_val_score": 0.8133243322372437 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4419805407524109, + "lr": 0.00011474342318251729, + "credit_card_default_test_accuracy": 0.8183344602584839, + "credit_card_default_test_precision": 0.8021625876426697, + "credit_card_default_test_recall": 0.8183344602584839, + "credit_card_default_test_f1-score": 0.7995673418045044, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7660477757453918, + "credit_card_default_test_score": 0.8183344602584839 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.44273221492767334, + "std": 0.0014322150175994754, + "n": 5, + "ci95": 0.0017780447324779584 + }, + "lr": { + "mean": 7.791498428559862e-05, + "std": 0.00011203107934943392, + "n": 5, + "ci95": 0.00013908265732679757 + }, + "credit_card_default_test_accuracy": { + "mean": 0.816521155834198, + "std": 0.001275359280154312, + "n": 5, + "ci95": 0.0015833138336281652 + }, + "credit_card_default_test_precision": { + "mean": 0.7998962998390198, + "std": 0.001601286384709838, + "n": 5, + "ci95": 0.001987940907290654 + }, + "credit_card_default_test_recall": { + "mean": 0.816521155834198, + "std": 0.001275359280154312, + "n": 5, + "ci95": 0.0015833138336281652 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7950961470603943, + "std": 0.0032084041798730836, + "n": 5, + "ci95": 0.003983121306216424 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7651958227157593, + "std": 0.0018436408287198583, + "n": 5, + "ci95": 0.0022888154528507877 + }, + "credit_card_default_test_score": { + "mean": 0.816521155834198, + "std": 0.001275359280154312, + "n": 5, + "ci95": 0.0015833138336281652 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/downstream_results/dcnv2_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/downstream_results/dcnv2_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..8dd570895f72ba3bf1d35ae768ff0288eee438ed --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/downstream_results/dcnv2_credit_card_default_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "credit_card_default", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpa_yovf0e", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpa_yovf0e/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.449384868144989, + "lr": 0.00018128802184946835, + "credit_card_default_val_accuracy": 0.8173824548721313, + "credit_card_default_val_precision": 0.8002381324768066, + "credit_card_default_val_recall": 0.8173824548721313, + "credit_card_default_val_f1-score": 0.7930015325546265, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7482926845550537, + "credit_card_default_val_score": 0.8173824548721313 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4423121511936188, + "lr": 0.00018128802184946835, + "credit_card_default_test_accuracy": 0.8164875507354736, + "credit_card_default_test_precision": 0.7997376918792725, + "credit_card_default_test_recall": 0.8164875507354736, + "credit_card_default_test_f1-score": 0.7953307628631592, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7643378376960754, + "credit_card_default_test_score": 0.8164875507354736 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.4507068693637848, + "lr": 0.00011474337225081399, + "credit_card_default_val_accuracy": 0.8126479387283325, + "credit_card_default_val_precision": 0.7933984398841858, + "credit_card_default_val_recall": 0.8126479387283325, + "credit_card_default_val_f1-score": 0.7903753519058228, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7484633922576904, + "credit_card_default_val_score": 0.8126479387283325 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.443997859954834, + "lr": 0.00011474342318251729, + "credit_card_default_test_accuracy": 0.8154801726341248, + "credit_card_default_test_precision": 0.7984010577201843, + "credit_card_default_test_recall": 0.8154801726341248, + "credit_card_default_test_f1-score": 0.7960041165351868, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7629562616348267, + "credit_card_default_test_score": 0.8154801726341248 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4502916634082794, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8150151968002319, + "credit_card_default_val_precision": 0.7968494296073914, + "credit_card_default_val_recall": 0.8150151968002319, + "credit_card_default_val_f1-score": 0.7902107238769531, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.749933660030365, + "credit_card_default_val_score": 0.8150151968002319 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4467161297798157, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8161517977714539, + "credit_card_default_test_precision": 0.7993505597114563, + "credit_card_default_test_recall": 0.8161517977714539, + "credit_card_default_test_f1-score": 0.794178307056427, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7612230181694031, + "credit_card_default_test_score": 0.8161517977714539 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.45529431104660034, + "lr": 2.447174301778432e-05, + "credit_card_default_val_accuracy": 0.8136624693870544, + "credit_card_default_val_precision": 0.7947797775268555, + "credit_card_default_val_recall": 0.8136624693870544, + "credit_card_default_val_f1-score": 0.7901690006256104, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7480678558349609, + "credit_card_default_val_score": 0.8136624693870544 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4470069706439972, + "lr": 2.4471739379805513e-05, + "credit_card_default_test_accuracy": 0.8146407008171082, + "credit_card_default_test_precision": 0.7972304224967957, + "credit_card_default_test_recall": 0.8146407008171082, + "credit_card_default_test_f1-score": 0.7935130596160889, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.761884868144989, + "credit_card_default_test_score": 0.8146407008171082 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.4540325403213501, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.81400066614151, + "credit_card_default_val_precision": 0.7952589988708496, + "credit_card_default_val_recall": 0.81400066614151, + "credit_card_default_val_f1-score": 0.79044508934021, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7479915618896484, + "credit_card_default_val_score": 0.81400066614151 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44810545444488525, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.815648078918457, + "credit_card_default_test_precision": 0.7985900640487671, + "credit_card_default_test_recall": 0.815648078918457, + "credit_card_default_test_f1-score": 0.7946352958679199, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7593523263931274, + "credit_card_default_test_score": 0.815648078918457 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.4456277132034302, + "std": 0.0023914209397125036, + "n": 5, + "ci95": 0.0029688652560843415 + }, + "lr": { + "mean": 6.410063688235823e-05, + "std": 8.075550726481286e-05, + "n": 5, + "ci95": 0.00010025513107064813 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8156816601753235, + "std": 0.0007063657462252608, + "n": 5, + "ci95": 0.0008769283095381693 + }, + "credit_card_default_test_precision": { + "mean": 0.7986619591712951, + "std": 0.000968750134089611, + "n": 5, + "ci95": 0.0012026693281658128 + }, + "credit_card_default_test_recall": { + "mean": 0.8156816601753235, + "std": 0.0007063657462252608, + "n": 5, + "ci95": 0.0008769283095381693 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7947323083877563, + "std": 0.000971923375880583, + "n": 5, + "ci95": 0.001206608796598963 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7619508624076843, + "std": 0.0018704932858864618, + "n": 5, + "ci95": 0.002322151837005727 + }, + "credit_card_default_test_score": { + "mean": 0.8156816601753235, + "std": 0.0007063657462252608, + "n": 5, + "ci95": 0.0008769283095381693 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/downstream_results/fttransformer_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/downstream_results/fttransformer_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..f0ac338414b65b113d916838dfc1573b2fa6f205 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/downstream_results/fttransformer_credit_card_default_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "credit_card_default", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpt46z2wg3", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpt46z2wg3/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.4566786587238312, + "lr": 0.00040630935109220445, + "credit_card_default_val_accuracy": 0.8096043467521667, + "credit_card_default_val_precision": 0.7952069044113159, + "credit_card_default_val_recall": 0.8096043467521667, + "credit_card_default_val_f1-score": 0.7700688242912292, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7338567972183228, + "credit_card_default_val_score": 0.8096043467521667 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4494597017765045, + "lr": 0.0004063093801960349, + "credit_card_default_test_accuracy": 0.8104432225227356, + "credit_card_default_test_precision": 0.7968187928199768, + "credit_card_default_test_recall": 0.8104432225227356, + "credit_card_default_test_f1-score": 0.7743884325027466, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7524568438529968, + "credit_card_default_test_score": 0.8104432225227356 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.4495929777622223, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8163679242134094, + "credit_card_default_val_precision": 0.7991825938224792, + "credit_card_default_val_recall": 0.8163679242134094, + "credit_card_default_val_f1-score": 0.7902233600616455, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7436548471450806, + "credit_card_default_val_score": 0.8163679242134094 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44352373480796814, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8179986476898193, + "credit_card_default_test_precision": 0.8020936250686646, + "credit_card_default_test_recall": 0.8179986476898193, + "credit_card_default_test_f1-score": 0.7952129244804382, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7594689726829529, + "credit_card_default_test_score": 0.8179986476898193 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4501277804374695, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8153533935546875, + "credit_card_default_val_precision": 0.7973188161849976, + "credit_card_default_val_recall": 0.8153533935546875, + "credit_card_default_val_f1-score": 0.7907015681266785, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7428151965141296, + "credit_card_default_val_score": 0.8153533935546875 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4434935450553894, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8181665539741516, + "credit_card_default_test_precision": 0.8021530508995056, + "credit_card_default_test_recall": 0.8181665539741516, + "credit_card_default_test_f1-score": 0.7962392568588257, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7565823197364807, + "credit_card_default_test_score": 0.8181665539741516 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.4477646052837372, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8177206516265869, + "credit_card_default_val_precision": 0.800749659538269, + "credit_card_default_val_recall": 0.8177206516265869, + "credit_card_default_val_f1-score": 0.7932789325714111, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7455721497535706, + "credit_card_default_val_score": 0.8177206516265869 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4431977868080139, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8198455572128296, + "credit_card_default_test_precision": 0.804335355758667, + "credit_card_default_test_recall": 0.8198455572128296, + "credit_card_default_test_f1-score": 0.7987916469573975, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7580954432487488, + "credit_card_default_test_score": 0.8198455572128296 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.4472001791000366, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8143388628959656, + "credit_card_default_val_precision": 0.7959071397781372, + "credit_card_default_val_recall": 0.8143388628959656, + "credit_card_default_val_f1-score": 0.7892271280288696, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7508561015129089, + "credit_card_default_val_score": 0.8143388628959656 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4399498999118805, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8176628351211548, + "credit_card_default_test_precision": 0.8015589714050293, + "credit_card_default_test_recall": 0.8176628351211548, + "credit_card_default_test_f1-score": 0.7951334118843079, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7630558609962463, + "credit_card_default_test_score": 0.8176628351211548 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.4439249336719513, + "std": 0.003439126000656882, + "n": 5, + "ci95": 0.004269554357868086 + }, + "lr": { + "mean": 8.126187603920698e-05, + "std": 0.00018170707880282817, + "n": 5, + "ci95": 0.00022558296788483854 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8168233633041382, + "std": 0.003665038195812223, + "n": 5, + "ci95": 0.0045500164279221635 + }, + "credit_card_default_test_precision": { + "mean": 0.8013919591903687, + "std": 0.0027693375870359777, + "n": 5, + "ci95": 0.003438035524397467 + }, + "credit_card_default_test_recall": { + "mean": 0.8168233633041382, + "std": 0.003665038195812223, + "n": 5, + "ci95": 0.0045500164279221635 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7919531345367432, + "std": 0.009929693458446106, + "n": 5, + "ci95": 0.012327366304609125 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7579318881034851, + "std": 0.003887220144838578, + "n": 5, + "ci95": 0.004825847528185402 + }, + "credit_card_default_test_score": { + "mean": 0.8168233633041382, + "std": 0.003665038195812223, + "n": 5, + "ci95": 0.0045500164279221635 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/downstream_results/mlp_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/downstream_results/mlp_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..ef043528ad66b2eb7ed4509b5493755381b2e083 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/downstream_results/mlp_credit_card_default_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "credit_card_default", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpsgqtq3mh", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpsgqtq3mh/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.483579158782959, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8096043467521667, + "credit_card_default_val_precision": 0.7897605299949646, + "credit_card_default_val_recall": 0.8096043467521667, + "credit_card_default_val_f1-score": 0.7894893288612366, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7337478399276733, + "credit_card_default_val_score": 0.8096043467521667 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.461852490901947, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8190060257911682, + "credit_card_default_test_precision": 0.8031627535820007, + "credit_card_default_test_recall": 0.8190060257911682, + "credit_card_default_test_f1-score": 0.8018050789833069, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7516908049583435, + "credit_card_default_test_score": 0.8190060257911682 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.4896552860736847, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8099424839019775, + "credit_card_default_val_precision": 0.7968769669532776, + "credit_card_default_val_recall": 0.8099424839019775, + "credit_card_default_val_f1-score": 0.769448459148407, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7414311766624451, + "credit_card_default_val_score": 0.8099424839019775 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4766939878463745, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8089321851730347, + "credit_card_default_test_precision": 0.7947275042533875, + "credit_card_default_test_recall": 0.8089321851730347, + "credit_card_default_test_f1-score": 0.7718437314033508, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.76018226146698, + "credit_card_default_test_score": 0.8089321851730347 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4582194983959198, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8092661499977112, + "credit_card_default_val_precision": 0.7894629836082458, + "credit_card_default_val_recall": 0.8092661499977112, + "credit_card_default_val_f1-score": 0.789602518081665, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7361729145050049, + "credit_card_default_val_score": 0.8092661499977112 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.45436614751815796, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8169912695884705, + "credit_card_default_test_precision": 0.8009130954742432, + "credit_card_default_test_recall": 0.8169912695884705, + "credit_card_default_test_f1-score": 0.8009700775146484, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7478117942810059, + "credit_card_default_test_score": 0.8169912695884705 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.47695407271385193, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8143388628959656, + "credit_card_default_val_precision": 0.8017020225524902, + "credit_card_default_val_recall": 0.8143388628959656, + "credit_card_default_val_f1-score": 0.7782884836196899, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7377161383628845, + "credit_card_default_val_score": 0.8143388628959656 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.46317508816719055, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8109469413757324, + "credit_card_default_test_precision": 0.7959011197090149, + "credit_card_default_test_recall": 0.8109469413757324, + "credit_card_default_test_f1-score": 0.7772707939147949, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7483590245246887, + "credit_card_default_test_score": 0.8109469413757324 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.4634089469909668, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8035171031951904, + "credit_card_default_val_precision": 0.7802319526672363, + "credit_card_default_val_recall": 0.8035171031951904, + "credit_card_default_val_f1-score": 0.7716054916381836, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7276750802993774, + "credit_card_default_val_score": 0.8035171031951904 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.46026745438575745, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8119543194770813, + "credit_card_default_test_precision": 0.7941941022872925, + "credit_card_default_test_recall": 0.8119543194770813, + "credit_card_default_test_f1-score": 0.7852121591567993, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7459552884101868, + "credit_card_default_test_score": 0.8119543194770813 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.4632710337638855, + "std": 0.00822368004362028, + "n": 5, + "ci95": 0.010209410461043124 + }, + "lr": { + "mean": 0.0009999994654208422, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8135661482810974, + "std": 0.004250163946492216, + "n": 5, + "ci95": 0.005276429533530806 + }, + "credit_card_default_test_precision": { + "mean": 0.7977797150611877, + "std": 0.0040154983129823095, + "n": 5, + "ci95": 0.004985100376645387 + }, + "credit_card_default_test_recall": { + "mean": 0.8135661482810974, + "std": 0.004250163946492216, + "n": 5, + "ci95": 0.005276429533530806 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7874203681945801, + "std": 0.013610989916928806, + "n": 5, + "ci95": 0.01689756679564001 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.750799834728241, + "std": 0.005638690642573505, + "n": 5, + "ci95": 0.007000236746508088 + }, + "credit_card_default_test_score": { + "mean": 0.8135661482810974, + "std": 0.004250163946492216, + "n": 5, + "ci95": 0.005276429533530806 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/downstream_results/resnet_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/downstream_results/resnet_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..11b7aad39e5df56161015cae63117a1d50638ce1 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/downstream_results/resnet_credit_card_default_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "credit_card_default", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpwcv1lnya", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpwcv1lnya/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.4507269859313965, + "lr": 0.0005000000237487257, + "credit_card_default_val_accuracy": 0.8116334080696106, + "credit_card_default_val_precision": 0.7919314503669739, + "credit_card_default_val_recall": 0.8116334080696106, + "credit_card_default_val_f1-score": 0.7874582409858704, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.750039279460907, + "credit_card_default_val_score": 0.8116334080696106 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44497960805892944, + "lr": 0.0004999997327104211, + "credit_card_default_test_accuracy": 0.8193418383598328, + "credit_card_default_test_precision": 0.8035263419151306, + "credit_card_default_test_recall": 0.8193418383598328, + "credit_card_default_test_f1-score": 0.7992178797721863, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7605851292610168, + "credit_card_default_test_score": 0.8193418383598328 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.4531923830509186, + "lr": 0.0005000000237487257, + "credit_card_default_val_accuracy": 0.8096043467521667, + "credit_card_default_val_precision": 0.7897065877914429, + "credit_card_default_val_recall": 0.8096043467521667, + "credit_card_default_val_f1-score": 0.7892932295799255, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7523709535598755, + "credit_card_default_val_score": 0.8096043467521667 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4441319406032562, + "lr": 0.0004999997327104211, + "credit_card_default_test_accuracy": 0.8179986476898193, + "credit_card_default_test_precision": 0.8019250631332397, + "credit_card_default_test_recall": 0.8179986476898193, + "credit_card_default_test_f1-score": 0.800961434841156, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7659971714019775, + "credit_card_default_test_score": 0.8179986476898193 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.45320925116539, + "lr": 0.0005000000237487257, + "credit_card_default_val_accuracy": 0.8136624693870544, + "credit_card_default_val_precision": 0.7947996854782104, + "credit_card_default_val_recall": 0.8136624693870544, + "credit_card_default_val_f1-score": 0.7897477149963379, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7517858743667603, + "credit_card_default_val_score": 0.8136624693870544 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4465925693511963, + "lr": 0.0004999997327104211, + "credit_card_default_test_accuracy": 0.8196776509284973, + "credit_card_default_test_precision": 0.8040286898612976, + "credit_card_default_test_recall": 0.8196776509284973, + "credit_card_default_test_f1-score": 0.7991241216659546, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7647441625595093, + "credit_card_default_test_score": 0.8196776509284973 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.45513319969177246, + "lr": 0.0005000000237487257, + "credit_card_default_val_accuracy": 0.8116334080696106, + "credit_card_default_val_precision": 0.7925723791122437, + "credit_card_default_val_recall": 0.8116334080696106, + "credit_card_default_val_f1-score": 0.7923093438148499, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7499633431434631, + "credit_card_default_val_score": 0.8116334080696106 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4473074674606323, + "lr": 0.0004999997327104211, + "credit_card_default_test_accuracy": 0.8203492164611816, + "credit_card_default_test_precision": 0.8048912882804871, + "credit_card_default_test_recall": 0.8203492164611816, + "credit_card_default_test_f1-score": 0.803446888923645, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7633171677589417, + "credit_card_default_test_score": 0.8203492164611816 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.4562824070453644, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8092661499977112, + "credit_card_default_val_precision": 0.7898598313331604, + "credit_card_default_val_recall": 0.8092661499977112, + "credit_card_default_val_f1-score": 0.7907483577728271, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7533894777297974, + "credit_card_default_val_score": 0.8092661499977112 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4499596059322357, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8163197040557861, + "credit_card_default_test_precision": 0.8001025319099426, + "credit_card_default_test_recall": 0.8163197040557861, + "credit_card_default_test_f1-score": 0.800324022769928, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7647920250892639, + "credit_card_default_test_score": 0.8163197040557861 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.44659423828125, + "std": 0.0022640858355987468, + "n": 5, + "ci95": 0.0028107831885547906 + }, + "lr": { + "mean": 0.00039999978616833687, + "std": 0.00022360667821444535, + "n": 5, + "ci95": 0.0002775998516008258 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8187374114990235, + "std": 0.0015998733151188723, + "n": 5, + "ci95": 0.0019861866309341244 + }, + "credit_card_default_test_precision": { + "mean": 0.8028947830200195, + "std": 0.0018978716723153054, + "n": 5, + "ci95": 0.0023561411439012403 + }, + "credit_card_default_test_recall": { + "mean": 0.8187374114990235, + "std": 0.0015998733151188723, + "n": 5, + "ci95": 0.0019861866309341244 + }, + "credit_card_default_test_f1-score": { + "mean": 0.800614869594574, + "std": 0.0017606066267280523, + "n": 5, + "ci95": 0.0021857314021650905 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7638871312141419, + "std": 0.002075673413928748, + "n": 5, + "ci95": 0.002576875772582252 + }, + "credit_card_default_test_score": { + "mean": 0.8187374114990235, + "std": 0.0015998733151188723, + "n": 5, + "ci95": 0.0019861866309341244 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..5773e007309706762e16801dfef09c17aa123902 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f70944bf7d7a7a8dad9dc1476478874840ff2a42d376abc0c1ce7a771ceadd46 +size 176640128 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..15618fad95f67fa7e1ec8081313249061665e24f --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:56e5e4175ccdd98b84f174ed7b710e2ef37965eeee59d762fd6ab4888f3428d6 +size 15531742 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..265239207234b7b2e76b261d71847a8e15fcbaa4 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0003125_sx0.00025_sy2.5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_233123/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: credit_card_default +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.001 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.3 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.1 +mask_min_trgt_share: 0.1 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 256 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.1 +recon_y_weight: 1.0 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.00025 +sw_s_y_weight: 2.5e-05 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 15 +vtjepa_kl_x_end_weight: 0.0001 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 15 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 15 +wjepa_joint_end_weight: 0.0003125 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..46246f0478aef8d234664badb7836169d7551ab5 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-15T23:40:39.711373+00:00", + "spec": { + "dataset": "credit_card_default", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.0005, + "sw_z_weight": 0.0, + "sw_s_y_weight": 5e-05, + "lambda_joint": 0.000625, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..9f0372db1f1c48951d845a9463290073028a9892 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "credit_card_default", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.0005, + "sw_z_weight": 0.0, + "sw_s_y_weight": 5e-05, + "lambda_joint": 0.000625, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "credit_card_default", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.000625", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.0005", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "5e-05", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..8635ada77bbe392c68304c781cda1fb0e18a3eb1 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "credit_card_default", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.000625, + "s_x": 0.0005, + "z": 0.0, + "s_y": 5e-05 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 15 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "credit_card_default", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.000625", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.0005", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "5e-05", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/downstream_results/autoint_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/downstream_results/autoint_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..0c7b189083a7dc288a5ed6d51034b357669a1eaf --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/downstream_results/autoint_credit_card_default_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "credit_card_default", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpysuij006", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpysuij006/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.4485049843788147, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8150151968002319, + "credit_card_default_val_precision": 0.7966398596763611, + "credit_card_default_val_recall": 0.8150151968002319, + "credit_card_default_val_f1-score": 0.792923092842102, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7486228942871094, + "credit_card_default_val_score": 0.8150151968002319 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4454190731048584, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8163197040557861, + "credit_card_default_test_precision": 0.7994815111160278, + "credit_card_default_test_recall": 0.8163197040557861, + "credit_card_default_test_f1-score": 0.7963292002677917, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7619565725326538, + "credit_card_default_test_score": 0.8163197040557861 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.4500960409641266, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.816706120967865, + "credit_card_default_val_precision": 0.7992650270462036, + "credit_card_default_val_recall": 0.816706120967865, + "credit_card_default_val_f1-score": 0.7922348380088806, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7485522031784058, + "credit_card_default_val_score": 0.816706120967865 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4457777738571167, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8146407008171082, + "credit_card_default_test_precision": 0.7972332835197449, + "credit_card_default_test_recall": 0.8146407008171082, + "credit_card_default_test_f1-score": 0.7933194637298584, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7620750069618225, + "credit_card_default_test_score": 0.8146407008171082 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.44990822672843933, + "lr": 2.447174301778432e-05, + "credit_card_default_val_accuracy": 0.8153533935546875, + "credit_card_default_val_precision": 0.7975366115570068, + "credit_card_default_val_recall": 0.8153533935546875, + "credit_card_default_val_f1-score": 0.7896167635917664, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7489857077598572, + "credit_card_default_val_score": 0.8153533935546875 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4442165195941925, + "lr": 2.4471739379805513e-05, + "credit_card_default_test_accuracy": 0.8171591758728027, + "credit_card_default_test_precision": 0.8007490038871765, + "credit_card_default_test_recall": 0.8171591758728027, + "credit_card_default_test_f1-score": 0.7952083945274353, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.763892650604248, + "credit_card_default_test_score": 0.8171591758728027 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.4489027261734009, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8156915903091431, + "credit_card_default_val_precision": 0.7976931929588318, + "credit_card_default_val_recall": 0.8156915903091431, + "credit_card_default_val_f1-score": 0.791827380657196, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7504896521568298, + "credit_card_default_val_score": 0.8156915903091431 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4436771869659424, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8148086071014404, + "credit_card_default_test_precision": 0.7974595427513123, + "credit_card_default_test_recall": 0.8148086071014404, + "credit_card_default_test_f1-score": 0.7935551404953003, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7638221383094788, + "credit_card_default_test_score": 0.8148086071014404 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.44905272126197815, + "lr": 9.549151582177728e-05, + "credit_card_default_val_accuracy": 0.8126479387283325, + "credit_card_default_val_precision": 0.7934079170227051, + "credit_card_default_val_recall": 0.8126479387283325, + "credit_card_default_val_f1-score": 0.787851870059967, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7527189254760742, + "credit_card_default_val_score": 0.8126479387283325 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4447745084762573, + "lr": 9.549150126986206e-05, + "credit_card_default_test_accuracy": 0.8158159852027893, + "credit_card_default_test_precision": 0.7988327741622925, + "credit_card_default_test_recall": 0.8158159852027893, + "credit_card_default_test_f1-score": 0.7944850921630859, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7634971737861633, + "credit_card_default_test_score": 0.8158159852027893 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.4447730123996735, + "std": 0.0008569509733871353, + "n": 5, + "ci95": 0.001063874589708403 + }, + "lr": { + "mean": 2.3992648129933515e-05, + "std": 4.134990030704784e-05, + "n": 5, + "ci95": 5.133445154950598e-05 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8157488346099854, + "std": 0.0010525506964760816, + "n": 5, + "ci95": 0.00130670478841374 + }, + "credit_card_default_test_precision": { + "mean": 0.7987512230873108, + "std": 0.0014580533193454992, + "n": 5, + "ci95": 0.0018101220782334136 + }, + "credit_card_default_test_recall": { + "mean": 0.8157488346099854, + "std": 0.0010525506964760816, + "n": 5, + "ci95": 0.00130670478841374 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7945794582366943, + "std": 0.0012351999819502407, + "n": 5, + "ci95": 0.001533457472848314 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7630487084388733, + "std": 0.0009555638015102553, + "n": 5, + "ci95": 0.001186298958566754 + }, + "credit_card_default_test_score": { + "mean": 0.8157488346099854, + "std": 0.0010525506964760816, + "n": 5, + "ci95": 0.00130670478841374 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/downstream_results/dcnv2_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/downstream_results/dcnv2_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..18090b83075091233342e95e16c95ccd3f4cdfb8 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/downstream_results/dcnv2_credit_card_default_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "credit_card_default", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp2oj7uajl", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp2oj7uajl/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.4540533721446991, + "lr": 0.00013551565643865615, + "credit_card_default_val_accuracy": 0.8153533935546875, + "credit_card_default_val_precision": 0.7972539663314819, + "credit_card_default_val_recall": 0.8153533935546875, + "credit_card_default_val_f1-score": 0.7911281585693359, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7491999864578247, + "credit_card_default_val_score": 0.8153533935546875 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4485446512699127, + "lr": 0.00013551570009440184, + "credit_card_default_test_accuracy": 0.8163197040557861, + "credit_card_default_test_precision": 0.7995871901512146, + "credit_card_default_test_recall": 0.8163197040557861, + "credit_card_default_test_f1-score": 0.7943171858787537, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7580667734146118, + "credit_card_default_test_score": 0.8163197040557861 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.45348984003067017, + "lr": 8.856372005539015e-06, + "credit_card_default_val_accuracy": 0.8146770596504211, + "credit_card_default_val_precision": 0.7961969971656799, + "credit_card_default_val_recall": 0.8146770596504211, + "credit_card_default_val_f1-score": 0.7914155721664429, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.743869423866272, + "credit_card_default_val_score": 0.8146770596504211 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4515991508960724, + "lr": 8.856375643517822e-06, + "credit_card_default_test_accuracy": 0.8121222257614136, + "credit_card_default_test_precision": 0.7940826416015625, + "credit_card_default_test_recall": 0.8121222257614136, + "credit_card_default_test_f1-score": 0.7928528785705566, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7561061382293701, + "credit_card_default_test_score": 0.8121222257614136 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4532528817653656, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8143388628959656, + "credit_card_default_val_precision": 0.7957558035850525, + "credit_card_default_val_recall": 0.8143388628959656, + "credit_card_default_val_f1-score": 0.790510892868042, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7472051382064819, + "credit_card_default_val_score": 0.8143388628959656 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4497765898704529, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8166554570198059, + "credit_card_default_test_precision": 0.7999332547187805, + "credit_card_default_test_recall": 0.8166554570198059, + "credit_card_default_test_f1-score": 0.7961377501487732, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7588075995445251, + "credit_card_default_test_score": 0.8166554570198059 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.45046207308769226, + "lr": 2.447174301778432e-05, + "credit_card_default_val_accuracy": 0.8153533935546875, + "credit_card_default_val_precision": 0.7971200346946716, + "credit_card_default_val_recall": 0.8153533935546875, + "credit_card_default_val_f1-score": 0.7925893068313599, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7519813179969788, + "credit_card_default_val_score": 0.8153533935546875 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4491949677467346, + "lr": 2.4471739379805513e-05, + "credit_card_default_test_accuracy": 0.8158159852027893, + "credit_card_default_test_precision": 0.7988073825836182, + "credit_card_default_test_recall": 0.8158159852027893, + "credit_card_default_test_f1-score": 0.7952517867088318, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7586456537246704, + "credit_card_default_test_score": 0.8158159852027893 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.45108580589294434, + "lr": 2.447174301778432e-05, + "credit_card_default_val_accuracy": 0.8123098015785217, + "credit_card_default_val_precision": 0.7928943037986755, + "credit_card_default_val_recall": 0.8123098015785217, + "credit_card_default_val_f1-score": 0.7880076766014099, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7477099299430847, + "credit_card_default_val_score": 0.8123098015785217 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44689032435417175, + "lr": 2.4471739379805513e-05, + "credit_card_default_test_accuracy": 0.815144419670105, + "credit_card_default_test_precision": 0.797907292842865, + "credit_card_default_test_recall": 0.815144419670105, + "credit_card_default_test_f1-score": 0.7943142056465149, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7598724365234375, + "credit_card_default_test_score": 0.815144419670105 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.44920113682746887, + "std": 0.001721371786503602, + "n": 5, + "ci95": 0.0021370227235564658 + }, + "lr": { + "mean": 3.866311089950614e-05, + "std": 5.515091045491374e-05, + "n": 5, + "ci95": 6.846792179995561e-05 + }, + "credit_card_default_test_accuracy": { + "mean": 0.81521155834198, + "std": 0.0018184227739673506, + "n": 5, + "ci95": 0.00225750812199261 + }, + "credit_card_default_test_precision": { + "mean": 0.7980635523796081, + "std": 0.0023581035777798906, + "n": 5, + "ci95": 0.0029275029193148007 + }, + "credit_card_default_test_recall": { + "mean": 0.81521155834198, + "std": 0.0018184227739673506, + "n": 5, + "ci95": 0.00225750812199261 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7945747613906861, + "std": 0.0012247869228708469, + "n": 5, + "ci95": 0.0015205300250715613 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.758299720287323, + "std": 0.0013889847130897522, + "n": 5, + "ci95": 0.001724375825035719 + }, + "credit_card_default_test_score": { + "mean": 0.81521155834198, + "std": 0.0018184227739673506, + "n": 5, + "ci95": 0.00225750812199261 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/downstream_results/fttransformer_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/downstream_results/fttransformer_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..88df529cb6a9ed5556494e111780fcacf1e79e7b --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/downstream_results/fttransformer_credit_card_default_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "credit_card_default", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpeegil83a", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpeegil83a/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.4543362855911255, + "lr": 0.00040630935109220445, + "credit_card_default_val_accuracy": 0.811295211315155, + "credit_card_default_val_precision": 0.7920544147491455, + "credit_card_default_val_recall": 0.811295211315155, + "credit_card_default_val_f1-score": 0.7916486859321594, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7374506592750549, + "credit_card_default_val_score": 0.811295211315155 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44870761036872864, + "lr": 0.0004063093801960349, + "credit_card_default_test_accuracy": 0.8185023665428162, + "credit_card_default_test_precision": 0.8026673793792725, + "credit_card_default_test_recall": 0.8185023665428162, + "credit_card_default_test_f1-score": 0.80206698179245, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7490655779838562, + "credit_card_default_test_score": 0.8185023665428162 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.4462408125400543, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8156915903091431, + "credit_card_default_val_precision": 0.7976679801940918, + "credit_card_default_val_recall": 0.8156915903091431, + "credit_card_default_val_f1-score": 0.7920372486114502, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7511783838272095, + "credit_card_default_val_score": 0.8156915903091431 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.43976593017578125, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.819509744644165, + "credit_card_default_test_precision": 0.8037140369415283, + "credit_card_default_test_recall": 0.819509744644165, + "credit_card_default_test_f1-score": 0.7999123334884644, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7649862766265869, + "credit_card_default_test_score": 0.819509744644165 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4516635537147522, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8153533935546875, + "credit_card_default_val_precision": 0.797396183013916, + "credit_card_default_val_recall": 0.8153533935546875, + "credit_card_default_val_f1-score": 0.7902708053588867, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.748105525970459, + "credit_card_default_val_score": 0.8153533935546875 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4440714120864868, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8164875507354736, + "credit_card_default_test_precision": 0.7998243570327759, + "credit_card_default_test_recall": 0.8164875507354736, + "credit_card_default_test_f1-score": 0.7944561243057251, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7606873512268066, + "credit_card_default_test_score": 0.8164875507354736 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.44778093695640564, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8126479387283325, + "credit_card_default_val_precision": 0.7933658957481384, + "credit_card_default_val_recall": 0.8126479387283325, + "credit_card_default_val_f1-score": 0.789758026599884, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7529269456863403, + "credit_card_default_val_score": 0.8126479387283325 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4405474066734314, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8198455572128296, + "credit_card_default_test_precision": 0.8042052984237671, + "credit_card_default_test_recall": 0.8198455572128296, + "credit_card_default_test_f1-score": 0.7997312545776367, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7655860185623169, + "credit_card_default_test_score": 0.8198455572128296 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.45011061429977417, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.81400066614151, + "credit_card_default_val_precision": 0.7952854037284851, + "credit_card_default_val_recall": 0.81400066614151, + "credit_card_default_val_f1-score": 0.7900235652923584, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7450293302536011, + "credit_card_default_val_score": 0.81400066614151 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44211912155151367, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8185023665428162, + "credit_card_default_test_precision": 0.8024287819862366, + "credit_card_default_test_recall": 0.8185023665428162, + "credit_card_default_test_f1-score": 0.7979564666748047, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.760788083076477, + "credit_card_default_test_score": 0.8185023665428162 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.44304229617118834, + "std": 0.003569529508361765, + "n": 5, + "ci95": 0.00443144574088119 + }, + "lr": { + "mean": 8.126187603920698e-05, + "std": 0.00018170707880282817, + "n": 5, + "ci95": 0.00022558296788483854 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8185695171356201, + "std": 0.0013091894956456936, + "n": 5, + "ci95": 0.0016253128601108403 + }, + "credit_card_default_test_precision": { + "mean": 0.802567970752716, + "std": 0.001699336741461206, + "n": 5, + "ci95": 0.002109666987660598 + }, + "credit_card_default_test_recall": { + "mean": 0.8185695171356201, + "std": 0.0013091894956456936, + "n": 5, + "ci95": 0.0016253128601108403 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7988246321678162, + "std": 0.0028440930764348604, + "n": 5, + "ci95": 0.003530841843641542 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7602226614952088, + "std": 0.006642171684391056, + "n": 5, + "ci95": 0.00824602327899099 + }, + "credit_card_default_test_score": { + "mean": 0.8185695171356201, + "std": 0.0013091894956456936, + "n": 5, + "ci95": 0.0016253128601108403 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/downstream_results/mlp_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/downstream_results/mlp_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..865cc649457dbfed73afed30fddedee0d1a43ee5 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/downstream_results/mlp_credit_card_default_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "credit_card_default", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp_lqf22ad", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp_lqf22ad/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.45473188161849976, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8126479387283325, + "credit_card_default_val_precision": 0.7940841317176819, + "credit_card_default_val_recall": 0.8126479387283325, + "credit_card_default_val_f1-score": 0.7842506766319275, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7407041192054749, + "credit_card_default_val_score": 0.8126479387283325 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4469430148601532, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8161517977714539, + "credit_card_default_test_precision": 0.7998897433280945, + "credit_card_default_test_recall": 0.8161517977714539, + "credit_card_default_test_f1-score": 0.7915464043617249, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7556383013725281, + "credit_card_default_test_score": 0.8161517977714539 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.455850750207901, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8150151968002319, + "credit_card_default_val_precision": 0.7979454398155212, + "credit_card_default_val_recall": 0.8150151968002319, + "credit_card_default_val_f1-score": 0.7863955497741699, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7409055233001709, + "credit_card_default_val_score": 0.8150151968002319 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4468882381916046, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8163197040557861, + "credit_card_default_test_precision": 0.8009631633758545, + "credit_card_default_test_recall": 0.8163197040557861, + "credit_card_default_test_f1-score": 0.7895458936691284, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7616845965385437, + "credit_card_default_test_score": 0.8163197040557861 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4676104485988617, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8045316338539124, + "credit_card_default_val_precision": 0.7857528924942017, + "credit_card_default_val_recall": 0.8045316338539124, + "credit_card_default_val_f1-score": 0.7890622615814209, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7275779843330383, + "credit_card_default_val_score": 0.8045316338539124 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4662460684776306, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8139691352844238, + "credit_card_default_test_precision": 0.7988900542259216, + "credit_card_default_test_recall": 0.8139691352844238, + "credit_card_default_test_f1-score": 0.801418125629425, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7366665005683899, + "credit_card_default_test_score": 0.8139691352844238 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.4549866020679474, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.811295211315155, + "credit_card_default_val_precision": 0.791545569896698, + "credit_card_default_val_recall": 0.811295211315155, + "credit_card_default_val_f1-score": 0.7888618111610413, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7406513094902039, + "credit_card_default_val_score": 0.811295211315155 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4476670026779175, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8161517977714539, + "credit_card_default_test_precision": 0.7992894649505615, + "credit_card_default_test_recall": 0.8161517977714539, + "credit_card_default_test_f1-score": 0.796838641166687, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7605608105659485, + "credit_card_default_test_score": 0.8161517977714539 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.4754534959793091, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8129861354827881, + "credit_card_default_val_precision": 0.7965248227119446, + "credit_card_default_val_recall": 0.8129861354827881, + "credit_card_default_val_f1-score": 0.7803699970245361, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7478076815605164, + "credit_card_default_val_score": 0.8129861354827881 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4649517238140106, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8126259446144104, + "credit_card_default_test_precision": 0.7965449690818787, + "credit_card_default_test_recall": 0.8126259446144104, + "credit_card_default_test_f1-score": 0.7826716899871826, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7616662383079529, + "credit_card_default_test_score": 0.8126259446144104 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.4545392096042633, + "std": 0.010111104823848513, + "n": 5, + "ci95": 0.012552582154674726 + }, + "lr": { + "mean": 0.0009999994654208422, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8150436758995057, + "std": 0.0016646454669230109, + "n": 5, + "ci95": 0.0020665989865590803 + }, + "credit_card_default_test_precision": { + "mean": 0.7991154789924622, + "std": 0.0016356979765116827, + "n": 5, + "ci95": 0.0020306616920803596 + }, + "credit_card_default_test_recall": { + "mean": 0.8150436758995057, + "std": 0.0016646454669230109, + "n": 5, + "ci95": 0.0020665989865590803 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7924041509628296, + "std": 0.007150919578790655, + "n": 5, + "ci95": 0.008877615953750547 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7552432894706727, + "std": 0.010680283931504487, + "n": 5, + "ci95": 0.013259198062040687 + }, + "credit_card_default_test_score": { + "mean": 0.8150436758995057, + "std": 0.0016646454669230109, + "n": 5, + "ci95": 0.0020665989865590803 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/downstream_results/resnet_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/downstream_results/resnet_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..f0583c6fc2568d7b22a1cec3d8389956b55af895 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/downstream_results/resnet_credit_card_default_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "credit_card_default", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmplmowax2k", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmplmowax2k/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.4538520276546478, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8075752258300781, + "credit_card_default_val_precision": 0.7877365946769714, + "credit_card_default_val_recall": 0.8075752258300781, + "credit_card_default_val_f1-score": 0.7889881134033203, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7571464776992798, + "credit_card_default_val_score": 0.8075752258300781 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44576552510261536, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8161517977714539, + "credit_card_default_test_precision": 0.8000396490097046, + "credit_card_default_test_recall": 0.8161517977714539, + "credit_card_default_test_f1-score": 0.8006024360656738, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7614629864692688, + "credit_card_default_test_score": 0.8161517977714539 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.45287635922431946, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8068988919258118, + "credit_card_default_val_precision": 0.7869548797607422, + "credit_card_default_val_recall": 0.8068988919258118, + "credit_card_default_val_f1-score": 0.7884361743927002, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7506144642829895, + "credit_card_default_val_score": 0.8068988919258118 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44696736335754395, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8153122663497925, + "credit_card_default_test_precision": 0.7990486025810242, + "credit_card_default_test_recall": 0.8153122663497925, + "credit_card_default_test_f1-score": 0.7998172044754028, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7600182890892029, + "credit_card_default_test_score": 0.8153122663497925 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4632794260978699, + "lr": 0.0005000000237487257, + "credit_card_default_val_accuracy": 0.8163679242134094, + "credit_card_default_val_precision": 0.7986240386962891, + "credit_card_default_val_recall": 0.8163679242134094, + "credit_card_default_val_f1-score": 0.7928004264831543, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7470420002937317, + "credit_card_default_val_score": 0.8163679242134094 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4555114507675171, + "lr": 0.0004999997327104211, + "credit_card_default_test_accuracy": 0.8168233633041382, + "credit_card_default_test_precision": 0.8002089262008667, + "credit_card_default_test_recall": 0.8168233633041382, + "credit_card_default_test_f1-score": 0.795512855052948, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7624208927154541, + "credit_card_default_test_score": 0.8168233633041382 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.45316916704177856, + "lr": 0.0005000000237487257, + "credit_card_default_val_accuracy": 0.8109570741653442, + "credit_card_default_val_precision": 0.791806161403656, + "credit_card_default_val_recall": 0.8109570741653442, + "credit_card_default_val_f1-score": 0.7919448614120483, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7518815994262695, + "credit_card_default_val_score": 0.8109570741653442 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44620370864868164, + "lr": 0.0004999997327104211, + "credit_card_default_test_accuracy": 0.8201813101768494, + "credit_card_default_test_precision": 0.8047459721565247, + "credit_card_default_test_recall": 0.8201813101768494, + "credit_card_default_test_f1-score": 0.803729772567749, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7616191506385803, + "credit_card_default_test_score": 0.8201813101768494 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.4542137384414673, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8065606951713562, + "credit_card_default_val_precision": 0.7847191691398621, + "credit_card_default_val_recall": 0.8065606951713562, + "credit_card_default_val_f1-score": 0.777481734752655, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.751011312007904, + "credit_card_default_val_score": 0.8065606951713562 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.447592556476593, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8124580383300781, + "credit_card_default_test_precision": 0.7945823669433594, + "credit_card_default_test_recall": 0.8124580383300781, + "credit_card_default_test_f1-score": 0.7872516512870789, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7608055472373962, + "credit_card_default_test_score": 0.8124580383300781 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.44840812087059023, + "std": 0.004032399061159157, + "n": 5, + "ci95": 0.005006082062985436 + }, + "lr": { + "mean": 0.0005999996792525053, + "std": 0.00041832978963653156, + "n": 5, + "ci95": 0.0005193417676547898 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8161853551864624, + "std": 0.00278527430462983, + "n": 5, + "ci95": 0.003457820400566572 + }, + "credit_card_default_test_precision": { + "mean": 0.7997251033782959, + "std": 0.0036210336277782596, + "n": 5, + "ci95": 0.004495386299459401 + }, + "credit_card_default_test_recall": { + "mean": 0.8161853551864624, + "std": 0.00278527430462983, + "n": 5, + "ci95": 0.003457820400566572 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7973827838897705, + "std": 0.006377956386414832, + "n": 5, + "ci95": 0.007918009249649137 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7612653732299804, + "std": 0.0009036168333594097, + "n": 5, + "ci95": 0.0011218086188106314 + }, + "credit_card_default_test_score": { + "mean": 0.8161853551864624, + "std": 0.00278527430462983, + "n": 5, + "ci95": 0.003457820400566572 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..e010397910e07f0e0eab5fdd5add2ebf47c3e0ca --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fe98c6d4f88ae30617ef1f8aec90abb08a17bc3fc6104e09e403a283caefb7ce +size 176640128 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..668d51044f2bf603c4e75a50d4bf1848a0564ccf --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e07cbe02dbd4c2316257c3cd531cb94d9b9d3f0100d11e777c72c33a2db5c830 +size 15531358 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..280af1afab60617260a1730b682427cec2257ee3 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.000625_sx0.0005_sy5e-05/seed_42/wjepa_deterministic/credit_card_default/20260915_225412/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: credit_card_default +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.001 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.3 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.1 +mask_min_trgt_share: 0.1 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 256 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.1 +recon_y_weight: 1.0 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.0005 +sw_s_y_weight: 5.0e-05 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 15 +vtjepa_kl_x_end_weight: 0.0001 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 15 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 15 +wjepa_joint_end_weight: 0.000625 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..8eded80cb657917a5d09838d4caf181933f6c4ad --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-16T00:26:59.150998+00:00", + "spec": { + "dataset": "credit_card_default", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.0005, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0001, + "lambda_joint": 0.00125, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..1c5511fc84c579c49e58418ca3431248064cf518 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "credit_card_default", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.0005, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0001, + "lambda_joint": 0.00125, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "credit_card_default", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.00125", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.0005", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "0.0001", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..5952504ef77c079563537be589ac16a4397bf007 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "credit_card_default", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.00125, + "s_x": 0.0005, + "z": 0.0, + "s_y": 0.0001 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 15 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "credit_card_default", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.00125", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.0005", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "0.0001", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/downstream_results/autoint_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/downstream_results/autoint_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..908ce0ab2bd107dc520a7e3a25409fa97c775ea9 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/downstream_results/autoint_credit_card_default_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "credit_card_default", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmppj7zaawj", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmppj7zaawj/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.4466967284679413, + "lr": 3.942649982491275e-06, + "credit_card_default_val_accuracy": 0.8150151968002319, + "credit_card_default_val_precision": 0.7966541051864624, + "credit_card_default_val_recall": 0.8150151968002319, + "credit_card_default_val_f1-score": 0.7921065092086792, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7572253942489624, + "credit_card_default_val_score": 0.8150151968002319 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44273442029953003, + "lr": 3.942650891985977e-06, + "credit_card_default_test_accuracy": 0.8179986476898193, + "credit_card_default_test_precision": 0.8017714023590088, + "credit_card_default_test_recall": 0.8179986476898193, + "credit_card_default_test_f1-score": 0.7971588373184204, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7668476700782776, + "credit_card_default_test_score": 0.8179986476898193 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.4473523497581482, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8119716048240662, + "credit_card_default_val_precision": 0.7924553751945496, + "credit_card_default_val_recall": 0.8119716048240662, + "credit_card_default_val_f1-score": 0.7894130945205688, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7560443878173828, + "credit_card_default_val_score": 0.8119716048240662 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4431668519973755, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8163197040557861, + "credit_card_default_test_precision": 0.799483060836792, + "credit_card_default_test_recall": 0.8163197040557861, + "credit_card_default_test_f1-score": 0.7957643866539001, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7665897011756897, + "credit_card_default_test_score": 0.8163197040557861 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.44735637307167053, + "lr": 1.5708415958215483e-05, + "credit_card_default_val_accuracy": 0.81400066614151, + "credit_card_default_val_precision": 0.7953028082847595, + "credit_card_default_val_recall": 0.81400066614151, + "credit_card_default_val_f1-score": 0.7898112535476685, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.752339243888855, + "credit_card_default_val_score": 0.81400066614151 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4421602785587311, + "lr": 1.57084250531625e-05, + "credit_card_default_test_accuracy": 0.8191739320755005, + "credit_card_default_test_precision": 0.8033663034439087, + "credit_card_default_test_recall": 0.8191739320755005, + "credit_card_default_test_f1-score": 0.7984214425086975, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7643096446990967, + "credit_card_default_test_score": 0.8191739320755005 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.4472765624523163, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8129861354827881, + "credit_card_default_val_precision": 0.7938289046287537, + "credit_card_default_val_recall": 0.8129861354827881, + "credit_card_default_val_f1-score": 0.7898260951042175, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7563230395317078, + "credit_card_default_val_score": 0.8129861354827881 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44396623969078064, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8176628351211548, + "credit_card_default_test_precision": 0.8012838363647461, + "credit_card_default_test_recall": 0.8176628351211548, + "credit_card_default_test_f1-score": 0.7972578406333923, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7663176655769348, + "credit_card_default_test_score": 0.8176628351211548 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.44765597581863403, + "lr": 0.00011474337225081399, + "credit_card_default_val_accuracy": 0.8156915903091431, + "credit_card_default_val_precision": 0.7975848913192749, + "credit_card_default_val_recall": 0.8156915903091431, + "credit_card_default_val_f1-score": 0.793071448802948, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7490474581718445, + "credit_card_default_val_score": 0.8156915903091431 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4428730607032776, + "lr": 0.00011474342318251729, + "credit_card_default_test_accuracy": 0.815648078918457, + "credit_card_default_test_precision": 0.7985898852348328, + "credit_card_default_test_recall": 0.815648078918457, + "credit_card_default_test_f1-score": 0.795584499835968, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7636952996253967, + "credit_card_default_test_score": 0.815648078918457 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.44298017024993896, + "std": 0.0006616779608335716, + "n": 5, + "ci95": 0.0008214499906786341 + }, + "lr": { + "mean": 2.6878899825533152e-05, + "std": 4.953780374167267e-05, + "n": 5, + "ci95": 6.149944660476953e-05 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8173606395721436, + "std": 0.001396676139318127, + "n": 5, + "ci95": 0.0017339244610453642 + }, + "credit_card_default_test_precision": { + "mean": 0.8008988976478577, + "std": 0.0018929914056988284, + "n": 5, + "ci95": 0.0023500824639936246 + }, + "credit_card_default_test_recall": { + "mean": 0.8173606395721436, + "std": 0.001396676139318127, + "n": 5, + "ci95": 0.0017339244610453642 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7968374013900756, + "std": 0.0011737119969536247, + "n": 5, + "ci95": 0.0014571222951756477 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7655519962310791, + "std": 0.001443313153142626, + "n": 5, + "ci95": 0.0017918226786664433 + }, + "credit_card_default_test_score": { + "mean": 0.8173606395721436, + "std": 0.001396676139318127, + "n": 5, + "ci95": 0.0017339244610453642 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/downstream_results/dcnv2_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/downstream_results/dcnv2_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..5aef892f13956c8f78adaf27db3165c2fe22509c --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/downstream_results/dcnv2_credit_card_default_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "credit_card_default", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpo9oyoloe", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpo9oyoloe/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.4506665766239166, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8136624693870544, + "credit_card_default_val_precision": 0.7947738766670227, + "credit_card_default_val_recall": 0.8136624693870544, + "credit_card_default_val_f1-score": 0.7903781533241272, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7519800066947937, + "credit_card_default_val_score": 0.8136624693870544 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4451627731323242, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8158159852027893, + "credit_card_default_test_precision": 0.7988141775131226, + "credit_card_default_test_recall": 0.8158159852027893, + "credit_card_default_test_f1-score": 0.795817494392395, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7618768215179443, + "credit_card_default_test_score": 0.8158159852027893 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.4499523639678955, + "lr": 8.856372005539015e-06, + "credit_card_default_val_accuracy": 0.8153533935546875, + "credit_card_default_val_precision": 0.7971306443214417, + "credit_card_default_val_recall": 0.8153533935546875, + "credit_card_default_val_f1-score": 0.7923835515975952, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7534835934638977, + "credit_card_default_val_score": 0.8153533935546875 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44925761222839355, + "lr": 8.856375643517822e-06, + "credit_card_default_test_accuracy": 0.811786413192749, + "credit_card_default_test_precision": 0.7934576272964478, + "credit_card_default_test_recall": 0.811786413192749, + "credit_card_default_test_f1-score": 0.7912546396255493, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7594878673553467, + "credit_card_default_test_score": 0.811786413192749 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4507373869419098, + "lr": 3.942649982491275e-06, + "credit_card_default_val_accuracy": 0.8136624693870544, + "credit_card_default_val_precision": 0.7947738766670227, + "credit_card_default_val_recall": 0.8136624693870544, + "credit_card_default_val_f1-score": 0.7903781533241272, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7504051327705383, + "credit_card_default_val_score": 0.8136624693870544 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.45056381821632385, + "lr": 3.942650891985977e-06, + "credit_card_default_test_accuracy": 0.8146407008171082, + "credit_card_default_test_precision": 0.7972332835197449, + "credit_card_default_test_recall": 0.8146407008171082, + "credit_card_default_test_f1-score": 0.7933194637298584, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7578086256980896, + "credit_card_default_test_score": 0.8146407008171082 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.45088261365890503, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8150151968002319, + "credit_card_default_val_precision": 0.7967853546142578, + "credit_card_default_val_recall": 0.8150151968002319, + "credit_card_default_val_f1-score": 0.7906391620635986, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7538487911224365, + "credit_card_default_val_score": 0.8150151968002319 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44990894198417664, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8158159852027893, + "credit_card_default_test_precision": 0.7988226413726807, + "credit_card_default_test_recall": 0.8158159852027893, + "credit_card_default_test_f1-score": 0.7946781516075134, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7595716118812561, + "credit_card_default_test_score": 0.8158159852027893 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.45530930161476135, + "lr": 2.447174301778432e-05, + "credit_card_default_val_accuracy": 0.811295211315155, + "credit_card_default_val_precision": 0.7914618253707886, + "credit_card_default_val_recall": 0.811295211315155, + "credit_card_default_val_f1-score": 0.7876092791557312, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7453367114067078, + "credit_card_default_val_score": 0.811295211315155 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4486435651779175, + "lr": 2.4471739379805513e-05, + "credit_card_default_test_accuracy": 0.8126259446144104, + "credit_card_default_test_precision": 0.7945024967193604, + "credit_card_default_test_recall": 0.8126259446144104, + "credit_card_default_test_f1-score": 0.7909747958183289, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7581478953361511, + "credit_card_default_test_score": 0.8126259446144104 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.44870734214782715, + "std": 0.0021072007337186054, + "n": 5, + "ci95": 0.0026160158347884564 + }, + "lr": { + "mean": 7.454153183061862e-06, + "std": 1.0186067880756795e-05, + "n": 5, + "ci95": 1.2645646161704635e-05 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8141370058059693, + "std": 0.0018506920227376739, + "n": 5, + "ci95": 0.0022975692630168554 + }, + "credit_card_default_test_precision": { + "mean": 0.7965660452842712, + "std": 0.002475481082248109, + "n": 5, + "ci95": 0.0030732229759868274 + }, + "credit_card_default_test_recall": { + "mean": 0.8141370058059693, + "std": 0.0018506920227376739, + "n": 5, + "ci95": 0.0022975692630168554 + }, + "credit_card_default_test_f1-score": { + "mean": 0.793208909034729, + "std": 0.002108675202222726, + "n": 5, + "ci95": 0.0026178463357430896 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7593785643577575, + "std": 0.0016023230764968374, + "n": 5, + "ci95": 0.0019892279237989485 + }, + "credit_card_default_test_score": { + "mean": 0.8141370058059693, + "std": 0.0018506920227376739, + "n": 5, + "ci95": 0.0022975692630168554 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/downstream_results/fttransformer_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/downstream_results/fttransformer_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..c0a348a417a30939bdf31a8d718b0aeb9eb4443a --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/downstream_results/fttransformer_credit_card_default_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "credit_card_default", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpmtkqvsce", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpmtkqvsce/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.44665321707725525, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8146770596504211, + "credit_card_default_val_precision": 0.7962096333503723, + "credit_card_default_val_recall": 0.8146770596504211, + "credit_card_default_val_f1-score": 0.7912070751190186, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7545708417892456, + "credit_card_default_val_score": 0.8146770596504211 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4404583275318146, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8188381195068359, + "credit_card_default_test_precision": 0.8028231263160706, + "credit_card_default_test_recall": 0.8188381195068359, + "credit_card_default_test_f1-score": 0.799167811870575, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7659869194030762, + "credit_card_default_test_score": 0.8188381195068359 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.44520750641822815, + "lr": 8.856372005539015e-06, + "credit_card_default_val_accuracy": 0.81400066614151, + "credit_card_default_val_precision": 0.7953231334686279, + "credit_card_default_val_recall": 0.81400066614151, + "credit_card_default_val_f1-score": 0.7895979285240173, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7541888356208801, + "credit_card_default_val_score": 0.81400066614151 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4413626790046692, + "lr": 8.856375643517822e-06, + "credit_card_default_test_accuracy": 0.8174949884414673, + "credit_card_default_test_precision": 0.8010461330413818, + "credit_card_default_test_recall": 0.8174949884414673, + "credit_card_default_test_f1-score": 0.7973995208740234, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7644074559211731, + "credit_card_default_test_score": 0.8174949884414673 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4468969702720642, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8126479387283325, + "credit_card_default_val_precision": 0.7933566570281982, + "credit_card_default_val_recall": 0.8126479387283325, + "credit_card_default_val_f1-score": 0.7893415689468384, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7531666159629822, + "credit_card_default_val_score": 0.8126479387283325 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44223999977111816, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.817327082157135, + "credit_card_default_test_precision": 0.800818920135498, + "credit_card_default_test_recall": 0.817327082157135, + "credit_card_default_test_f1-score": 0.7976316809654236, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7633288502693176, + "credit_card_default_test_score": 0.817327082157135 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.44827666878700256, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8109570741653442, + "credit_card_default_val_precision": 0.7910169363021851, + "credit_card_default_val_recall": 0.8109570741653442, + "credit_card_default_val_f1-score": 0.7877559065818787, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7547550797462463, + "credit_card_default_val_score": 0.8109570741653442 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.43966251611709595, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8171591758728027, + "credit_card_default_test_precision": 0.8006656169891357, + "credit_card_default_test_recall": 0.8171591758728027, + "credit_card_default_test_f1-score": 0.7958877086639404, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.768369197845459, + "credit_card_default_test_score": 0.8171591758728027 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.4470387101173401, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8170443177223206, + "credit_card_default_val_precision": 0.7994793653488159, + "credit_card_default_val_recall": 0.8170443177223206, + "credit_card_default_val_f1-score": 0.7943868637084961, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7504265904426575, + "credit_card_default_val_score": 0.8170443177223206 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44190946221351624, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8190060257911682, + "credit_card_default_test_precision": 0.8030440211296082, + "credit_card_default_test_recall": 0.8190060257911682, + "credit_card_default_test_f1-score": 0.7993999123573303, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7621228098869324, + "credit_card_default_test_score": 0.8190060257911682 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.44112659692764283, + "std": 0.001060471039207338, + "n": 5, + "ci95": 0.0013165376162361535 + }, + "lr": { + "mean": 1.7712751287035645e-06, + "std": 3.960691594635859e-06, + "n": 5, + "ci95": 4.917059757281094e-06 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8179650783538819, + "std": 0.0008836394773564739, + "n": 5, + "ci95": 0.0010970074317169555 + }, + "credit_card_default_test_precision": { + "mean": 0.8016795635223388, + "std": 0.001155368903311339, + "n": 5, + "ci95": 0.0014343499875072911 + }, + "credit_card_default_test_recall": { + "mean": 0.8179650783538819, + "std": 0.0008836394773564739, + "n": 5, + "ci95": 0.0010970074317169555 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7978973269462586, + "std": 0.0014342976655663333, + "n": 5, + "ci95": 0.0017806302669134823 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7648430466651917, + "std": 0.002430247165306601, + "n": 5, + "ci95": 0.003017066653954959 + }, + "credit_card_default_test_score": { + "mean": 0.8179650783538819, + "std": 0.0008836394773564739, + "n": 5, + "ci95": 0.0010970074317169555 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/downstream_results/mlp_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/downstream_results/mlp_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..078ec80093c5fc3d86a8d573c46e1d1505f8e47a --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/downstream_results/mlp_credit_card_default_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "credit_card_default", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpxgg6e7bl", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpxgg6e7bl/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.4594225287437439, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8123098015785217, + "credit_card_default_val_precision": 0.7939954996109009, + "credit_card_default_val_recall": 0.8123098015785217, + "credit_card_default_val_f1-score": 0.7825542688369751, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7457811236381531, + "credit_card_default_val_score": 0.8123098015785217 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4544026255607605, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8131296038627625, + "credit_card_default_test_precision": 0.7960661053657532, + "credit_card_default_test_recall": 0.8131296038627625, + "credit_card_default_test_f1-score": 0.7861685752868652, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.763243556022644, + "credit_card_default_test_score": 0.8131296038627625 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.45449942350387573, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8062224984169006, + "credit_card_default_val_precision": 0.785197377204895, + "credit_card_default_val_recall": 0.8062224984169006, + "credit_card_default_val_f1-score": 0.7724267840385437, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7439695000648499, + "credit_card_default_val_score": 0.8062224984169006 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.448397159576416, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8107790350914001, + "credit_card_default_test_precision": 0.7933811545372009, + "credit_card_default_test_recall": 0.8107790350914001, + "credit_card_default_test_f1-score": 0.7810633182525635, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.758385181427002, + "credit_card_default_test_score": 0.8107790350914001 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4734494090080261, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.7937098145484924, + "credit_card_default_val_precision": 0.7795546054840088, + "credit_card_default_val_recall": 0.7937098145484924, + "credit_card_default_val_f1-score": 0.7845005393028259, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7333582043647766, + "credit_card_default_val_score": 0.7937098145484924 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4754425883293152, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8032236695289612, + "credit_card_default_test_precision": 0.7936435341835022, + "credit_card_default_test_recall": 0.8032236695289612, + "credit_card_default_test_f1-score": 0.797307014465332, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7409855723381042, + "credit_card_default_test_score": 0.8032236695289612 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.4619128108024597, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8173824548721313, + "credit_card_default_val_precision": 0.8015634417533875, + "credit_card_default_val_recall": 0.8173824548721313, + "credit_card_default_val_f1-score": 0.7892444133758545, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7317764163017273, + "credit_card_default_val_score": 0.8173824548721313 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4523678719997406, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8176628351211548, + "credit_card_default_test_precision": 0.8023170828819275, + "credit_card_default_test_recall": 0.8176628351211548, + "credit_card_default_test_f1-score": 0.7925816178321838, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7484805583953857, + "credit_card_default_test_score": 0.8176628351211548 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.5030925273895264, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8133243322372437, + "credit_card_default_val_precision": 0.7961936593055725, + "credit_card_default_val_recall": 0.8133243322372437, + "credit_card_default_val_f1-score": 0.7823938131332397, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7401190996170044, + "credit_card_default_val_score": 0.8133243322372437 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.49112600088119507, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8136333227157593, + "credit_card_default_test_precision": 0.7978288531303406, + "credit_card_default_test_recall": 0.8136333227157593, + "credit_card_default_test_f1-score": 0.7844238877296448, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7524433732032776, + "credit_card_default_test_score": 0.8136333227157593 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.46434724926948545, + "std": 0.018282604221159234, + "n": 5, + "ci95": 0.022697212172720183 + }, + "lr": { + "mean": 0.0009999994654208422, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8116856932640075, + "std": 0.005339293909624962, + "n": 5, + "ci95": 0.0066285461990702324 + }, + "credit_card_default_test_precision": { + "mean": 0.7966473460197449, + "std": 0.0036595972519126853, + "n": 5, + "ci95": 0.004543261686824352 + }, + "credit_card_default_test_recall": { + "mean": 0.8116856932640075, + "std": 0.005339293909624962, + "n": 5, + "ci95": 0.0066285461990702324 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7883088827133179, + "std": 0.006545894642454088, + "n": 5, + "ci95": 0.008126498706792672 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7527076482772828, + "std": 0.008639805107656778, + "n": 5, + "ci95": 0.01072601513916071 + }, + "credit_card_default_test_score": { + "mean": 0.8116856932640075, + "std": 0.005339293909624962, + "n": 5, + "ci95": 0.0066285461990702324 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/downstream_results/resnet_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/downstream_results/resnet_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..66e6dd431878de22f0bd865f2a101cdf094c8922 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/downstream_results/resnet_credit_card_default_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "credit_card_default", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp752hu84k", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp752hu84k/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.4531463086605072, + "lr": 0.0005000000237487257, + "credit_card_default_val_accuracy": 0.8146770596504211, + "credit_card_default_val_precision": 0.7971151471138, + "credit_card_default_val_recall": 0.8146770596504211, + "credit_card_default_val_f1-score": 0.7868182063102722, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7511764168739319, + "credit_card_default_val_score": 0.8146770596504211 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4479611814022064, + "lr": 0.0004999997327104211, + "credit_card_default_test_accuracy": 0.815144419670105, + "credit_card_default_test_precision": 0.7986536026000977, + "credit_card_default_test_recall": 0.815144419670105, + "credit_card_default_test_f1-score": 0.7896633744239807, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7605858445167542, + "credit_card_default_test_score": 0.815144419670105 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.4540707767009735, + "lr": 0.0005000000237487257, + "credit_card_default_val_accuracy": 0.8048697710037231, + "credit_card_default_val_precision": 0.7856640219688416, + "credit_card_default_val_recall": 0.8048697710037231, + "credit_card_default_val_f1-score": 0.7886255383491516, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7527823448181152, + "credit_card_default_val_score": 0.8048697710037231 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4465164244174957, + "lr": 0.0004999997327104211, + "credit_card_default_test_accuracy": 0.8143048882484436, + "credit_card_default_test_precision": 0.798718273639679, + "credit_card_default_test_recall": 0.8143048882484436, + "credit_card_default_test_f1-score": 0.8007622361183167, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7661420702934265, + "credit_card_default_test_score": 0.8143048882484436 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.45368295907974243, + "lr": 0.0005000000237487257, + "credit_card_default_val_accuracy": 0.8126479387283325, + "credit_card_default_val_precision": 0.7933623194694519, + "credit_card_default_val_recall": 0.8126479387283325, + "credit_card_default_val_f1-score": 0.788709282875061, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7508742809295654, + "credit_card_default_val_score": 0.8126479387283325 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44604626297950745, + "lr": 0.0004999997327104211, + "credit_card_default_test_accuracy": 0.8188381195068359, + "credit_card_default_test_precision": 0.8029491305351257, + "credit_card_default_test_recall": 0.8188381195068359, + "credit_card_default_test_f1-score": 0.7977620363235474, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7657611966133118, + "credit_card_default_test_score": 0.8188381195068359 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.45522627234458923, + "lr": 0.0005000000237487257, + "credit_card_default_val_accuracy": 0.8096043467521667, + "credit_card_default_val_precision": 0.7904725074768066, + "credit_card_default_val_recall": 0.8096043467521667, + "credit_card_default_val_f1-score": 0.7915864586830139, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7497424483299255, + "credit_card_default_val_score": 0.8096043467521667 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44650569558143616, + "lr": 0.0004999997327104211, + "credit_card_default_test_accuracy": 0.8190060257911682, + "credit_card_default_test_precision": 0.8034393787384033, + "credit_card_default_test_recall": 0.8190060257911682, + "credit_card_default_test_f1-score": 0.8032443523406982, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7639819979667664, + "credit_card_default_test_score": 0.8190060257911682 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.45477432012557983, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8065606951713562, + "credit_card_default_val_precision": 0.7848050594329834, + "credit_card_default_val_recall": 0.8065606951713562, + "credit_card_default_val_f1-score": 0.776755154132843, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7520853281021118, + "credit_card_default_val_score": 0.8065606951713562 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4478449523448944, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8111148476600647, + "credit_card_default_test_precision": 0.7927166223526001, + "credit_card_default_test_recall": 0.8111148476600647, + "credit_card_default_test_f1-score": 0.7852959632873535, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7621224522590637, + "credit_card_default_test_score": 0.8111148476600647 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.44697490334510803, + "std": 0.0008692613889274515, + "n": 5, + "ci95": 0.0010791575390121752 + }, + "lr": { + "mean": 0.00039999978616833687, + "std": 0.00022360667821444535, + "n": 5, + "ci95": 0.0002775998516008258 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8156816601753235, + "std": 0.0033186799750774027, + "n": 5, + "ci95": 0.004120024839815379 + }, + "credit_card_default_test_precision": { + "mean": 0.7992954015731811, + "std": 0.004317047621924733, + "n": 5, + "ci95": 0.005359463271712716 + }, + "credit_card_default_test_recall": { + "mean": 0.8156816601753235, + "std": 0.0033186799750774027, + "n": 5, + "ci95": 0.004120024839815379 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7953455924987793, + "std": 0.007596911962046749, + "n": 5, + "ci95": 0.00943129986156414 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7637187123298645, + "std": 0.002370476330948227, + "n": 5, + "ci95": 0.0029428632585982717 + }, + "credit_card_default_test_score": { + "mean": 0.8156816601753235, + "std": 0.0033186799750774027, + "n": 5, + "ci95": 0.004120024839815379 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..a226f8ad7e4965886b16097e4b0efaf3aceede38 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ae87d2f54c91f400be0ca1ef241d6593b3e1edb9ecebd0640bdd3c5c0b975db6 +size 176640128 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..138c443c12d041c029933b1bcb271d7cc94050bb --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0a1211b0dd7a714304192bae23066de5fa3c1e8a7edd07af7681cf755ca6359f +size 15531358 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0bfc06579cd2adb176c3ca49b01722514db86a61 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.0005_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260915_234054/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: credit_card_default +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.001 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.3 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.1 +mask_min_trgt_share: 0.1 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 256 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.1 +recon_y_weight: 1.0 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.0005 +sw_s_y_weight: 0.0001 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 15 +vtjepa_kl_x_end_weight: 0.0001 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 15 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 15 +wjepa_joint_end_weight: 0.00125 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..5f55fe3ee455a884f3530d81337f693d4a326f7c --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-16T02:42:24.488824+00:00", + "spec": { + "dataset": "credit_card_default", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.001, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0001, + "lambda_joint": 0.00125, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..a951fbc77c9de10ed675f161e2c39bd588e9f1af --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "credit_card_default", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.001, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0001, + "lambda_joint": 0.00125, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "credit_card_default", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.00125", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.001", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "0.0001", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..0ed7622ccce64dc58c0f964ada2a46f9d321a6f8 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "credit_card_default", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.00125, + "s_x": 0.001, + "z": 0.0, + "s_y": 0.0001 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 15 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "credit_card_default", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.00125", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.001", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "0.0001", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/downstream_results/autoint_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/downstream_results/autoint_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..54def9e1c2ff777f91338a194d5810fab0485ff8 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/downstream_results/autoint_credit_card_default_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "credit_card_default", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmptks3ag5k", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmptks3ag5k/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.44546744227409363, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8150151968002319, + "credit_card_default_val_precision": 0.7966643571853638, + "credit_card_default_val_recall": 0.8150151968002319, + "credit_card_default_val_f1-score": 0.7918998599052429, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7653359770774841, + "credit_card_default_val_score": 0.8150151968002319 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4415125846862793, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8183344602584839, + "credit_card_default_test_precision": 0.8022464513778687, + "credit_card_default_test_recall": 0.8183344602584839, + "credit_card_default_test_f1-score": 0.7973429560661316, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7686120867729187, + "credit_card_default_test_score": 0.8183344602584839 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.4460710883140564, + "lr": 6.184668018249795e-05, + "credit_card_default_val_accuracy": 0.8153533935546875, + "credit_card_default_val_precision": 0.7971121072769165, + "credit_card_default_val_recall": 0.8153533935546875, + "credit_card_default_val_f1-score": 0.7927940487861633, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7558637857437134, + "credit_card_default_val_score": 0.8153533935546875 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4429546296596527, + "lr": 6.184665107866749e-05, + "credit_card_default_test_accuracy": 0.8171591758728027, + "credit_card_default_test_precision": 0.8005972504615784, + "credit_card_default_test_recall": 0.8171591758728027, + "credit_card_default_test_f1-score": 0.7974919676780701, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7640913128852844, + "credit_card_default_test_score": 0.8171591758728027 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.44784149527549744, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8129861354827881, + "credit_card_default_val_precision": 0.7938451170921326, + "credit_card_default_val_recall": 0.8129861354827881, + "credit_card_default_val_f1-score": 0.7904467582702637, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7518703937530518, + "credit_card_default_val_score": 0.8129861354827881 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44063782691955566, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8171591758728027, + "credit_card_default_test_precision": 0.8006010055541992, + "credit_card_default_test_recall": 0.8171591758728027, + "credit_card_default_test_f1-score": 0.7976765632629395, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.767403781414032, + "credit_card_default_test_score": 0.8171591758728027 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.4451896846294403, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8123098015785217, + "credit_card_default_val_precision": 0.7928875684738159, + "credit_card_default_val_recall": 0.8123098015785217, + "credit_card_default_val_f1-score": 0.7882214188575745, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7605947256088257, + "credit_card_default_val_score": 0.8123098015785217 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44132500886917114, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8149765133857727, + "credit_card_default_test_precision": 0.7976836562156677, + "credit_card_default_test_recall": 0.8149765133857727, + "credit_card_default_test_f1-score": 0.7938871383666992, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7687815427780151, + "credit_card_default_test_score": 0.8149765133857727 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.4478146731853485, + "lr": 9.549151582177728e-05, + "credit_card_default_val_accuracy": 0.8150151968002319, + "credit_card_default_val_precision": 0.7967334389686584, + "credit_card_default_val_recall": 0.8150151968002319, + "credit_card_default_val_f1-score": 0.7910634279251099, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7499752640724182, + "credit_card_default_val_score": 0.8150151968002319 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4436773359775543, + "lr": 9.549150126986206e-05, + "credit_card_default_test_accuracy": 0.815144419670105, + "credit_card_default_test_precision": 0.797907292842865, + "credit_card_default_test_recall": 0.815144419670105, + "credit_card_default_test_f1-score": 0.7943142056465149, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7638025879859924, + "credit_card_default_test_score": 0.815144419670105 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.4420214772224426, + "std": 0.001252115173326898, + "n": 5, + "ci95": 0.001554457089914564 + }, + "lr": { + "mean": 3.146763046970591e-05, + "std": 4.4700604659021924e-05, + "n": 5, + "ci95": 5.5494233530499417e-05 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8165547490119934, + "std": 0.0014472294501463845, + "n": 5, + "ci95": 0.0017966846240955753 + }, + "credit_card_default_test_precision": { + "mean": 0.7998071312904358, + "std": 0.001957251878288529, + "n": 5, + "ci95": 0.002429859587812762 + }, + "credit_card_default_test_recall": { + "mean": 0.8165547490119934, + "std": 0.0014472294501463845, + "n": 5, + "ci95": 0.0017966846240955753 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7961425662040711, + "std": 0.0018738212423817407, + "n": 5, + "ci95": 0.002326283378320148 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7665382623672485, + "std": 0.002426605343621657, + "n": 5, + "ci95": 0.0030125454600113347 + }, + "credit_card_default_test_score": { + "mean": 0.8165547490119934, + "std": 0.0014472294501463845, + "n": 5, + "ci95": 0.0017966846240955753 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/downstream_results/dcnv2_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/downstream_results/dcnv2_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..a81cfdc29faeabda7c031ae45e4a37f96ff0d096 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/downstream_results/dcnv2_credit_card_default_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "credit_card_default", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmppixb22y_", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmppixb22y_/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.45164215564727783, + "lr": 3.511174872983247e-05, + "credit_card_default_val_accuracy": 0.8092661499977112, + "credit_card_default_val_precision": 0.7886391878128052, + "credit_card_default_val_recall": 0.8092661499977112, + "credit_card_default_val_f1-score": 0.7855390310287476, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7485185265541077, + "credit_card_default_val_score": 0.8092661499977112 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4460753798484802, + "lr": 3.51117632817477e-05, + "credit_card_default_test_accuracy": 0.8136333227157593, + "credit_card_default_test_precision": 0.7959308624267578, + "credit_card_default_test_recall": 0.8136333227157593, + "credit_card_default_test_f1-score": 0.7935397028923035, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7620080709457397, + "credit_card_default_test_score": 0.8136333227157593 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.45115557312965393, + "lr": 8.856372005539015e-06, + "credit_card_default_val_accuracy": 0.8116334080696106, + "credit_card_default_val_precision": 0.7919310331344604, + "credit_card_default_val_recall": 0.8116334080696106, + "credit_card_default_val_f1-score": 0.787671685218811, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7514497637748718, + "credit_card_default_val_score": 0.8116334080696106 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4473575949668884, + "lr": 8.856375643517822e-06, + "credit_card_default_test_accuracy": 0.8158159852027893, + "credit_card_default_test_precision": 0.7988327741622925, + "credit_card_default_test_recall": 0.8158159852027893, + "credit_card_default_test_f1-score": 0.7944850921630859, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7607871294021606, + "credit_card_default_test_score": 0.8158159852027893 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4507676362991333, + "lr": 3.942649982491275e-06, + "credit_card_default_val_accuracy": 0.8109570741653442, + "credit_card_default_val_precision": 0.7910326719284058, + "credit_card_default_val_recall": 0.8109570741653442, + "credit_card_default_val_f1-score": 0.7879649996757507, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7503932118415833, + "credit_card_default_val_score": 0.8109570741653442 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44750547409057617, + "lr": 3.942650891985977e-06, + "credit_card_default_test_accuracy": 0.8144727945327759, + "credit_card_default_test_precision": 0.7970324158668518, + "credit_card_default_test_recall": 0.8144727945327759, + "credit_card_default_test_f1-score": 0.794328510761261, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7594409584999084, + "credit_card_default_test_score": 0.8144727945327759 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.4513669013977051, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8106188774108887, + "credit_card_default_val_precision": 0.7906190752983093, + "credit_card_default_val_recall": 0.8106188774108887, + "credit_card_default_val_f1-score": 0.7881050705909729, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7520156502723694, + "credit_card_default_val_score": 0.8106188774108887 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44810912013053894, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8126259446144104, + "credit_card_default_test_precision": 0.7946444153785706, + "credit_card_default_test_recall": 0.8126259446144104, + "credit_card_default_test_f1-score": 0.7928016185760498, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7593835592269897, + "credit_card_default_test_score": 0.8126259446144104 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.45220452547073364, + "lr": 9.86635996014229e-07, + "credit_card_default_val_accuracy": 0.8129861354827881, + "credit_card_default_val_precision": 0.7938436269760132, + "credit_card_default_val_recall": 0.8129861354827881, + "credit_card_default_val_f1-score": 0.7889845967292786, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7494129538536072, + "credit_card_default_val_score": 0.8129861354827881 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44776108860969543, + "lr": 9.866355412668781e-07, + "credit_card_default_test_accuracy": 0.8158159852027893, + "credit_card_default_test_precision": 0.7988069653511047, + "credit_card_default_test_recall": 0.8158159852027893, + "credit_card_default_test_f1-score": 0.7953466176986694, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7596734166145325, + "credit_card_default_test_score": 0.8158159852027893 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.44736173152923586, + "std": 0.0007735398310272306, + "n": 5, + "ci95": 0.0009603225807708227 + }, + "lr": { + "mean": 9.779485071703676e-06, + "std": 1.4573928903195385e-05, + "n": 5, + "ci95": 1.8093021787515935e-05 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8144728064537048, + "std": 0.001389594356623852, + "n": 5, + "ci95": 0.0017251326761098772 + }, + "credit_card_default_test_precision": { + "mean": 0.7970494866371155, + "std": 0.001823788990087108, + "n": 5, + "ci95": 0.002264170091171697 + }, + "credit_card_default_test_recall": { + "mean": 0.8144728064537048, + "std": 0.001389594356623852, + "n": 5, + "ci95": 0.0017251326761098772 + }, + "credit_card_default_test_f1-score": { + "mean": 0.794100308418274, + "std": 0.0009688002276090045, + "n": 5, + "ci95": 0.0012027315175139163 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7602586269378662, + "std": 0.0011310023747805697, + "n": 5, + "ci95": 0.0014040997965998358 + }, + "credit_card_default_test_score": { + "mean": 0.8144728064537048, + "std": 0.001389594356623852, + "n": 5, + "ci95": 0.0017251326761098772 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/downstream_results/fttransformer_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/downstream_results/fttransformer_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..3f83b2741ad106d675ede0288f4b84b9b423bbce --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/downstream_results/fttransformer_credit_card_default_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "credit_card_default", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpagtdlcvv", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpagtdlcvv/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.44726452231407166, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8143388628959656, + "credit_card_default_val_precision": 0.7957068681716919, + "credit_card_default_val_recall": 0.8143388628959656, + "credit_card_default_val_f1-score": 0.791758120059967, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7507679462432861, + "credit_card_default_val_score": 0.8143388628959656 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4400160014629364, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8190060257911682, + "credit_card_default_test_precision": 0.8030407428741455, + "credit_card_default_test_recall": 0.8190060257911682, + "credit_card_default_test_f1-score": 0.7994917035102844, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7647790312767029, + "credit_card_default_test_score": 0.8190060257911682 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.44776108860969543, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8150151968002319, + "credit_card_default_val_precision": 0.7966932058334351, + "credit_card_default_val_recall": 0.8150151968002319, + "credit_card_default_val_f1-score": 0.7914837002754211, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7525211572647095, + "credit_card_default_val_score": 0.8150151968002319 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44099584221839905, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8185023665428162, + "credit_card_default_test_precision": 0.8026161789894104, + "credit_card_default_test_recall": 0.8185023665428162, + "credit_card_default_test_f1-score": 0.7966155409812927, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7639510631561279, + "credit_card_default_test_score": 0.8185023665428162 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4494698643684387, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8160297870635986, + "credit_card_default_val_precision": 0.7982540726661682, + "credit_card_default_val_recall": 0.8160297870635986, + "credit_card_default_val_f1-score": 0.7916812300682068, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7508789300918579, + "credit_card_default_val_score": 0.8160297870635986 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4420795738697052, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8176628351211548, + "credit_card_default_test_precision": 0.8013173937797546, + "credit_card_default_test_recall": 0.8176628351211548, + "credit_card_default_test_f1-score": 0.7967845797538757, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.761913001537323, + "credit_card_default_test_score": 0.8176628351211548 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.44779202342033386, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8156915903091431, + "credit_card_default_val_precision": 0.7979090213775635, + "credit_card_default_val_recall": 0.8156915903091431, + "credit_card_default_val_f1-score": 0.7905467748641968, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7489837408065796, + "credit_card_default_val_score": 0.8156915903091431 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44313955307006836, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8181665539741516, + "credit_card_default_test_precision": 0.8021530508995056, + "credit_card_default_test_recall": 0.8181665539741516, + "credit_card_default_test_f1-score": 0.7962392568588257, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.759540319442749, + "credit_card_default_test_score": 0.8181665539741516 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.44690796732902527, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8133243322372437, + "credit_card_default_val_precision": 0.7943000197410583, + "credit_card_default_val_recall": 0.8133243322372437, + "credit_card_default_val_f1-score": 0.7903100252151489, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.752513587474823, + "credit_card_default_val_score": 0.8133243322372437 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.43984168767929077, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8179986476898193, + "credit_card_default_test_precision": 0.8018132448196411, + "credit_card_default_test_recall": 0.8179986476898193, + "credit_card_default_test_f1-score": 0.7967769503593445, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7656611800193787, + "credit_card_default_test_score": 0.8179986476898193 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.44121453166007996, + "std": 0.0013984818334808497, + "n": 5, + "ci95": 0.0017361661670427474 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_accuracy": { + "mean": 0.818267285823822, + "std": 0.0005120238491346953, + "n": 5, + "ci95": 0.0006356596577118361 + }, + "credit_card_default_test_precision": { + "mean": 0.8021881222724915, + "std": 0.0006727045520112721, + "n": 5, + "ci95": 0.0008351391170456787 + }, + "credit_card_default_test_recall": { + "mean": 0.818267285823822, + "std": 0.0005120238491346953, + "n": 5, + "ci95": 0.0006356596577118361 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7971816062927246, + "std": 0.0013101887310540189, + "n": 5, + "ci95": 0.0016265533758381898 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7631689190864563, + "std": 0.0024576228063520936, + "n": 5, + "ci95": 0.0030510525525532925 + }, + "credit_card_default_test_score": { + "mean": 0.818267285823822, + "std": 0.0005120238491346953, + "n": 5, + "ci95": 0.0006356596577118361 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/downstream_results/mlp_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/downstream_results/mlp_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..26d39116325b6c3cf157fc6c675a2f99f36912c0 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/downstream_results/mlp_credit_card_default_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "credit_card_default", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpjq3mwp91", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpjq3mwp91/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.456789493560791, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8123098015785217, + "credit_card_default_val_precision": 0.7928820848464966, + "credit_card_default_val_recall": 0.8123098015785217, + "credit_card_default_val_f1-score": 0.7886457443237305, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.73887699842453, + "credit_card_default_val_score": 0.8123098015785217 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.45091870427131653, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8183344602584839, + "credit_card_default_test_precision": 0.8021538257598877, + "credit_card_default_test_recall": 0.8183344602584839, + "credit_card_default_test_f1-score": 0.7986555695533752, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7522221207618713, + "credit_card_default_test_score": 0.8183344602584839 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.4575861692428589, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8099424839019775, + "credit_card_default_val_precision": 0.7934061884880066, + "credit_card_default_val_recall": 0.8099424839019775, + "credit_card_default_val_f1-score": 0.773453950881958, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7348690629005432, + "credit_card_default_val_score": 0.8099424839019775 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.45597708225250244, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8107790350914001, + "credit_card_default_test_precision": 0.7945290803909302, + "credit_card_default_test_recall": 0.8107790350914001, + "credit_card_default_test_f1-score": 0.7787691950798035, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7376885414123535, + "credit_card_default_test_score": 0.8107790350914001 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4570556879043579, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8133243322372437, + "credit_card_default_val_precision": 0.7943274974822998, + "credit_card_default_val_recall": 0.8133243322372437, + "credit_card_default_val_f1-score": 0.7892600297927856, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7351929545402527, + "credit_card_default_val_score": 0.8133243322372437 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4498947262763977, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8176628351211548, + "credit_card_default_test_precision": 0.8012663125991821, + "credit_card_default_test_recall": 0.8176628351211548, + "credit_card_default_test_f1-score": 0.7977256178855896, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7465294599533081, + "credit_card_default_test_score": 0.8176628351211548 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.4659510552883148, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8089279532432556, + "credit_card_default_val_precision": 0.7939008474349976, + "credit_card_default_val_recall": 0.8089279532432556, + "credit_card_default_val_f1-score": 0.7692520022392273, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7287012338638306, + "credit_card_default_val_score": 0.8089279532432556 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.46061089634895325, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.809603750705719, + "credit_card_default_test_precision": 0.7940508723258972, + "credit_card_default_test_recall": 0.809603750705719, + "credit_card_default_test_f1-score": 0.7750416398048401, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7370550632476807, + "credit_card_default_test_score": 0.809603750705719 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.47154441475868225, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8116334080696106, + "credit_card_default_val_precision": 0.7974916100502014, + "credit_card_default_val_recall": 0.8116334080696106, + "credit_card_default_val_f1-score": 0.774224579334259, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7265719771385193, + "credit_card_default_val_score": 0.8116334080696106 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.462413489818573, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8101074695587158, + "credit_card_default_test_precision": 0.7948746085166931, + "credit_card_default_test_recall": 0.8101074695587158, + "credit_card_default_test_f1-score": 0.775701642036438, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.74432772397995, + "credit_card_default_test_score": 0.8101074695587158 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.4559629797935486, + "std": 0.005600911154912593, + "n": 5, + "ci95": 0.006953334837084047 + }, + "lr": { + "mean": 0.0009999994654208422, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8132975101470947, + "std": 0.004318271150767074, + "n": 5, + "ci95": 0.005360982239874915 + }, + "credit_card_default_test_precision": { + "mean": 0.797374939918518, + "std": 0.003980594495634056, + "n": 5, + "ci95": 0.004941768511096697 + }, + "credit_card_default_test_recall": { + "mean": 0.8132975101470947, + "std": 0.004318271150767074, + "n": 5, + "ci95": 0.005360982239874915 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7851787328720092, + "std": 0.011965654304592491, + "n": 5, + "ci95": 0.014854940316568206 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7435645818710327, + "std": 0.006348770584675134, + "n": 5, + "ci95": 0.007881776100011176 + }, + "credit_card_default_test_score": { + "mean": 0.8132975101470947, + "std": 0.004318271150767074, + "n": 5, + "ci95": 0.005360982239874915 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/downstream_results/resnet_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/downstream_results/resnet_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..a46ab751f49ed9dc714685759bd0461b5e00d99a --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/downstream_results/resnet_credit_card_default_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "credit_card_default", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpwa750ghg", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpwa750ghg/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.45115408301353455, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8068988919258118, + "credit_card_default_val_precision": 0.7871233224868774, + "credit_card_default_val_recall": 0.8068988919258118, + "credit_card_default_val_f1-score": 0.7888129949569702, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7570883631706238, + "credit_card_default_val_score": 0.8068988919258118 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.443339079618454, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8171591758728027, + "credit_card_default_test_precision": 0.8012056350708008, + "credit_card_default_test_recall": 0.8171591758728027, + "credit_card_default_test_f1-score": 0.8014456629753113, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7628201842308044, + "credit_card_default_test_score": 0.8171591758728027 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.4531085193157196, + "lr": 0.0005000000237487257, + "credit_card_default_val_accuracy": 0.8048697710037231, + "credit_card_default_val_precision": 0.7840221524238586, + "credit_card_default_val_recall": 0.8048697710037231, + "credit_card_default_val_f1-score": 0.7854407429695129, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7518139481544495, + "credit_card_default_val_score": 0.8048697710037231 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4463523030281067, + "lr": 0.0004999997327104211, + "credit_card_default_test_accuracy": 0.8166554570198059, + "credit_card_default_test_precision": 0.8007574677467346, + "credit_card_default_test_recall": 0.8166554570198059, + "credit_card_default_test_f1-score": 0.8014382123947144, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7634868025779724, + "credit_card_default_test_score": 0.8166554570198059 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4580335021018982, + "lr": 0.0005000000237487257, + "credit_card_default_val_accuracy": 0.8109570741653442, + "credit_card_default_val_precision": 0.791175901889801, + "credit_card_default_val_recall": 0.8109570741653442, + "credit_card_default_val_f1-score": 0.7891989350318909, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7497589588165283, + "credit_card_default_val_score": 0.8109570741653442 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44910746812820435, + "lr": 0.0004999997327104211, + "credit_card_default_test_accuracy": 0.8174949884414673, + "credit_card_default_test_precision": 0.8012334108352661, + "credit_card_default_test_recall": 0.8174949884414673, + "credit_card_default_test_f1-score": 0.8001068234443665, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7649736404418945, + "credit_card_default_test_score": 0.8174949884414673 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.45125502347946167, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8116334080696106, + "credit_card_default_val_precision": 0.791999340057373, + "credit_card_default_val_recall": 0.8116334080696106, + "credit_card_default_val_f1-score": 0.7891373634338379, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7552090287208557, + "credit_card_default_val_score": 0.8116334080696106 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4443984627723694, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8174949884414673, + "credit_card_default_test_precision": 0.8010596632957458, + "credit_card_default_test_recall": 0.8174949884414673, + "credit_card_default_test_f1-score": 0.7985050678253174, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.764964759349823, + "credit_card_default_test_score": 0.8174949884414673 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.45458972454071045, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8099424839019775, + "credit_card_default_val_precision": 0.789559543132782, + "credit_card_default_val_recall": 0.8099424839019775, + "credit_card_default_val_f1-score": 0.783901572227478, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7507177591323853, + "credit_card_default_val_score": 0.8099424839019775 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44696947932243347, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8168233633041382, + "credit_card_default_test_precision": 0.8004374504089355, + "credit_card_default_test_recall": 0.8168233633041382, + "credit_card_default_test_f1-score": 0.7939406037330627, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7622969746589661, + "credit_card_default_test_score": 0.8168233633041382 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.4460333585739136, + "std": 0.0022561150538143587, + "n": 5, + "ci95": 0.0028008877424162515 + }, + "lr": { + "mean": 0.0005999996792525053, + "std": 0.00041832978963653156, + "n": 5, + "ci95": 0.0005193417676547898 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8171255946159363, + "std": 0.00038288523823073767, + "n": 5, + "ci95": 0.00047533859973120053 + }, + "credit_card_default_test_precision": { + "mean": 0.8009387254714966, + "std": 0.00033794866706583895, + "n": 5, + "ci95": 0.0004195514220563793 + }, + "credit_card_default_test_recall": { + "mean": 0.8171255946159363, + "std": 0.00038288523823073767, + "n": 5, + "ci95": 0.00047533859973120053 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7990872740745545, + "std": 0.003120719862088489, + "n": 5, + "ci95": 0.003874264299801887 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7637084722518921, + "std": 0.0012257065644966931, + "n": 5, + "ci95": 0.001521671727908433 + }, + "credit_card_default_test_score": { + "mean": 0.8171255946159363, + "std": 0.00038288523823073767, + "n": 5, + "ci95": 0.00047533859973120053 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..1fc7c49f8c6d4ee1f2710b357e47a0f3cc159e94 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:db71cf3f6d0f18efc7a9e05d8937e97b529b985ee334e4f17fe3c202cc535494 +size 176640128 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..a6133af641fe80080a1964323ddccd55ef7a7f14 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1e9ad677e26f2d8a4cb101425ea45cf47d8c6d7a9eb3039e68d486aa2aac1a03 +size 15531742 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..397113492a0b53de8865bde61f6801222b3a5d37 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.00125_sx0.001_sy0.0001/seed_42/wjepa_deterministic/credit_card_default/20260916_015657/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: credit_card_default +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.001 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.3 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.1 +mask_min_trgt_share: 0.1 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 256 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.1 +recon_y_weight: 1.0 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.001 +sw_s_y_weight: 0.0001 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 15 +vtjepa_kl_x_end_weight: 0.0001 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 15 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 15 +wjepa_joint_end_weight: 0.00125 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..76d861038ca2aa6c899999f85e5e66335d560ce7 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-16T02:06:46.213016+00:00", + "spec": { + "dataset": "credit_card_default", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.002, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0002, + "lambda_joint": 0.0025, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..ee9abb67b8306edf3f771504086c2102155e3a3f --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "credit_card_default", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.002, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0002, + "lambda_joint": 0.0025, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "credit_card_default", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.0025", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.002", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "0.0002", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..9069de1c6b88bdba5c144b7fccb5a22d5fd2d343 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "credit_card_default", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.0025, + "s_x": 0.002, + "z": 0.0, + "s_y": 0.0002 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 15 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "credit_card_default", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.0025", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.002", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "0.0002", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/downstream_results/autoint_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/downstream_results/autoint_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..523436033253b3ee3574bcc6d02b2344df5fae41 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/downstream_results/autoint_credit_card_default_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "credit_card_default", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpvxwroc9o", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpvxwroc9o/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.44814422726631165, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8136624693870544, + "credit_card_default_val_precision": 0.7947720289230347, + "credit_card_default_val_recall": 0.8136624693870544, + "credit_card_default_val_f1-score": 0.7909994721412659, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7529870271682739, + "credit_card_default_val_score": 0.8136624693870544 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4424729645252228, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8183344602584839, + "credit_card_default_test_precision": 0.8021537065505981, + "credit_card_default_test_recall": 0.8183344602584839, + "credit_card_default_test_f1-score": 0.7992960214614868, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7644686102867126, + "credit_card_default_test_score": 0.8183344602584839 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.4491504728794098, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8119716048240662, + "credit_card_default_val_precision": 0.7924071550369263, + "credit_card_default_val_recall": 0.8119716048240662, + "credit_card_default_val_f1-score": 0.7881590127944946, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7481553554534912, + "credit_card_default_val_score": 0.8119716048240662 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4436683654785156, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8148086071014404, + "credit_card_default_test_precision": 0.7974652051925659, + "credit_card_default_test_recall": 0.8148086071014404, + "credit_card_default_test_f1-score": 0.7944173216819763, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7652016282081604, + "credit_card_default_test_score": 0.8148086071014404 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4515869915485382, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8180588483810425, + "credit_card_default_val_precision": 0.8011156320571899, + "credit_card_default_val_recall": 0.8180588483810425, + "credit_card_default_val_f1-score": 0.7941885590553284, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7427580952644348, + "credit_card_default_val_score": 0.8180588483810425 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4479624330997467, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8119543194770813, + "credit_card_default_test_precision": 0.7936046719551086, + "credit_card_default_test_recall": 0.8119543194770813, + "credit_card_default_test_f1-score": 0.7905204892158508, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7579166293144226, + "credit_card_default_test_score": 0.8119543194770813 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.4459420442581177, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8150151968002319, + "credit_card_default_val_precision": 0.7967579364776611, + "credit_card_default_val_recall": 0.8150151968002319, + "credit_card_default_val_f1-score": 0.7908518314361572, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7572458386421204, + "credit_card_default_val_score": 0.8150151968002319 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4425094723701477, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8179986476898193, + "credit_card_default_test_precision": 0.8017909526824951, + "credit_card_default_test_recall": 0.8179986476898193, + "credit_card_default_test_f1-score": 0.7969683408737183, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7645324468612671, + "credit_card_default_test_score": 0.8179986476898193 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.45179978013038635, + "lr": 0.00015772644837852567, + "credit_card_default_val_accuracy": 0.8136624693870544, + "credit_card_default_val_precision": 0.7947996854782104, + "credit_card_default_val_recall": 0.8136624693870544, + "credit_card_default_val_f1-score": 0.7897477149963379, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7513718605041504, + "credit_card_default_val_score": 0.8136624693870544 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4449441432952881, + "lr": 0.00015772644837852567, + "credit_card_default_test_accuracy": 0.8188381195068359, + "credit_card_default_test_precision": 0.8029009103775024, + "credit_card_default_test_recall": 0.8188381195068359, + "credit_card_default_test_f1-score": 0.7981417179107666, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7674136757850647, + "credit_card_default_test_score": 0.8188381195068359 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.44431147575378416, + "std": 0.00227874353089866, + "n": 5, + "ci95": 0.0028289802033870746 + }, + "lr": { + "mean": 3.154528967570513e-05, + "std": 7.053741208479897e-05, + "n": 5, + "ci95": 8.756972413975744e-05 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8163868308067321, + "std": 0.002938451003020982, + "n": 5, + "ci95": 0.003647983901413843 + }, + "credit_card_default_test_precision": { + "mean": 0.799583089351654, + "std": 0.003959721502497829, + "n": 5, + "ci95": 0.0049158554219020855 + }, + "credit_card_default_test_recall": { + "mean": 0.8163868308067321, + "std": 0.002938451003020982, + "n": 5, + "ci95": 0.003647983901413843 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7958687782287598, + "std": 0.0034940023775502, + "n": 5, + "ci95": 0.004337681455876162 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7639065980911255, + "std": 0.0035554172502079585, + "n": 5, + "ci95": 0.004413925867143375 + }, + "credit_card_default_test_score": { + "mean": 0.8163868308067321, + "std": 0.002938451003020982, + "n": 5, + "ci95": 0.003647983901413843 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/downstream_results/dcnv2_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/downstream_results/dcnv2_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..96fb1b72dc53f64bfad702f05fa158cf7e0d9f09 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/downstream_results/dcnv2_credit_card_default_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "credit_card_default", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpd3u_3r_8", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpd3u_3r_8/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.4500962495803833, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8126479387283325, + "credit_card_default_val_precision": 0.7933695316314697, + "credit_card_default_val_recall": 0.8126479387283325, + "credit_card_default_val_f1-score": 0.7884964346885681, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7485284209251404, + "credit_card_default_val_score": 0.8126479387283325 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4447903633117676, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8148086071014404, + "credit_card_default_test_precision": 0.7974621653556824, + "credit_card_default_test_recall": 0.8148086071014404, + "credit_card_default_test_f1-score": 0.7943224310874939, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.764095664024353, + "credit_card_default_test_score": 0.8148086071014404 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.4501810669898987, + "lr": 0.00011474337225081399, + "credit_card_default_val_accuracy": 0.8163679242134094, + "credit_card_default_val_precision": 0.7985750436782837, + "credit_card_default_val_recall": 0.8163679242134094, + "credit_card_default_val_f1-score": 0.7932155728340149, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7453614473342896, + "credit_card_default_val_score": 0.8163679242134094 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44515568017959595, + "lr": 0.00011474342318251729, + "credit_card_default_test_accuracy": 0.8154801726341248, + "credit_card_default_test_precision": 0.7983880043029785, + "credit_card_default_test_recall": 0.8154801726341248, + "credit_card_default_test_f1-score": 0.7958186864852905, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7616989016532898, + "credit_card_default_test_score": 0.8154801726341248 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4535575807094574, + "lr": 3.942649982491275e-06, + "credit_card_default_val_accuracy": 0.8119716048240662, + "credit_card_default_val_precision": 0.7924088835716248, + "credit_card_default_val_recall": 0.8119716048240662, + "credit_card_default_val_f1-score": 0.7883705496788025, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7444469332695007, + "credit_card_default_val_score": 0.8119716048240662 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4458691477775574, + "lr": 3.942650891985977e-06, + "credit_card_default_test_accuracy": 0.8174949884414673, + "credit_card_default_test_precision": 0.801041305065155, + "credit_card_default_test_recall": 0.8174949884414673, + "credit_card_default_test_f1-score": 0.7978639006614685, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7611337900161743, + "credit_card_default_test_score": 0.8174949884414673 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.4504140317440033, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.81400066614151, + "credit_card_default_val_precision": 0.7952708005905151, + "credit_card_default_val_recall": 0.81400066614151, + "credit_card_default_val_f1-score": 0.7902348041534424, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7495315074920654, + "credit_card_default_val_score": 0.81400066614151 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44611531496047974, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8164875507354736, + "credit_card_default_test_precision": 0.7997313737869263, + "credit_card_default_test_recall": 0.8164875507354736, + "credit_card_default_test_f1-score": 0.7954267859458923, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7607540488243103, + "credit_card_default_test_score": 0.8164875507354736 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.4471879005432129, + "lr": 0.00013551565643865615, + "credit_card_default_val_accuracy": 0.818397045135498, + "credit_card_default_val_precision": 0.802642285823822, + "credit_card_default_val_recall": 0.818397045135498, + "credit_card_default_val_f1-score": 0.7914347052574158, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7496649026870728, + "credit_card_default_val_score": 0.818397045135498 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4424583613872528, + "lr": 0.00013551570009440184, + "credit_card_default_test_accuracy": 0.8166554570198059, + "credit_card_default_test_precision": 0.8004509210586548, + "credit_card_default_test_recall": 0.8166554570198059, + "credit_card_default_test_f1-score": 0.7926862835884094, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7640072107315063, + "credit_card_default_test_score": 0.8166554570198059 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.4448777735233307, + "std": 0.0014536814867335472, + "n": 5, + "ci95": 0.001804694601317283 + }, + "lr": { + "mean": 5.084035483378102e-05, + "std": 6.823194675819088e-05, + "n": 5, + "ci95": 8.470756976383366e-05 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8161853551864624, + "std": 0.0010512105959205932, + "n": 5, + "ci95": 0.0013050411005565419 + }, + "credit_card_default_test_precision": { + "mean": 0.7994147539138794, + "std": 0.0014736329245035318, + "n": 5, + "ci95": 0.0018294636118334144 + }, + "credit_card_default_test_recall": { + "mean": 0.8161853551864624, + "std": 0.0010512105959205932, + "n": 5, + "ci95": 0.0013050411005565419 + }, + "credit_card_default_test_f1-score": { + "mean": 0.795223617553711, + "std": 0.0019115883458781257, + "n": 5, + "ci95": 0.002373169913238103 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7623379230499268, + "std": 0.0016002434945935737, + "n": 5, + "ci95": 0.001986646195773884 + }, + "credit_card_default_test_score": { + "mean": 0.8161853551864624, + "std": 0.0010512105959205932, + "n": 5, + "ci95": 0.0013050411005565419 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/downstream_results/fttransformer_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/downstream_results/fttransformer_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..168df84b404015299d9fe37c61df391d6cf413fc --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/downstream_results/fttransformer_credit_card_default_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "credit_card_default", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp8f3v206v", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp8f3v206v/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.4480937719345093, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8129861354827881, + "credit_card_default_val_precision": 0.7938436269760132, + "credit_card_default_val_recall": 0.8129861354827881, + "credit_card_default_val_f1-score": 0.7889845967292786, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7513830661773682, + "credit_card_default_val_score": 0.8129861354827881 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44127994775772095, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8196776509284973, + "credit_card_default_test_precision": 0.80404132604599, + "credit_card_default_test_recall": 0.8196776509284973, + "credit_card_default_test_f1-score": 0.799030065536499, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7642912864685059, + "credit_card_default_test_score": 0.8196776509284973 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.4463687241077423, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.811295211315155, + "credit_card_default_val_precision": 0.7914800047874451, + "credit_card_default_val_recall": 0.811295211315155, + "credit_card_default_val_f1-score": 0.7880308032035828, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7496252655982971, + "credit_card_default_val_score": 0.811295211315155 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4416617453098297, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8185023665428162, + "credit_card_default_test_precision": 0.8023770451545715, + "credit_card_default_test_recall": 0.8185023665428162, + "credit_card_default_test_f1-score": 0.7996172904968262, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7623673677444458, + "credit_card_default_test_score": 0.8185023665428162 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4481878876686096, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8143388628959656, + "credit_card_default_val_precision": 0.7957068681716919, + "credit_card_default_val_recall": 0.8143388628959656, + "credit_card_default_val_f1-score": 0.791758120059967, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7483878135681152, + "credit_card_default_val_score": 0.8143388628959656 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44245412945747375, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8158159852027893, + "credit_card_default_test_precision": 0.7988305687904358, + "credit_card_default_test_recall": 0.8158159852027893, + "credit_card_default_test_f1-score": 0.7961902618408203, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7612103819847107, + "credit_card_default_test_score": 0.8158159852027893 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.46258658170700073, + "lr": 0.00023208656057249755, + "credit_card_default_val_accuracy": 0.8089279532432556, + "credit_card_default_val_precision": 0.7897705435752869, + "credit_card_default_val_recall": 0.8089279532432556, + "credit_card_default_val_f1-score": 0.7912172079086304, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7044989466667175, + "credit_card_default_val_score": 0.8089279532432556 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4565331041812897, + "lr": 0.0002320865314686671, + "credit_card_default_test_accuracy": 0.815648078918457, + "credit_card_default_test_precision": 0.7998767495155334, + "credit_card_default_test_recall": 0.815648078918457, + "credit_card_default_test_f1-score": 0.8012471199035645, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7112227082252502, + "credit_card_default_test_score": 0.815648078918457 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.44638821482658386, + "lr": 9.86635996014229e-07, + "credit_card_default_val_accuracy": 0.8150151968002319, + "credit_card_default_val_precision": 0.7966774106025696, + "credit_card_default_val_recall": 0.8150151968002319, + "credit_card_default_val_f1-score": 0.7916922569274902, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7519859671592712, + "credit_card_default_val_score": 0.8150151968002319 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44145703315734863, + "lr": 9.866355412668781e-07, + "credit_card_default_test_accuracy": 0.8190060257911682, + "credit_card_default_test_precision": 0.8031989932060242, + "credit_card_default_test_recall": 0.8190060257911682, + "credit_card_default_test_f1-score": 0.7978062629699707, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7633851766586304, + "credit_card_default_test_score": 0.8190060257911682 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.44467719197273253, + "std": 0.006642820765181367, + "n": 5, + "ci95": 0.00824682909003611 + }, + "lr": { + "mean": 4.6614633401986796e-05, + "std": 0.00010368282319511726, + "n": 5, + "ci95": 0.0001287185899918253 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8177300214767456, + "std": 0.0018718963691182926, + "n": 5, + "ci95": 0.0023238937156475014 + }, + "credit_card_default_test_precision": { + "mean": 0.801664936542511, + "std": 0.0022214245987620057, + "n": 5, + "ci95": 0.002757820758677676 + }, + "credit_card_default_test_recall": { + "mean": 0.8177300214767456, + "std": 0.0018718963691182926, + "n": 5, + "ci95": 0.0023238937156475014 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7987782001495362, + "std": 0.0019042820239290352, + "n": 5, + "ci95": 0.00236409937068986 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7524953842163086, + "std": 0.02310071382948042, + "n": 5, + "ci95": 0.028678726333865968 + }, + "credit_card_default_test_score": { + "mean": 0.8177300214767456, + "std": 0.0018718963691182926, + "n": 5, + "ci95": 0.0023238937156475014 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/downstream_results/mlp_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/downstream_results/mlp_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..6cab04d48f3dbd006db335ac53af8cd6f86d0a79 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/downstream_results/mlp_credit_card_default_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "credit_card_default", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp3vr8k1lp", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp3vr8k1lp/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.45163774490356445, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8116334080696106, + "credit_card_default_val_precision": 0.7919310331344604, + "credit_card_default_val_recall": 0.8116334080696106, + "credit_card_default_val_f1-score": 0.787671685218811, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7453951239585876, + "credit_card_default_val_score": 0.8116334080696106 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44573184847831726, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8163197040557861, + "credit_card_default_test_precision": 0.7995646595954895, + "credit_card_default_test_recall": 0.8163197040557861, + "credit_card_default_test_f1-score": 0.7945131063461304, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7585065960884094, + "credit_card_default_test_score": 0.8163197040557861 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.4491868019104004, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.81400066614151, + "credit_card_default_val_precision": 0.7952436804771423, + "credit_card_default_val_recall": 0.81400066614151, + "credit_card_default_val_f1-score": 0.7908625602722168, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7520341873168945, + "credit_card_default_val_score": 0.81400066614151 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4471282660961151, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8169912695884705, + "credit_card_default_test_precision": 0.8003919124603271, + "credit_card_default_test_recall": 0.8169912695884705, + "credit_card_default_test_f1-score": 0.7963217496871948, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7499787211418152, + "credit_card_default_test_score": 0.8169912695884705 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.49760714173316956, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.7954007387161255, + "credit_card_default_val_precision": 0.782047688961029, + "credit_card_default_val_recall": 0.7954007387161255, + "credit_card_default_val_f1-score": 0.7867961525917053, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7253882884979248, + "credit_card_default_val_score": 0.7954007387161255 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.5010742545127869, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.7988582849502563, + "credit_card_default_test_precision": 0.7889818549156189, + "credit_card_default_test_recall": 0.7988582849502563, + "credit_card_default_test_f1-score": 0.7928103804588318, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7286624312400818, + "credit_card_default_test_score": 0.7988582849502563 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.4692836105823517, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8116334080696106, + "credit_card_default_val_precision": 0.8010846972465515, + "credit_card_default_val_recall": 0.8116334080696106, + "credit_card_default_val_f1-score": 0.7707598805427551, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7232521176338196, + "credit_card_default_val_score": 0.8116334080696106 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.485039621591568, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8082605600357056, + "credit_card_default_test_precision": 0.7954111695289612, + "credit_card_default_test_recall": 0.8082605600357056, + "credit_card_default_test_f1-score": 0.7689459919929504, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7412496209144592, + "credit_card_default_test_score": 0.8082605600357056 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.48940467834472656, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8075752258300781, + "credit_card_default_val_precision": 0.7959722876548767, + "credit_card_default_val_recall": 0.8075752258300781, + "credit_card_default_val_f1-score": 0.7630168199539185, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7367837429046631, + "credit_card_default_val_score": 0.8075752258300781 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.48302894830703735, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8030557632446289, + "credit_card_default_test_precision": 0.7879353761672974, + "credit_card_default_test_recall": 0.8030557632446289, + "credit_card_default_test_f1-score": 0.7597498893737793, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7535432577133179, + "credit_card_default_test_score": 0.8030557632446289 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.47240058779716493, + "std": 0.02472245075998964, + "n": 5, + "ci95": 0.030692055876793082 + }, + "lr": { + "mean": 0.0009999994654208422, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8086971163749694, + "std": 0.007995543616028306, + "n": 5, + "ci95": 0.00992618708439809 + }, + "credit_card_default_test_precision": { + "mean": 0.7944569945335388, + "std": 0.005803675625310408, + "n": 5, + "ci95": 0.007205059818385242 + }, + "credit_card_default_test_recall": { + "mean": 0.8086971163749694, + "std": 0.007995543616028306, + "n": 5, + "ci95": 0.00992618708439809 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7824682235717774, + "std": 0.016903646437664413, + "n": 5, + "ci95": 0.02098528442924353 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7463881254196167, + "std": 0.011742551786419194, + "n": 5, + "ci95": 0.014577966361983172 + }, + "credit_card_default_test_score": { + "mean": 0.8086971163749694, + "std": 0.007995543616028306, + "n": 5, + "ci95": 0.00992618708439809 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/downstream_results/resnet_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/downstream_results/resnet_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..560e825ddaa83b39e31bd3a573cf0aa0d176d194 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/downstream_results/resnet_credit_card_default_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "credit_card_default", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpxbeexic8", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpxbeexic8/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.45282119512557983, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8052079677581787, + "credit_card_default_val_precision": 0.7855105996131897, + "credit_card_default_val_recall": 0.8052079677581787, + "credit_card_default_val_f1-score": 0.7879908084869385, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7529222965240479, + "credit_card_default_val_score": 0.8052079677581787 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44551023840904236, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8159838914871216, + "credit_card_default_test_precision": 0.80002361536026, + "credit_card_default_test_recall": 0.8159838914871216, + "credit_card_default_test_f1-score": 0.8009580969810486, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7607998847961426, + "credit_card_default_test_score": 0.8159838914871216 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.45112910866737366, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8079134225845337, + "credit_card_default_val_precision": 0.7867250442504883, + "credit_card_default_val_recall": 0.8079134225845337, + "credit_card_default_val_f1-score": 0.7795129418373108, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7525303959846497, + "credit_card_default_val_score": 0.8079134225845337 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4451403021812439, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8129616975784302, + "credit_card_default_test_precision": 0.7955929636955261, + "credit_card_default_test_recall": 0.8129616975784302, + "credit_card_default_test_f1-score": 0.7866913676261902, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7629019021987915, + "credit_card_default_test_score": 0.8129616975784302 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4599471986293793, + "lr": 0.0005000000237487257, + "credit_card_default_val_accuracy": 0.8126479387283325, + "credit_card_default_val_precision": 0.7934526205062866, + "credit_card_default_val_recall": 0.8126479387283325, + "credit_card_default_val_f1-score": 0.7909838557243347, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7490652799606323, + "credit_card_default_val_score": 0.8126479387283325 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.45053836703300476, + "lr": 0.0004999997327104211, + "credit_card_default_test_accuracy": 0.8186702728271484, + "credit_card_default_test_precision": 0.8026127815246582, + "credit_card_default_test_recall": 0.8186702728271484, + "credit_card_default_test_f1-score": 0.80020672082901, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.764349639415741, + "credit_card_default_test_score": 0.8186702728271484 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.45513010025024414, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.811295211315155, + "credit_card_default_val_precision": 0.7915080189704895, + "credit_card_default_val_recall": 0.811295211315155, + "credit_card_default_val_f1-score": 0.7884483337402344, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.749649703502655, + "credit_card_default_val_score": 0.811295211315155 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44853633642196655, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.815648078918457, + "credit_card_default_test_precision": 0.7985826730728149, + "credit_card_default_test_recall": 0.815648078918457, + "credit_card_default_test_f1-score": 0.7952074408531189, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7627364993095398, + "credit_card_default_test_score": 0.815648078918457 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.45586690306663513, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8085898160934448, + "credit_card_default_val_precision": 0.787668764591217, + "credit_card_default_val_recall": 0.8085898160934448, + "credit_card_default_val_f1-score": 0.7809908986091614, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7510380744934082, + "credit_card_default_val_score": 0.8085898160934448 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44754841923713684, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8143048882484436, + "credit_card_default_test_precision": 0.7971880435943604, + "credit_card_default_test_recall": 0.8143048882484436, + "credit_card_default_test_f1-score": 0.7896102070808411, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7629051804542542, + "credit_card_default_test_score": 0.8143048882484436 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.4474547326564789, + "std": 0.0022262688040763858, + "n": 5, + "ci95": 0.0027638346697430073 + }, + "lr": { + "mean": 0.00029999983962625265, + "std": 0.0004472133564288907, + "n": 5, + "ci95": 0.0005551997032016516 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8155137658119201, + "std": 0.002131726526206059, + "n": 5, + "ci95": 0.0026464637463145175 + }, + "credit_card_default_test_precision": { + "mean": 0.798800015449524, + "std": 0.0026909688686707982, + "n": 5, + "ci95": 0.003340743508067539 + }, + "credit_card_default_test_recall": { + "mean": 0.8155137658119201, + "std": 0.002131726526206059, + "n": 5, + "ci95": 0.0026464637463145175 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7945347666740418, + "std": 0.006317691875420071, + "n": 5, + "ci95": 0.00784319297205613 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7627386212348938, + "std": 0.0012657493538665529, + "n": 5, + "ci95": 0.001571383447055281 + }, + "credit_card_default_test_score": { + "mean": 0.8155137658119201, + "std": 0.002131726526206059, + "n": 5, + "ci95": 0.0026464637463145175 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..8b95d78a317933b2d33e6d38ab3e1c7d29a6d586 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:20b1b8bb30bc042f482968ad5d9876785c0ac34d58d50248b062cf2b3a278f30 +size 176640128 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..50149bc8befb058085000638e88b3b2e51770ba3 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:036516bedd2946384d24e537fc0985d86b4d98efcbb4ea877918b3bc6525e66f +size 15531742 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0dc643101f795937f992439165f85727046b670e --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.0025_sx0.002_sy0.0002/seed_42/wjepa_deterministic/credit_card_default/20260916_012036/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: credit_card_default +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.001 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.3 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.1 +mask_min_trgt_share: 0.1 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 256 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.1 +recon_y_weight: 1.0 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.002 +sw_s_y_weight: 0.0002 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 15 +vtjepa_kl_x_end_weight: 0.0001 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 15 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 15 +wjepa_joint_end_weight: 0.0025 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..6c8b9738ebf5fd946c7062b2fcef3be20a0a0c13 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-16T01:01:47.402400+00:00", + "spec": { + "dataset": "credit_card_default", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.004, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0004, + "lambda_joint": 0.005, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..bd7f8a442517539ef7407af4e0975bf35dd2de0f --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "credit_card_default", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.004, + "sw_z_weight": 0.0, + "sw_s_y_weight": 0.0004, + "lambda_joint": 0.005, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "credit_card_default", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.005", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.004", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "0.0004", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..7d986c8ec24632ddaeb14f958d0f16504e326a41 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "credit_card_default", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.005, + "s_x": 0.004, + "z": 0.0, + "s_y": 0.0004 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 15 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "credit_card_default", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "256", + "--model_num_layers", + "8", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.1", + "--mask_max_ctx_share", + "0.3", + "--mask_min_trgt_share", + "0.1", + "--mask_max_trgt_share", + "0.6", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.001", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "0.0001", + "--vtjepa_kl_x_annealing_epochs", + "15", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "15", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.1", + "--recon_y_weight", + "1.0", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.005", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "15", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.004", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "0.0004", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/downstream_results/autoint_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/downstream_results/autoint_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..ae59c67b640a6044c5b56f7c27040eef311c12cd --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/downstream_results/autoint_credit_card_default_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "credit_card_default", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpb0wgezs0", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpb0wgezs0/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.447949081659317, + "lr": 8.856372005539015e-06, + "credit_card_default_val_accuracy": 0.8160297870635986, + "credit_card_default_val_precision": 0.7981109023094177, + "credit_card_default_val_recall": 0.8160297870635986, + "credit_card_default_val_f1-score": 0.7927311062812805, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7546127438545227, + "credit_card_default_val_score": 0.8160297870635986 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4435780644416809, + "lr": 8.856375643517822e-06, + "credit_card_default_test_accuracy": 0.8171591758728027, + "credit_card_default_test_precision": 0.800597071647644, + "credit_card_default_test_recall": 0.8171591758728027, + "credit_card_default_test_f1-score": 0.797120213508606, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7669686675071716, + "credit_card_default_test_score": 0.8171591758728027 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.4483162462711334, + "lr": 3.942649982491275e-06, + "credit_card_default_val_accuracy": 0.8160297870635986, + "credit_card_default_val_precision": 0.7980486750602722, + "credit_card_default_val_recall": 0.8160297870635986, + "credit_card_default_val_f1-score": 0.7935530543327332, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.757779061794281, + "credit_card_default_val_score": 0.8160297870635986 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4445836544036865, + "lr": 3.942650891985977e-06, + "credit_card_default_test_accuracy": 0.8190060257911682, + "credit_card_default_test_precision": 0.803084671497345, + "credit_card_default_test_recall": 0.8190060257911682, + "credit_card_default_test_f1-score": 0.7987513542175293, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7645808458328247, + "credit_card_default_test_score": 0.8190060257911682 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.4505448341369629, + "lr": 2.447174301778432e-05, + "credit_card_default_val_accuracy": 0.8173824548721313, + "credit_card_default_val_precision": 0.8001925349235535, + "credit_card_default_val_recall": 0.8173824548721313, + "credit_card_default_val_f1-score": 0.7932130098342896, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.748562753200531, + "credit_card_default_val_score": 0.8173824548721313 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4437708258628845, + "lr": 2.4471739379805513e-05, + "credit_card_default_test_accuracy": 0.8144727945327759, + "credit_card_default_test_precision": 0.7970197796821594, + "credit_card_default_test_recall": 0.8144727945327759, + "credit_card_default_test_f1-score": 0.792693018913269, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7646008729934692, + "credit_card_default_test_score": 0.8144727945327759 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.4482191801071167, + "lr": 4.758646900882013e-05, + "credit_card_default_val_accuracy": 0.8136624693870544, + "credit_card_default_val_precision": 0.7948768734931946, + "credit_card_default_val_recall": 0.8136624693870544, + "credit_card_default_val_f1-score": 0.7926135063171387, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.754078209400177, + "credit_card_default_val_score": 0.8136624693870544 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4441959857940674, + "lr": 4.7586476284777746e-05, + "credit_card_default_test_accuracy": 0.8164875507354736, + "credit_card_default_test_precision": 0.7998494505882263, + "credit_card_default_test_recall": 0.8164875507354736, + "credit_card_default_test_f1-score": 0.7983850836753845, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7642012238502502, + "credit_card_default_test_score": 0.8164875507354736 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.44996148347854614, + "lr": 0.00015772644837852567, + "credit_card_default_val_accuracy": 0.8146770596504211, + "credit_card_default_val_precision": 0.796315610408783, + "credit_card_default_val_recall": 0.8146770596504211, + "credit_card_default_val_f1-score": 0.7901495099067688, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7526938319206238, + "credit_card_default_val_score": 0.8146770596504211 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4443097412586212, + "lr": 0.00015772644837852567, + "credit_card_default_test_accuracy": 0.8163197040557861, + "credit_card_default_test_precision": 0.7995755672454834, + "credit_card_default_test_recall": 0.8163197040557861, + "credit_card_default_test_f1-score": 0.794415295124054, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7652780413627625, + "credit_card_default_test_score": 0.8163197040557861 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.4440876543521881, + "std": 0.0004083921261851868, + "n": 5, + "ci95": 0.0005070045068834161 + }, + "lr": { + "mean": 4.851673811572255e-05, + "std": 6.337503800832335e-05, + "n": 5, + "ci95": 7.867788782871299e-05 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8166890501976013, + "std": 0.0016338759002178386, + "n": 5, + "ci95": 0.0020283996482415283 + }, + "credit_card_default_test_precision": { + "mean": 0.8000253081321717, + "std": 0.0021767528409619963, + "n": 5, + "ci95": 0.002702362337511302 + }, + "credit_card_default_test_recall": { + "mean": 0.8166890501976013, + "std": 0.0016338759002178386, + "n": 5, + "ci95": 0.0020283996482415283 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7962729930877686, + "std": 0.0026262254185313964, + "n": 5, + "ci95": 0.003260366784553106 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7651259303092957, + "std": 0.001100761379661293, + "n": 5, + "ci95": 0.0013665566613750393 + }, + "credit_card_default_test_score": { + "mean": 0.8166890501976013, + "std": 0.0016338759002178386, + "n": 5, + "ci95": 0.0020283996482415283 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/downstream_results/dcnv2_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/downstream_results/dcnv2_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..cf5e0eee248ef0cc175b63f067abda0e9fcdb7c4 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/downstream_results/dcnv2_credit_card_default_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "credit_card_default", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp28mvql24", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp28mvql24/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.453788161277771, + "lr": 1.5708415958215483e-05, + "credit_card_default_val_accuracy": 0.8126479387283325, + "credit_card_default_val_precision": 0.793379545211792, + "credit_card_default_val_recall": 0.8126479387283325, + "credit_card_default_val_f1-score": 0.7882826328277588, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7470809817314148, + "credit_card_default_val_score": 0.8126479387283325 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4495033025741577, + "lr": 1.57084250531625e-05, + "credit_card_default_test_accuracy": 0.8132975101470947, + "credit_card_default_test_precision": 0.7954416871070862, + "credit_card_default_test_recall": 0.8132975101470947, + "credit_card_default_test_f1-score": 0.7925962805747986, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7574616074562073, + "credit_card_default_test_score": 0.8132975101470947 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.44916775822639465, + "lr": 0.00011474337225081399, + "credit_card_default_val_accuracy": 0.8153533935546875, + "credit_card_default_val_precision": 0.7971200346946716, + "credit_card_default_val_recall": 0.8153533935546875, + "credit_card_default_val_f1-score": 0.7925893068313599, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7473573088645935, + "credit_card_default_val_score": 0.8153533935546875 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44524434208869934, + "lr": 0.00011474342318251729, + "credit_card_default_test_accuracy": 0.8158159852027893, + "credit_card_default_test_precision": 0.7988305687904358, + "credit_card_default_test_recall": 0.8158159852027893, + "credit_card_default_test_f1-score": 0.7961902618408203, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.761531412601471, + "credit_card_default_test_score": 0.8158159852027893 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.45480045676231384, + "lr": 3.942649982491275e-06, + "credit_card_default_val_accuracy": 0.8106188774108887, + "credit_card_default_val_precision": 0.790574848651886, + "credit_card_default_val_recall": 0.8106188774108887, + "credit_card_default_val_f1-score": 0.7876900434494019, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7467442154884338, + "credit_card_default_val_score": 0.8106188774108887 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4513683319091797, + "lr": 3.942650891985977e-06, + "credit_card_default_test_accuracy": 0.8126259446144104, + "credit_card_default_test_precision": 0.7946200370788574, + "credit_card_default_test_recall": 0.8126259446144104, + "credit_card_default_test_f1-score": 0.7926130890846252, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.755546510219574, + "credit_card_default_test_score": 0.8126259446144104 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.45112428069114685, + "lr": 3.511174872983247e-05, + "credit_card_default_val_accuracy": 0.8153533935546875, + "credit_card_default_val_precision": 0.7972261309623718, + "credit_card_default_val_recall": 0.8153533935546875, + "credit_card_default_val_f1-score": 0.7913399338722229, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7457246780395508, + "credit_card_default_val_score": 0.8153533935546875 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4458002746105194, + "lr": 3.51117632817477e-05, + "credit_card_default_test_accuracy": 0.8153122663497925, + "credit_card_default_test_precision": 0.7981326580047607, + "credit_card_default_test_recall": 0.8153122663497925, + "credit_card_default_test_f1-score": 0.7945488095283508, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7595342397689819, + "credit_card_default_test_score": 0.8153122663497925 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.4490184485912323, + "lr": 0.00013551565643865615, + "credit_card_default_val_accuracy": 0.8150151968002319, + "credit_card_default_val_precision": 0.796886146068573, + "credit_card_default_val_recall": 0.8150151968002319, + "credit_card_default_val_f1-score": 0.7899949550628662, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7498002648353577, + "credit_card_default_val_score": 0.8150151968002319 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44480782747268677, + "lr": 0.00013551570009440184, + "credit_card_default_test_accuracy": 0.815144419670105, + "credit_card_default_test_precision": 0.7979660630226135, + "credit_card_default_test_recall": 0.815144419670105, + "credit_card_default_test_f1-score": 0.7930505275726318, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7621772289276123, + "credit_card_default_test_score": 0.815144419670105 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.44734481573104856, + "std": 0.0029189747389208884, + "n": 5, + "ci95": 0.0036238048023498194 + }, + "lr": { + "mean": 6.100439250076306e-05, + "std": 6.0037516704646254e-05, + "n": 5, + "ci95": 7.453447213999721e-05 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8144392251968384, + "std": 0.0013916238661438555, + "n": 5, + "ci95": 0.001727652241026606 + }, + "credit_card_default_test_precision": { + "mean": 0.7969982028007507, + "std": 0.001847952951547777, + "n": 5, + "ci95": 0.0022941688021633996 + }, + "credit_card_default_test_recall": { + "mean": 0.8144392251968384, + "std": 0.0013916238661438555, + "n": 5, + "ci95": 0.001727652241026606 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7937997937202453, + "std": 0.001556717223816043, + "n": 5, + "ci95": 0.001932609856586411 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7592501997947693, + "std": 0.0027734675563225587, + "n": 5, + "ci95": 0.00344316273647461 + }, + "credit_card_default_test_score": { + "mean": 0.8144392251968384, + "std": 0.0013916238661438555, + "n": 5, + "ci95": 0.001727652241026606 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/downstream_results/fttransformer_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/downstream_results/fttransformer_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..b21131ed7ffedd894d218aced36bf6ed814fed7b --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/downstream_results/fttransformer_credit_card_default_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "credit_card_default", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpgaw4vyek", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpgaw4vyek/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.45644640922546387, + "lr": 0.00025912324781529605, + "credit_card_default_val_accuracy": 0.8153533935546875, + "credit_card_default_val_precision": 0.7972539663314819, + "credit_card_default_val_recall": 0.8153533935546875, + "credit_card_default_val_f1-score": 0.7911281585693359, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7270500659942627, + "credit_card_default_val_score": 0.8153533935546875 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44839224219322205, + "lr": 0.0002591230731923133, + "credit_card_default_test_accuracy": 0.8169912695884705, + "credit_card_default_test_precision": 0.8003775477409363, + "credit_card_default_test_recall": 0.8169912695884705, + "credit_card_default_test_f1-score": 0.796699583530426, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7500472664833069, + "credit_card_default_test_score": 0.8169912695884705 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.4450884163379669, + "lr": 3.942649982491275e-06, + "credit_card_default_val_accuracy": 0.8156915903091431, + "credit_card_default_val_precision": 0.7977214455604553, + "credit_card_default_val_recall": 0.8156915903091431, + "credit_card_default_val_f1-score": 0.7916164994239807, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7552632093429565, + "credit_card_default_val_score": 0.8156915903091431 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44133660197257996, + "lr": 3.942650891985977e-06, + "credit_card_default_test_accuracy": 0.8191739320755005, + "credit_card_default_test_precision": 0.8034356832504272, + "credit_card_default_test_recall": 0.8191739320755005, + "credit_card_default_test_f1-score": 0.797946035861969, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7641297578811646, + "credit_card_default_test_score": 0.8191739320755005 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.44740623235702515, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8129861354827881, + "credit_card_default_val_precision": 0.7938436269760132, + "credit_card_default_val_recall": 0.8129861354827881, + "credit_card_default_val_f1-score": 0.7889845967292786, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7539299726486206, + "credit_card_default_val_score": 0.8129861354827881 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4400595426559448, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8179986476898193, + "credit_card_default_test_precision": 0.8017470240592957, + "credit_card_default_test_recall": 0.8179986476898193, + "credit_card_default_test_f1-score": 0.7974429726600647, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.766780436038971, + "credit_card_default_test_score": 0.8179986476898193 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.4472641050815582, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8160297870635986, + "credit_card_default_val_precision": 0.7981335520744324, + "credit_card_default_val_recall": 0.8160297870635986, + "credit_card_default_val_f1-score": 0.7925231456756592, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.751899778842926, + "credit_card_default_val_score": 0.8160297870635986 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4406627118587494, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8181665539741516, + "credit_card_default_test_precision": 0.80197674036026, + "credit_card_default_test_recall": 0.8181665539741516, + "credit_card_default_test_f1-score": 0.7975826263427734, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7647634744644165, + "credit_card_default_test_score": 0.8181665539741516 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.45124173164367676, + "lr": 0.0, + "credit_card_default_val_accuracy": 0.8200879096984863, + "credit_card_default_val_precision": 0.8048437833786011, + "credit_card_default_val_recall": 0.8200879096984863, + "credit_card_default_val_f1-score": 0.7941468954086304, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7391803860664368, + "credit_card_default_val_score": 0.8200879096984863 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4462711811065674, + "lr": 0.0, + "credit_card_default_test_accuracy": 0.8176628351211548, + "credit_card_default_test_precision": 0.8018006086349487, + "credit_card_default_test_recall": 0.8176628351211548, + "credit_card_default_test_f1-score": 0.794130802154541, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7518731355667114, + "credit_card_default_test_score": 0.8176628351211548 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.44334445595741273, + "std": 0.0037436529758514475, + "n": 5, + "ci95": 0.004647613921193768 + }, + "lr": { + "mean": 5.261314481685986e-05, + "std": 0.00011545518234263643, + "n": 5, + "ci95": 0.00014333356114760106 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8179986476898193, + "std": 0.0007964085992589712, + "n": 5, + "ci95": 0.0009887133547768503 + }, + "credit_card_default_test_precision": { + "mean": 0.8018675208091736, + "std": 0.001085135798225422, + "n": 5, + "ci95": 0.0013471580498379792 + }, + "credit_card_default_test_recall": { + "mean": 0.8179986476898193, + "std": 0.0007964085992589712, + "n": 5, + "ci95": 0.0009887133547768503 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7967604041099549, + "std": 0.001538357684670153, + "n": 5, + "ci95": 0.0019098171324018907 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.759518814086914, + "std": 0.007900396074966122, + "n": 5, + "ci95": 0.009808064747936768 + }, + "credit_card_default_test_score": { + "mean": 0.8179986476898193, + "std": 0.0007964085992589712, + "n": 5, + "ci95": 0.0009887133547768503 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/downstream_results/mlp_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/downstream_results/mlp_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..a049b0cc137ae814eb251d159b0b9b10acb9e1bd --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/downstream_results/mlp_credit_card_default_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "credit_card_default", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp_n1zgkyb", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp_n1zgkyb/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.46065470576286316, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8099424839019775, + "credit_card_default_val_precision": 0.7959985136985779, + "credit_card_default_val_recall": 0.8099424839019775, + "credit_card_default_val_f1-score": 0.7703313827514648, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.732129693031311, + "credit_card_default_val_score": 0.8099424839019775 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.45430588722229004, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8106111288070679, + "credit_card_default_test_precision": 0.796504020690918, + "credit_card_default_test_recall": 0.8106111288070679, + "credit_card_default_test_f1-score": 0.7753177881240845, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7501654624938965, + "credit_card_default_test_score": 0.8106111288070679 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.46262994408607483, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8102806806564331, + "credit_card_default_val_precision": 0.7900192737579346, + "credit_card_default_val_recall": 0.8102806806564331, + "credit_card_default_val_f1-score": 0.7859318852424622, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7327893376350403, + "credit_card_default_val_score": 0.8102806806564331 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4545239210128784, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8121222257614136, + "credit_card_default_test_precision": 0.79383385181427, + "credit_card_default_test_recall": 0.8121222257614136, + "credit_card_default_test_f1-score": 0.7907564640045166, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7550603747367859, + "credit_card_default_test_score": 0.8121222257614136 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.45644208788871765, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8170443177223206, + "credit_card_default_val_precision": 0.802611768245697, + "credit_card_default_val_recall": 0.8170443177223206, + "credit_card_default_val_f1-score": 0.7861231565475464, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7383232712745667, + "credit_card_default_val_score": 0.8170443177223206 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4519837498664856, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8136333227157593, + "credit_card_default_test_precision": 0.7979508638381958, + "credit_card_default_test_recall": 0.8136333227157593, + "credit_card_default_test_f1-score": 0.7841911911964417, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7496302127838135, + "credit_card_default_test_score": 0.8136333227157593 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.46032851934432983, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8126479387283325, + "credit_card_default_val_precision": 0.7944825887680054, + "credit_card_default_val_recall": 0.8126479387283325, + "credit_card_default_val_f1-score": 0.7830660939216614, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7413825988769531, + "credit_card_default_val_score": 0.8126479387283325 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.45080238580703735, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8144727945327759, + "credit_card_default_test_precision": 0.7980764508247375, + "credit_card_default_test_recall": 0.8144727945327759, + "credit_card_default_test_f1-score": 0.7877055406570435, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7600964903831482, + "credit_card_default_test_score": 0.8144727945327759 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.4584132432937622, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8092661499977112, + "credit_card_default_val_precision": 0.7897169589996338, + "credit_card_default_val_recall": 0.8092661499977112, + "credit_card_default_val_f1-score": 0.7774110436439514, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7484244108200073, + "credit_card_default_val_score": 0.8092661499977112 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4491511285305023, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8141369819641113, + "credit_card_default_test_precision": 0.7974941730499268, + "credit_card_default_test_recall": 0.8141369819641113, + "credit_card_default_test_f1-score": 0.7875410914421082, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7649203538894653, + "credit_card_default_test_score": 0.8141369819641113 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.45215341448783875, + "std": 0.0022978373314661684, + "n": 5, + "ci95": 0.002852684487384142 + }, + "lr": { + "mean": 0.0009999994654208422, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8129952907562256, + "std": 0.0016077959828790466, + "n": 5, + "ci95": 0.0019960223451984263 + }, + "credit_card_default_test_precision": { + "mean": 0.7967718720436097, + "std": 0.0017548057874551954, + "n": 5, + "ci95": 0.0021785298635788367 + }, + "credit_card_default_test_recall": { + "mean": 0.8129952907562256, + "std": 0.0016077959828790466, + "n": 5, + "ci95": 0.0019960223451984263 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7851024150848389, + "std": 0.00594266203065146, + "n": 5, + "ci95": 0.007377606567906768 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7559745788574219, + "std": 0.006554571468161879, + "n": 5, + "ci95": 0.008137270681708999 + }, + "credit_card_default_test_score": { + "mean": 0.8129952907562256, + "std": 0.0016077959828790466, + "n": 5, + "ci95": 0.0019960223451984263 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/downstream_results/resnet_credit_card_default_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/downstream_results/resnet_credit_card_default_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..86ecd728ccb99cf29336904e1c33c61c2413f14a --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/downstream_results/resnet_credit_card_default_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "credit_card_default", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp6h4s53i_", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "credit_card_default", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp6h4s53i_/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "credit_card_default_val_loss": 0.45229673385620117, + "lr": 0.0005000000237487257, + "credit_card_default_val_accuracy": 0.8116334080696106, + "credit_card_default_val_precision": 0.7920339107513428, + "credit_card_default_val_recall": 0.8116334080696106, + "credit_card_default_val_f1-score": 0.7857130765914917, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7508911490440369, + "credit_card_default_val_score": 0.8116334080696106 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4470994174480438, + "lr": 0.0004999997327104211, + "credit_card_default_test_accuracy": 0.8176628351211548, + "credit_card_default_test_precision": 0.8016462922096252, + "credit_card_default_test_recall": 0.8176628351211548, + "credit_card_default_test_f1-score": 0.7947351932525635, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7592672109603882, + "credit_card_default_test_score": 0.8176628351211548 + } + } + }, + { + "seed": 1, + "val_metrics": { + "credit_card_default_val_loss": 0.45245808362960815, + "lr": 0.0005000000237487257, + "credit_card_default_val_accuracy": 0.8068988919258118, + "credit_card_default_val_precision": 0.7868729829788208, + "credit_card_default_val_recall": 0.8068988919258118, + "credit_card_default_val_f1-score": 0.7882464528083801, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7518185377120972, + "credit_card_default_val_score": 0.8068988919258118 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4432048797607422, + "lr": 0.0004999997327104211, + "credit_card_default_test_accuracy": 0.8181665539741516, + "credit_card_default_test_precision": 0.8023433685302734, + "credit_card_default_test_recall": 0.8181665539741516, + "credit_card_default_test_f1-score": 0.8021225929260254, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7669070363044739, + "credit_card_default_test_score": 0.8181665539741516 + } + } + }, + { + "seed": 2, + "val_metrics": { + "credit_card_default_val_loss": 0.45458704233169556, + "lr": 0.0005000000237487257, + "credit_card_default_val_accuracy": 0.8156915903091431, + "credit_card_default_val_precision": 0.7977214455604553, + "credit_card_default_val_recall": 0.8156915903091431, + "credit_card_default_val_f1-score": 0.7916164994239807, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7489187121391296, + "credit_card_default_val_score": 0.8156915903091431 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4485466182231903, + "lr": 0.0004999997327104211, + "credit_card_default_test_accuracy": 0.8181665539741516, + "credit_card_default_test_precision": 0.8020133972167969, + "credit_card_default_test_recall": 0.8181665539741516, + "credit_card_default_test_f1-score": 0.7972033023834229, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7620669007301331, + "credit_card_default_test_score": 0.8181665539741516 + } + } + }, + { + "seed": 3, + "val_metrics": { + "credit_card_default_val_loss": 0.453709214925766, + "lr": 0.0005000000237487257, + "credit_card_default_val_accuracy": 0.811295211315155, + "credit_card_default_val_precision": 0.7920047640800476, + "credit_card_default_val_recall": 0.811295211315155, + "credit_card_default_val_f1-score": 0.7914557456970215, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7520936131477356, + "credit_card_default_val_score": 0.811295211315155 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.44556868076324463, + "lr": 0.0004999997327104211, + "credit_card_default_test_accuracy": 0.8216924071311951, + "credit_card_default_test_precision": 0.8065870404243469, + "credit_card_default_test_recall": 0.8216924071311951, + "credit_card_default_test_f1-score": 0.8048316836357117, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7632036209106445, + "credit_card_default_test_score": 0.8216924071311951 + } + } + }, + { + "seed": 4, + "val_metrics": { + "credit_card_default_val_loss": 0.4477183520793915, + "lr": 0.0010000000474974513, + "credit_card_default_val_accuracy": 0.8119716048240662, + "credit_card_default_val_precision": 0.7924178838729858, + "credit_card_default_val_recall": 0.8119716048240662, + "credit_card_default_val_f1-score": 0.787518322467804, + "credit_card_default_val_support": 2957.0, + "credit_card_default_val_roc_auc": 0.7552493214607239, + "credit_card_default_val_score": 0.8119716048240662 + }, + "test_metrics": { + "0": { + "credit_card_default_test_loss": 0.4430689215660095, + "lr": 0.0009999994654208422, + "credit_card_default_test_accuracy": 0.8191739320755005, + "credit_card_default_test_precision": 0.8034058213233948, + "credit_card_default_test_recall": 0.8191739320755005, + "credit_card_default_test_f1-score": 0.7981368899345398, + "credit_card_default_test_support": 5956.0, + "credit_card_default_test_roc_auc": 0.7638898491859436, + "credit_card_default_test_score": 0.8191739320755005 + } + } + } + ], + "summary": { + "0": { + "credit_card_default_test_loss": { + "mean": 0.4454977035522461, + "std": 0.0023990846543382512, + "n": 5, + "ci95": 0.002978379489110863 + }, + "lr": { + "mean": 0.0005999996792525053, + "std": 0.00022360667821444535, + "n": 5, + "ci95": 0.0002775998516008258 + }, + "credit_card_default_test_accuracy": { + "mean": 0.8189724564552308, + "std": 0.0016165387222340023, + "n": 5, + "ci95": 0.0020068761495968482 + }, + "credit_card_default_test_precision": { + "mean": 0.8031991839408874, + "std": 0.002004386032022293, + "n": 5, + "ci95": 0.0024883749872020196 + }, + "credit_card_default_test_recall": { + "mean": 0.8189724564552308, + "std": 0.0016165387222340023, + "n": 5, + "ci95": 0.0020068761495968482 + }, + "credit_card_default_test_f1-score": { + "mean": 0.7994059324264526, + "std": 0.0040341316549523795, + "n": 5, + "ci95": 0.005008233017436903 + }, + "credit_card_default_test_support": { + "mean": 5956.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "credit_card_default_test_roc_auc": { + "mean": 0.7630669236183166, + "std": 0.0027784987017129615, + "n": 5, + "ci95": 0.003449408727090411 + }, + "credit_card_default_test_score": { + "mean": 0.8189724564552308, + "std": 0.0016165387222340023, + "n": 5, + "ci95": 0.0020068761495968482 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..1cc458e9bfe1ed9b176b05de5bbdebba462877d3 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ed1aa4cfc1e45b3f96f992f66ee8fe42b9ef89f7c2b99f20b6c0149f031886fa +size 176640128 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..72f586f65f8f758c2025a16c6efea0f349fcc4a6 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7dac5b43def918cc1ebd9fb087d798763ee3f27c92430abf2422a3e10ebb394d +size 15531358 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..79ec19ae39ca63c08e1a1541068d79be151108ce --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/credit_card_default/deterministic_sw/sw_f0.005_sx0.004_sy0.0004/seed_42/wjepa_deterministic/credit_card_default/20260916_001729/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: credit_card_default +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.001 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.3 +mask_max_trgt_share: 0.6 +mask_min_ctx_share: 0.1 +mask_min_trgt_share: 0.1 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 256 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 8 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.1 +recon_y_weight: 1.0 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.004 +sw_s_y_weight: 0.0004 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 15 +vtjepa_kl_x_end_weight: 0.0001 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 15 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 15 +wjepa_joint_end_weight: 0.005 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/deterministic_shard00of01_protocol.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/deterministic_shard00of01_protocol.json new file mode 100644 index 0000000000000000000000000000000000000000..8a627ec846aebd15de9efe4f6a720c74764230f3 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/deterministic_shard00of01_protocol.json @@ -0,0 +1,50 @@ +{ + "created_at_utc": "2026-09-16T05:50:51.368416+00:00", + "stage": "deterministic", + "execute": true, + "batch_size": 512, + "shard_index": 0, + "num_shards": 1, + "total_specifications": 1, + "source_sha256": { + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5" + }, + "specifications": [ + { + "dataset": "sim", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 5e-05, + "sw_z_weight": 0.0, + "sw_s_y_weight": 1e-05, + "lambda_joint": 0.00025, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..017e98deb41f2fdd65434417760b13995056a7cc --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-16T00:05:57.830161+00:00", + "spec": { + "dataset": "electricity", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 2.5e-05, + "sw_z_weight": 0.0, + "sw_s_y_weight": 5e-06, + "lambda_joint": 0.000125, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..058c6eae3f6ece0e449eac3090edee26bd1d1b0e --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "electricity", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 2.5e-05, + "sw_z_weight": 0.0, + "sw_s_y_weight": 5e-06, + "lambda_joint": 0.000125, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "electricity", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.000125", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "2.5e-05", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "5e-06", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..9c41b9f7098bd37ebe4f3d32ab26ac6f41a9a9f6 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "electricity", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.000125, + "s_x": 2.5e-05, + "z": 0.0, + "s_y": 5e-06 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 40 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "electricity", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "64", + "--model_num_layers", + "6", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.15", + "--mask_max_ctx_share", + "0.6", + "--mask_min_trgt_share", + "0.15", + "--mask_max_trgt_share", + "0.9", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "128", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.0005", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "1e-06", + "--vtjepa_kl_x_annealing_epochs", + "40", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-05", + "--vtjepa_kl_z_annealing_epochs", + "40", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-06", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.001", + "--recon_y_weight", + "0.25", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.000125", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "40", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "2.5e-05", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "5e-06", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/downstream_results/autoint_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/downstream_results/autoint_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..8736ac2a1934ce01e8831a756e7ca2c1f30610d2 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/downstream_results/autoint_electricity_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "electricity", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpff0mpj6n", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpff0mpj6n/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.38456568121910095, + "lr": 0.0, + "electricity_val_accuracy": 0.8237150311470032, + "electricity_val_precision": 0.8231273293495178, + "electricity_val_recall": 0.8237150311470032, + "electricity_val_f1-score": 0.8230891823768616, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.904629647731781, + "electricity_val_score": 0.8237150311470032 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3919948637485504, + "lr": 0.0, + "electricity_test_accuracy": 0.8189569115638733, + "electricity_test_precision": 0.8183704018592834, + "electricity_test_recall": 0.8189569115638733, + "electricity_test_f1-score": 0.8184359073638916, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.901023805141449, + "electricity_test_score": 0.8189569115638733 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.38753265142440796, + "lr": 0.0, + "electricity_val_accuracy": 0.8239405155181885, + "electricity_val_precision": 0.8233599662780762, + "electricity_val_recall": 0.8239405155181885, + "electricity_val_f1-score": 0.823354959487915, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9029892683029175, + "electricity_val_score": 0.8239405155181885 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3893147110939026, + "lr": 0.0, + "electricity_test_accuracy": 0.8211715221405029, + "electricity_test_precision": 0.8207123279571533, + "electricity_test_recall": 0.8211715221405029, + "electricity_test_f1-score": 0.820841372013092, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9016156792640686, + "electricity_test_score": 0.8211715221405029 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.39341384172439575, + "lr": 0.0, + "electricity_val_accuracy": 0.8214607834815979, + "electricity_val_precision": 0.8210547566413879, + "electricity_val_recall": 0.8214607834815979, + "electricity_val_f1-score": 0.8202871680259705, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8996707201004028, + "electricity_val_score": 0.8214607834815979 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3975147604942322, + "lr": 0.0, + "electricity_test_accuracy": 0.8157457709312439, + "electricity_test_precision": 0.8152078986167908, + "electricity_test_recall": 0.8157457709312439, + "electricity_test_f1-score": 0.8146214485168457, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8968580961227417, + "electricity_test_score": 0.8157457709312439 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.3880448341369629, + "lr": 0.0, + "electricity_val_accuracy": 0.8270964622497559, + "electricity_val_precision": 0.8266586065292358, + "electricity_val_recall": 0.8270964622497559, + "electricity_val_f1-score": 0.8261107802391052, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9032466411590576, + "electricity_val_score": 0.8270964622497559 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3877507448196411, + "lr": 0.0, + "electricity_test_accuracy": 0.8212822675704956, + "electricity_test_precision": 0.8206790685653687, + "electricity_test_recall": 0.8212822675704956, + "electricity_test_f1-score": 0.8205593228340149, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9028655886650085, + "electricity_test_score": 0.8212822675704956 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.39925751090049744, + "lr": 0.0, + "electricity_val_accuracy": 0.814472496509552, + "electricity_val_precision": 0.8138003349304199, + "electricity_val_recall": 0.814472496509552, + "electricity_val_f1-score": 0.8137373924255371, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8962525725364685, + "electricity_val_score": 0.814472496509552 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.40180373191833496, + "lr": 0.0, + "electricity_test_accuracy": 0.8161886930465698, + "electricity_test_precision": 0.8155398368835449, + "electricity_test_recall": 0.8161886930465698, + "electricity_test_f1-score": 0.8154784440994263, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8950114250183105, + "electricity_test_score": 0.8161886930465698 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.39367576241493224, + "std": 0.005868385999508754, + "n": 5, + "ci95": 0.00728539547927846 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.8186690330505371, + "std": 0.0026396907154776863, + "n": 5, + "ci95": 0.0032770834786335318 + }, + "electricity_test_precision": { + "mean": 0.8181019067764282, + "std": 0.0026677535774850664, + "n": 5, + "ci95": 0.003311922537962842 + }, + "electricity_test_recall": { + "mean": 0.8186690330505371, + "std": 0.0026396907154776863, + "n": 5, + "ci95": 0.0032770834786335318 + }, + "electricity_test_f1-score": { + "mean": 0.8179872989654541, + "std": 0.002854189517973416, + "n": 5, + "ci95": 0.0035433762218416037 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8994749188423157, + "std": 0.0033633719709585046, + "n": 5, + "ci95": 0.004175508385849905 + }, + "electricity_test_score": { + "mean": 0.8186690330505371, + "std": 0.0026396907154776863, + "n": 5, + "ci95": 0.0032770834786335318 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/downstream_results/dcnv2_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/downstream_results/dcnv2_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..e11ef14809ef84cd10b3239dfd6298674c1c65e1 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/downstream_results/dcnv2_electricity_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "electricity", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpctsjogcz", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpctsjogcz/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.37542182207107544, + "lr": 0.0, + "electricity_val_accuracy": 0.829801619052887, + "electricity_val_precision": 0.8292719125747681, + "electricity_val_recall": 0.829801619052887, + "electricity_val_f1-score": 0.8292809128761292, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9105371236801147, + "electricity_val_score": 0.829801619052887 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3762483596801758, + "lr": 0.0, + "electricity_test_accuracy": 0.8304728269577026, + "electricity_test_precision": 0.8300725817680359, + "electricity_test_recall": 0.8304728269577026, + "electricity_test_f1-score": 0.8301867842674255, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9109489321708679, + "electricity_test_score": 0.8304728269577026 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.38015156984329224, + "lr": 0.0, + "electricity_val_accuracy": 0.8221370577812195, + "electricity_val_precision": 0.821533739566803, + "electricity_val_recall": 0.8221370577812195, + "electricity_val_f1-score": 0.82143235206604, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9072757959365845, + "electricity_val_score": 0.8221370577812195 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.37906211614608765, + "lr": 0.0, + "electricity_test_accuracy": 0.8292548060417175, + "electricity_test_precision": 0.8287531733512878, + "electricity_test_recall": 0.8292548060417175, + "electricity_test_f1-score": 0.8288174867630005, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9089200496673584, + "electricity_test_score": 0.8292548060417175 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.37066492438316345, + "lr": 0.0, + "electricity_val_accuracy": 0.8307033181190491, + "electricity_val_precision": 0.8301798105239868, + "electricity_val_recall": 0.8307033181190491, + "electricity_val_f1-score": 0.8300168514251709, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9136984944343567, + "electricity_val_score": 0.8307033181190491 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3772537112236023, + "lr": 0.0, + "electricity_test_accuracy": 0.833683967590332, + "electricity_test_precision": 0.8332021832466125, + "electricity_test_recall": 0.833683967590332, + "electricity_test_f1-score": 0.8332440853118896, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9114270210266113, + "electricity_test_score": 0.833683967590332 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.3764151632785797, + "lr": 0.0, + "electricity_val_accuracy": 0.8257439136505127, + "electricity_val_precision": 0.8252767324447632, + "electricity_val_recall": 0.8257439136505127, + "electricity_val_f1-score": 0.8247680068016052, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9101144671440125, + "electricity_val_score": 0.8257439136505127 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.38134896755218506, + "lr": 0.0, + "electricity_test_accuracy": 0.8271509408950806, + "electricity_test_precision": 0.8266164660453796, + "electricity_test_recall": 0.8271509408950806, + "electricity_test_f1-score": 0.8266534805297852, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9079077839851379, + "electricity_test_score": 0.8271509408950806 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.3728141784667969, + "lr": 0.0, + "electricity_val_accuracy": 0.8275473117828369, + "electricity_val_precision": 0.8269947171211243, + "electricity_val_recall": 0.8275473117828369, + "electricity_val_f1-score": 0.8268480896949768, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9121366739273071, + "electricity_val_score": 0.8275473117828369 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3759726285934448, + "lr": 0.0, + "electricity_test_accuracy": 0.8301406502723694, + "electricity_test_precision": 0.8296355605125427, + "electricity_test_recall": 0.8301406502723694, + "electricity_test_f1-score": 0.8296841979026794, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.91001957654953, + "electricity_test_score": 0.8301406502723694 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.37797715663909914, + "std": 0.0022403398651549954, + "n": 5, + "ci95": 0.002781303398756289 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.8301406383514405, + "std": 0.002365825496891641, + "n": 5, + "ci95": 0.002937089411170109 + }, + "electricity_test_precision": { + "mean": 0.8296559929847718, + "std": 0.002387636717516496, + "n": 5, + "ci95": 0.002964167276898636 + }, + "electricity_test_recall": { + "mean": 0.8301406383514405, + "std": 0.002365825496891641, + "n": 5, + "ci95": 0.002937089411170109 + }, + "electricity_test_f1-score": { + "mean": 0.8297172069549561, + "std": 0.0023904172975982604, + "n": 5, + "ci95": 0.002967619269586089 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.9098446726799011, + "std": 0.0014457067656135309, + "n": 5, + "ci95": 0.00179479426463167 + }, + "electricity_test_score": { + "mean": 0.8301406383514405, + "std": 0.002365825496891641, + "n": 5, + "ci95": 0.002937089411170109 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/downstream_results/fttransformer_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/downstream_results/fttransformer_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..9b4062aa7c43f26aec0ba06e4e7c3b31c5006e8e --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/downstream_results/fttransformer_electricity_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "electricity", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpiju2lm2o", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpiju2lm2o/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.43329522013664246, + "lr": 0.0, + "electricity_val_accuracy": 0.7948602437973022, + "electricity_val_precision": 0.7950323224067688, + "electricity_val_recall": 0.7948602437973022, + "electricity_val_f1-score": 0.7923453450202942, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8740659356117249, + "electricity_val_score": 0.7948602437973022 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4363137483596802, + "lr": 0.0, + "electricity_test_accuracy": 0.7932676076889038, + "electricity_test_precision": 0.7936683893203735, + "electricity_test_recall": 0.7932676076889038, + "electricity_test_f1-score": 0.7905468940734863, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8725687861442566, + "electricity_test_score": 0.7932676076889038 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.42924654483795166, + "lr": 0.0, + "electricity_val_accuracy": 0.797339916229248, + "electricity_val_precision": 0.797608494758606, + "electricity_val_recall": 0.797339916229248, + "electricity_val_f1-score": 0.7948424816131592, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8769302368164062, + "electricity_val_score": 0.797339916229248 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.43700656294822693, + "lr": 0.0, + "electricity_test_accuracy": 0.7932676076889038, + "electricity_test_precision": 0.7941017746925354, + "electricity_test_recall": 0.7932676076889038, + "electricity_test_f1-score": 0.7902488708496094, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8721259236335754, + "electricity_test_score": 0.7932676076889038 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.43134257197380066, + "lr": 0.0, + "electricity_val_accuracy": 0.7995942234992981, + "electricity_val_precision": 0.7988092303276062, + "electricity_val_recall": 0.7995942234992981, + "electricity_val_f1-score": 0.7983703017234802, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8768086433410645, + "electricity_val_score": 0.7995942234992981 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.44039347767829895, + "lr": 0.0, + "electricity_test_accuracy": 0.7931569218635559, + "electricity_test_precision": 0.7922785878181458, + "electricity_test_recall": 0.7931569218635559, + "electricity_test_f1-score": 0.7919566631317139, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8710671067237854, + "electricity_test_score": 0.7931569218635559 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.4341762661933899, + "lr": 0.0, + "electricity_val_accuracy": 0.7944093942642212, + "electricity_val_precision": 0.7937944531440735, + "electricity_val_recall": 0.7944093942642212, + "electricity_val_f1-score": 0.7926554679870605, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8747204542160034, + "electricity_val_score": 0.7944093942642212 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.44078996777534485, + "lr": 0.0, + "electricity_test_accuracy": 0.79271399974823, + "electricity_test_precision": 0.7920024394989014, + "electricity_test_recall": 0.79271399974823, + "electricity_test_f1-score": 0.7910707592964172, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8711503744125366, + "electricity_test_score": 0.79271399974823 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.4213695526123047, + "lr": 0.0, + "electricity_val_accuracy": 0.806582510471344, + "electricity_val_precision": 0.8058724403381348, + "electricity_val_recall": 0.806582510471344, + "electricity_val_f1-score": 0.8054701089859009, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8833061456680298, + "electricity_val_score": 0.806582510471344 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4307301938533783, + "lr": 0.0, + "electricity_test_accuracy": 0.7995792031288147, + "electricity_test_precision": 0.7988213300704956, + "electricity_test_recall": 0.7995792031288147, + "electricity_test_f1-score": 0.7983158230781555, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8768301010131836, + "electricity_test_score": 0.7995792031288147 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.43704679012298586, + "std": 0.004051236027916876, + "n": 5, + "ci95": 0.005029467496811959 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.7943970680236816, + "std": 0.0029058805404658205, + "n": 5, + "ci95": 0.003607548814035943 + }, + "electricity_test_precision": { + "mean": 0.7941745042800903, + "std": 0.002746215419267239, + "n": 5, + "ci95": 0.0034093301637501635 + }, + "electricity_test_recall": { + "mean": 0.7943970680236816, + "std": 0.0029058805404658205, + "n": 5, + "ci95": 0.003607548814035943 + }, + "electricity_test_f1-score": { + "mean": 0.7924278020858765, + "std": 0.0033547684173250076, + "n": 5, + "ci95": 0.004164827375644976 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8727484583854676, + "std": 0.0023696230695378485, + "n": 5, + "ci95": 0.0029418039644716868 + }, + "electricity_test_score": { + "mean": 0.7943970680236816, + "std": 0.0029058805404658205, + "n": 5, + "ci95": 0.003607548814035943 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/downstream_results/mlp_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/downstream_results/mlp_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..373798b8ba06c155f2dd9e356525cd2ec6c7edf2 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/downstream_results/mlp_electricity_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "electricity", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp913r6l6q", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp913r6l6q/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.44889727234840393, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7917042374610901, + "electricity_val_precision": 0.7907853126525879, + "electricity_val_recall": 0.7917042374610901, + "electricity_val_f1-score": 0.7906306982040405, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8670802116394043, + "electricity_val_score": 0.7917042374610901 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4536339342594147, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7897242903709412, + "electricity_test_precision": 0.7888091802597046, + "electricity_test_recall": 0.7897242903709412, + "electricity_test_f1-score": 0.7887915968894958, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8646141886711121, + "electricity_test_score": 0.7897242903709412 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.4482845962047577, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7840396761894226, + "electricity_val_precision": 0.7878449559211731, + "electricity_val_recall": 0.7840396761894226, + "electricity_val_f1-score": 0.7789176106452942, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8684053421020508, + "electricity_val_score": 0.7840396761894226 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.45561301708221436, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7824161052703857, + "electricity_test_precision": 0.7860459685325623, + "electricity_test_recall": 0.7824161052703857, + "electricity_test_f1-score": 0.7773237228393555, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8649308681488037, + "electricity_test_score": 0.7824161052703857 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.45404478907585144, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.783363401889801, + "electricity_val_precision": 0.7867265343666077, + "electricity_val_recall": 0.783363401889801, + "electricity_val_f1-score": 0.7784221768379211, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8638512492179871, + "electricity_val_score": 0.783363401889801 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.46528056263923645, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7815302610397339, + "electricity_test_precision": 0.7850295305252075, + "electricity_test_recall": 0.7815302610397339, + "electricity_test_f1-score": 0.7764526009559631, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8586947917938232, + "electricity_test_score": 0.7815302610397339 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.49843543767929077, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7497745752334595, + "electricity_val_precision": 0.7648465633392334, + "electricity_val_recall": 0.7497745752334595, + "electricity_val_f1-score": 0.7512147426605225, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8493627905845642, + "electricity_val_score": 0.7497745752334595 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4989757537841797, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7491971850395203, + "electricity_test_precision": 0.7638506889343262, + "electricity_test_recall": 0.7491971850395203, + "electricity_test_f1-score": 0.7506317496299744, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.848699688911438, + "electricity_test_score": 0.7491971850395203 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.44793573021888733, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.790802538394928, + "electricity_val_precision": 0.7906069755554199, + "electricity_val_recall": 0.790802538394928, + "electricity_val_f1-score": 0.7884482741355896, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8664727210998535, + "electricity_val_score": 0.790802538394928 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.45670565962791443, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7809766530990601, + "electricity_test_precision": 0.780791699886322, + "electricity_test_recall": 0.7809766530990601, + "electricity_test_f1-score": 0.778240442276001, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8612192869186401, + "electricity_test_score": 0.7809766530990601 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.46604178547859193, + "std": 0.0189413763795482, + "n": 5, + "ci95": 0.023515054711538058 + }, + "lr": { + "mean": 0.0009999994654208422, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.7767688989639282, + "std": 0.0158138252645006, + "n": 5, + "ci95": 0.01963230965068359 + }, + "electricity_test_precision": { + "mean": 0.7809054136276246, + "std": 0.009959665369242492, + "n": 5, + "ci95": 0.012364575381080852 + }, + "electricity_test_recall": { + "mean": 0.7767688989639282, + "std": 0.0158138252645006, + "n": 5, + "ci95": 0.01963230965068359 + }, + "electricity_test_f1-score": { + "mean": 0.774288022518158, + "std": 0.014137674974167933, + "n": 5, + "ci95": 0.017551427829207784 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8596317648887635, + "std": 0.006629733630725378, + "n": 5, + "ci95": 0.008230581871429433 + }, + "electricity_test_score": { + "mean": 0.7767688989639282, + "std": 0.0158138252645006, + "n": 5, + "ci95": 0.01963230965068359 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/downstream_results/resnet_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/downstream_results/resnet_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..8fb66801ee89c12e08045786915a1e76e3880544 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/downstream_results/resnet_electricity_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "electricity", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp13ef8maa", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp13ef8maa/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.4096173048019409, + "lr": 0.0, + "electricity_val_accuracy": 0.814472496509552, + "electricity_val_precision": 0.8137989640235901, + "electricity_val_recall": 0.814472496509552, + "electricity_val_f1-score": 0.8137029409408569, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8902553915977478, + "electricity_val_score": 0.814472496509552 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4122096300125122, + "lr": 0.0, + "electricity_test_accuracy": 0.8100985288619995, + "electricity_test_precision": 0.809472918510437, + "electricity_test_recall": 0.8100985288619995, + "electricity_test_f1-score": 0.8095842599868774, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8895787596702576, + "electricity_test_score": 0.8100985288619995 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.4212607741355896, + "lr": 0.0, + "electricity_val_accuracy": 0.8061316609382629, + "electricity_val_precision": 0.8127837777137756, + "electricity_val_recall": 0.8061316609382629, + "electricity_val_f1-score": 0.8071792125701904, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8945531249046326, + "electricity_val_score": 0.8061316609382629 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.42931613326072693, + "lr": 0.0, + "electricity_test_accuracy": 0.8027904033660889, + "electricity_test_precision": 0.8102234601974487, + "electricity_test_recall": 0.8027904033660889, + "electricity_test_f1-score": 0.8038804531097412, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8904880285263062, + "electricity_test_score": 0.8027904033660889 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.3943096399307251, + "lr": 0.0, + "electricity_val_accuracy": 0.8174030780792236, + "electricity_val_precision": 0.8178496360778809, + "electricity_val_recall": 0.8174030780792236, + "electricity_val_f1-score": 0.8175885677337646, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8998651504516602, + "electricity_val_score": 0.8174030780792236 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3973785638809204, + "lr": 0.0, + "electricity_test_accuracy": 0.8195105791091919, + "electricity_test_precision": 0.8203485608100891, + "electricity_test_recall": 0.8195105791091919, + "electricity_test_f1-score": 0.8198157548904419, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8987885117530823, + "electricity_test_score": 0.8195105791091919 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.3988509178161621, + "lr": 0.0, + "electricity_val_accuracy": 0.8128945231437683, + "electricity_val_precision": 0.8125920295715332, + "electricity_val_recall": 0.8128945231437683, + "electricity_val_f1-score": 0.8127148747444153, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8960681557655334, + "electricity_val_score": 0.8128945231437683 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4048568606376648, + "lr": 0.0, + "electricity_test_accuracy": 0.8120917081832886, + "electricity_test_precision": 0.8119973540306091, + "electricity_test_recall": 0.8120917081832886, + "electricity_test_f1-score": 0.812042236328125, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8937528729438782, + "electricity_test_score": 0.8120917081832886 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.3939351737499237, + "lr": 0.0, + "electricity_val_accuracy": 0.814021646976471, + "electricity_val_precision": 0.8133492469787598, + "electricity_val_recall": 0.814021646976471, + "electricity_val_f1-score": 0.8131622672080994, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8984636068344116, + "electricity_val_score": 0.814021646976471 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.39690616726875305, + "lr": 0.0, + "electricity_test_accuracy": 0.8168530464172363, + "electricity_test_precision": 0.8162230253219604, + "electricity_test_recall": 0.8168530464172363, + "electricity_test_f1-score": 0.8162349462509155, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8977286219596863, + "electricity_test_score": 0.8168530464172363 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.4081334710121155, + "std": 0.013395678448921245, + "n": 5, + "ci95": 0.016630265156690152 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.812268853187561, + "std": 0.006481514501639444, + "n": 5, + "ci95": 0.008046573019067702 + }, + "electricity_test_precision": { + "mean": 0.8136530637741088, + "std": 0.004566496271470452, + "n": 5, + "ci95": 0.005669145024730432 + }, + "electricity_test_recall": { + "mean": 0.812268853187561, + "std": 0.006481514501639444, + "n": 5, + "ci95": 0.008046573019067702 + }, + "electricity_test_f1-score": { + "mean": 0.8123115301132202, + "std": 0.006129848379444652, + "n": 5, + "ci95": 0.007609991857387507 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.894067358970642, + "std": 0.004145846169368799, + "n": 5, + "ci95": 0.005146922670497779 + }, + "electricity_test_score": { + "mean": 0.812268853187561, + "std": 0.006481514501639444, + "n": 5, + "ci95": 0.008046573019067702 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..73de94be36f0ffa25a38e0372fe8582311d036e4 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:75d0faad233e0e1864e385de6f119ac4de8bf4ca331693eb2ca3c2b68468d14c +size 92799104 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..8cfe5ae930d5d011c3934fac5044575faea21a84 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:09b51aeada46f222ea7559d1c729661d838f73a71fa1ec5d5819ad6d024b0b09 +size 8331758 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a4fedc4a2d1c1649c7988204863e538e9d49b971 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/electricity/20260915_225442/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: electricity +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.0005 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.6 +mask_max_trgt_share: 0.9 +mask_min_ctx_share: 0.15 +mask_min_trgt_share: 0.15 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 64 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 6 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 128 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.001 +recon_y_weight: 0.25 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 2.5e-05 +sw_s_y_weight: 5.0e-06 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 40 +vtjepa_kl_x_end_weight: 1.0e-06 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-06 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 40 +vtjepa_kl_z_end_weight: 1.0e-05 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 40 +wjepa_joint_end_weight: 0.000125 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..3f2360a4a78627d32006763862614173c073f6c8 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-16T01:20:23.006669+00:00", + "spec": { + "dataset": "electricity", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 2.5e-05, + "sw_z_weight": 0.0, + "sw_s_y_weight": 1e-05, + "lambda_joint": 0.00025, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..1473f695aa1cf00ad3e3e7b3259ef5bca7043902 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "electricity", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 2.5e-05, + "sw_z_weight": 0.0, + "sw_s_y_weight": 1e-05, + "lambda_joint": 0.00025, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "electricity", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.00025", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "2.5e-05", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "1e-05", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..db81aba2042f118569d89a140643ed8482c326f4 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "electricity", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.00025, + "s_x": 2.5e-05, + "z": 0.0, + "s_y": 1e-05 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 40 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "electricity", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "64", + "--model_num_layers", + "6", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.15", + "--mask_max_ctx_share", + "0.6", + "--mask_min_trgt_share", + "0.15", + "--mask_max_trgt_share", + "0.9", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "128", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.0005", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "1e-06", + "--vtjepa_kl_x_annealing_epochs", + "40", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-05", + "--vtjepa_kl_z_annealing_epochs", + "40", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-06", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.001", + "--recon_y_weight", + "0.25", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.00025", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "40", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "2.5e-05", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "1e-05", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/downstream_results/autoint_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/downstream_results/autoint_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..4c459cecb1eb790df1d9d6a5b704425473c99316 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/downstream_results/autoint_electricity_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "electricity", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpeg5zq0f_", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpeg5zq0f_/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.37676724791526794, + "lr": 0.0, + "electricity_val_accuracy": 0.824165940284729, + "electricity_val_precision": 0.823758065700531, + "electricity_val_recall": 0.824165940284729, + "electricity_val_f1-score": 0.8238867521286011, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9085114002227783, + "electricity_val_score": 0.824165940284729 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3778446912765503, + "lr": 0.0, + "electricity_test_accuracy": 0.826597273349762, + "electricity_test_precision": 0.8260992169380188, + "electricity_test_recall": 0.826597273349762, + "electricity_test_f1-score": 0.8261891603469849, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9077622890472412, + "electricity_test_score": 0.826597273349762 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.39786243438720703, + "lr": 0.0, + "electricity_val_accuracy": 0.8110910654067993, + "electricity_val_precision": 0.8104005455970764, + "electricity_val_recall": 0.8110910654067993, + "electricity_val_f1-score": 0.8101359605789185, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8985281586647034, + "electricity_val_score": 0.8110910654067993 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3984154760837555, + "lr": 0.0, + "electricity_test_accuracy": 0.8195105791091919, + "electricity_test_precision": 0.8188907504081726, + "electricity_test_recall": 0.8195105791091919, + "electricity_test_f1-score": 0.8188132047653198, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8984673023223877, + "electricity_test_score": 0.8195105791091919 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.3746926486492157, + "lr": 0.0, + "electricity_val_accuracy": 0.8277727961540222, + "electricity_val_precision": 0.8272218108177185, + "electricity_val_recall": 0.8277727961540222, + "electricity_val_f1-score": 0.8272076845169067, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9104304909706116, + "electricity_val_score": 0.8277727961540222 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.37780824303627014, + "lr": 0.0, + "electricity_test_accuracy": 0.8321337699890137, + "electricity_test_precision": 0.8316606283187866, + "electricity_test_recall": 0.8321337699890137, + "electricity_test_f1-score": 0.8317317366600037, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9088961482048035, + "electricity_test_score": 0.8321337699890137 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.3830503523349762, + "lr": 0.0, + "electricity_val_accuracy": 0.8291253447532654, + "electricity_val_precision": 0.8286396265029907, + "electricity_val_recall": 0.8291253447532654, + "electricity_val_f1-score": 0.8287276029586792, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.906862199306488, + "electricity_val_score": 0.8291253447532654 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3782241642475128, + "lr": 0.0, + "electricity_test_accuracy": 0.8285903930664062, + "electricity_test_precision": 0.8282761573791504, + "electricity_test_recall": 0.8285903930664062, + "electricity_test_f1-score": 0.8283909559249878, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9089770913124084, + "electricity_test_score": 0.8285903930664062 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.4047601819038391, + "lr": 0.0, + "electricity_val_accuracy": 0.8151487708091736, + "electricity_val_precision": 0.8145381808280945, + "electricity_val_recall": 0.8151487708091736, + "electricity_val_f1-score": 0.8141227960586548, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8940008878707886, + "electricity_val_score": 0.8151487708091736 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.40314340591430664, + "lr": 0.0, + "electricity_test_accuracy": 0.8158565163612366, + "electricity_test_precision": 0.8152197599411011, + "electricity_test_recall": 0.8158565163612366, + "electricity_test_f1-score": 0.8149707317352295, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8943537473678589, + "electricity_test_score": 0.8158565163612366 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.3870871961116791, + "std": 0.012611581978284757, + "n": 5, + "ci95": 0.015656836877948527 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.824537706375122, + "std": 0.006689569055104953, + "n": 5, + "ci95": 0.008304865453032988 + }, + "electricity_test_precision": { + "mean": 0.8240293025970459, + "std": 0.006792831029755497, + "n": 5, + "ci95": 0.008433061574311208 + }, + "electricity_test_recall": { + "mean": 0.824537706375122, + "std": 0.006689569055104953, + "n": 5, + "ci95": 0.008304865453032988 + }, + "electricity_test_f1-score": { + "mean": 0.8240191578865051, + "std": 0.006933277793011343, + "n": 5, + "ci95": 0.00860742130698542 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.90369131565094, + "std": 0.006820612643828555, + "n": 5, + "ci95": 0.0084675514741903 + }, + "electricity_test_score": { + "mean": 0.824537706375122, + "std": 0.006689569055104953, + "n": 5, + "ci95": 0.008304865453032988 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/downstream_results/dcnv2_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/downstream_results/dcnv2_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..037a681603cf20e5fe8601c58804ad4b38ab1681 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/downstream_results/dcnv2_electricity_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "electricity", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp9pcho2x0", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp9pcho2x0/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.3700176477432251, + "lr": 0.0, + "electricity_val_accuracy": 0.8284490704536438, + "electricity_val_precision": 0.8279020190238953, + "electricity_val_recall": 0.8284490704536438, + "electricity_val_f1-score": 0.8278785347938538, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9132524728775024, + "electricity_val_score": 0.8284490704536438 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3751595616340637, + "lr": 0.0, + "electricity_test_accuracy": 0.832023024559021, + "electricity_test_precision": 0.8315133452415466, + "electricity_test_recall": 0.832023024559021, + "electricity_test_f1-score": 0.8315108418464661, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9120344519615173, + "electricity_test_score": 0.832023024559021 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.368421345949173, + "lr": 0.0, + "electricity_val_accuracy": 0.8322813510894775, + "electricity_val_precision": 0.8317623138427734, + "electricity_val_recall": 0.8322813510894775, + "electricity_val_f1-score": 0.8317160606384277, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9140526652336121, + "electricity_val_score": 0.8322813510894775 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.378262996673584, + "lr": 0.0, + "electricity_test_accuracy": 0.8324659466743469, + "electricity_test_precision": 0.8320069313049316, + "electricity_test_recall": 0.8324659466743469, + "electricity_test_f1-score": 0.8320888876914978, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9107716083526611, + "electricity_test_score": 0.8324659466743469 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.37066081166267395, + "lr": 0.0, + "electricity_val_accuracy": 0.832055926322937, + "electricity_val_precision": 0.8315445184707642, + "electricity_val_recall": 0.832055926322937, + "electricity_val_f1-score": 0.8315567970275879, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9143795371055603, + "electricity_val_score": 0.832055926322937 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3715841472148895, + "lr": 0.0, + "electricity_test_accuracy": 0.8342376351356506, + "electricity_test_precision": 0.8338351249694824, + "electricity_test_recall": 0.8342376351356506, + "electricity_test_f1-score": 0.8339382410049438, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9133196473121643, + "electricity_test_score": 0.8342376351356506 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.37516358494758606, + "lr": 0.0, + "electricity_val_accuracy": 0.8284490704536438, + "electricity_val_precision": 0.827900767326355, + "electricity_val_recall": 0.8284490704536438, + "electricity_val_f1-score": 0.8277851939201355, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9116746783256531, + "electricity_val_score": 0.8284490704536438 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.37138354778289795, + "lr": 0.0, + "electricity_test_accuracy": 0.8377809524536133, + "electricity_test_precision": 0.8373184204101562, + "electricity_test_recall": 0.8377809524536133, + "electricity_test_f1-score": 0.8373348116874695, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9131065011024475, + "electricity_test_score": 0.8377809524536133 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.3669682741165161, + "lr": 0.0, + "electricity_val_accuracy": 0.8343101739883423, + "electricity_val_precision": 0.8338348865509033, + "electricity_val_recall": 0.8343101739883423, + "electricity_val_f1-score": 0.8338894248008728, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9141197800636292, + "electricity_val_score": 0.8343101739883423 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3763114809989929, + "lr": 0.0, + "electricity_test_accuracy": 0.8309157490730286, + "electricity_test_precision": 0.830554187297821, + "electricity_test_recall": 0.8309157490730286, + "electricity_test_f1-score": 0.830670177936554, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9112343788146973, + "electricity_test_score": 0.8309157490730286 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.3745403468608856, + "std": 0.0030034547384739088, + "n": 5, + "ci95": 0.0037286837600197034 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.833484661579132, + "std": 0.002683214432111806, + "n": 5, + "ci95": 0.0033311166469415054 + }, + "electricity_test_precision": { + "mean": 0.8330456018447876, + "std": 0.002669930640033729, + "n": 5, + "ci95": 0.0033146252847916064 + }, + "electricity_test_recall": { + "mean": 0.833484661579132, + "std": 0.002683214432111806, + "n": 5, + "ci95": 0.0033311166469415054 + }, + "electricity_test_f1-score": { + "mean": 0.8331085920333863, + "std": 0.0026498922569071592, + "n": 5, + "ci95": 0.003289748334663482 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.9120933175086975, + "std": 0.0011201095601950212, + "n": 5, + "ci95": 0.001390576749181889 + }, + "electricity_test_score": { + "mean": 0.833484661579132, + "std": 0.002683214432111806, + "n": 5, + "ci95": 0.0033311166469415054 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/downstream_results/fttransformer_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/downstream_results/fttransformer_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..b4158a85a98160600cf81d27f5ace03168bab1e8 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/downstream_results/fttransformer_electricity_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "electricity", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpki_1xno3", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpki_1xno3/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.42834633588790894, + "lr": 0.0, + "electricity_val_accuracy": 0.7989179491996765, + "electricity_val_precision": 0.7981755137443542, + "electricity_val_recall": 0.7989179491996765, + "electricity_val_f1-score": 0.7975538969039917, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8783780336380005, + "electricity_val_score": 0.7989179491996765 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.43527716398239136, + "lr": 0.0, + "electricity_test_accuracy": 0.7929354310035706, + "electricity_test_precision": 0.7921723127365112, + "electricity_test_recall": 0.7929354310035706, + "electricity_test_f1-score": 0.7913938760757446, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8737823367118835, + "electricity_test_score": 0.7929354310035706 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.43609222769737244, + "lr": 0.0, + "electricity_val_accuracy": 0.7901262640953064, + "electricity_val_precision": 0.7896750569343567, + "electricity_val_recall": 0.7901262640953064, + "electricity_val_f1-score": 0.7880067825317383, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8720516562461853, + "electricity_val_score": 0.7901262640953064 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.44743162393569946, + "lr": 0.0, + "electricity_test_accuracy": 0.7825268507003784, + "electricity_test_precision": 0.782005786895752, + "electricity_test_recall": 0.7825268507003784, + "electricity_test_f1-score": 0.7802097797393799, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8652223944664001, + "electricity_test_score": 0.7825268507003784 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.43264248967170715, + "lr": 0.0, + "electricity_val_accuracy": 0.7962127923965454, + "electricity_val_precision": 0.7960327863693237, + "electricity_val_recall": 0.7962127923965454, + "electricity_val_f1-score": 0.7940444350242615, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8752588629722595, + "electricity_val_score": 0.7962127923965454 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4366978108882904, + "lr": 0.0, + "electricity_test_accuracy": 0.7932676076889038, + "electricity_test_precision": 0.7932544946670532, + "electricity_test_recall": 0.7932676076889038, + "electricity_test_f1-score": 0.7908713221549988, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.872521162033081, + "electricity_test_score": 0.7932676076889038 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.43429580330848694, + "lr": 0.0, + "electricity_val_accuracy": 0.7937331199645996, + "electricity_val_precision": 0.7934473752975464, + "electricity_val_recall": 0.7937331199645996, + "electricity_val_f1-score": 0.7915757894515991, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8738759160041809, + "electricity_val_score": 0.7937331199645996 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.44147440791130066, + "lr": 0.0, + "electricity_test_accuracy": 0.7891706228256226, + "electricity_test_precision": 0.7887260317802429, + "electricity_test_recall": 0.7891706228256226, + "electricity_test_f1-score": 0.7870363593101501, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8702382445335388, + "electricity_test_score": 0.7891706228256226 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.41187557578086853, + "lr": 0.0, + "electricity_val_accuracy": 0.8119927644729614, + "electricity_val_precision": 0.811437726020813, + "electricity_val_recall": 0.8119927644729614, + "electricity_val_f1-score": 0.8107569813728333, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8886616826057434, + "electricity_val_score": 0.8119927644729614 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.413981169462204, + "lr": 0.0, + "electricity_test_accuracy": 0.8086590766906738, + "electricity_test_precision": 0.8080326914787292, + "electricity_test_recall": 0.8086590766906738, + "electricity_test_f1-score": 0.8074625730514526, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8874712586402893, + "electricity_test_score": 0.8086590766906738 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.4349724352359772, + "std": 0.012661062862785347, + "n": 5, + "ci95": 0.01571826566131102 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.7933119177818299, + "std": 0.009606599139622246, + "n": 5, + "ci95": 0.011926256035118172 + }, + "electricity_test_precision": { + "mean": 0.7928382635116578, + "std": 0.009562169987810893, + "n": 5, + "ci95": 0.011871098800781217 + }, + "electricity_test_recall": { + "mean": 0.7933119177818299, + "std": 0.009606599139622246, + "n": 5, + "ci95": 0.011926256035118172 + }, + "electricity_test_f1-score": { + "mean": 0.7913947820663452, + "std": 0.010031799990737505, + "n": 5, + "ci95": 0.012454127984706999 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8738470792770385, + "std": 0.008288386268573969, + "n": 5, + "ci95": 0.010289740970794568 + }, + "electricity_test_score": { + "mean": 0.7933119177818299, + "std": 0.009606599139622246, + "n": 5, + "ci95": 0.011926256035118172 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/downstream_results/mlp_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/downstream_results/mlp_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..61b30bc2e27ee816341fcca09ed74adfe64ad038 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/downstream_results/mlp_electricity_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "electricity", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpdr916ycy", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpdr916ycy/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.46313613653182983, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7793056964874268, + "electricity_val_precision": 0.7825559377670288, + "electricity_val_recall": 0.7793056964874268, + "electricity_val_f1-score": 0.7801650762557983, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8617111444473267, + "electricity_val_score": 0.7793056964874268 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4705192446708679, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7733362913131714, + "electricity_test_precision": 0.7759066224098206, + "electricity_test_recall": 0.7733362913131714, + "electricity_test_f1-score": 0.7740969657897949, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8568459749221802, + "electricity_test_score": 0.7733362913131714 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.5063551664352417, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7590171098709106, + "electricity_val_precision": 0.7836204171180725, + "electricity_val_recall": 0.7590171098709106, + "electricity_val_f1-score": 0.7436493635177612, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8581312894821167, + "electricity_val_score": 0.7590171098709106 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.5209208726882935, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7529619932174683, + "electricity_test_precision": 0.7779517769813538, + "electricity_test_recall": 0.7529619932174683, + "electricity_test_f1-score": 0.7366639375686646, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8485890626907349, + "electricity_test_score": 0.7529619932174683 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.4453205466270447, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7896753549575806, + "electricity_val_precision": 0.7887317538261414, + "electricity_val_recall": 0.7896753549575806, + "electricity_val_f1-score": 0.7885810732841492, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8678401112556458, + "electricity_val_score": 0.7896753549575806 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.45247143507003784, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7888384461402893, + "electricity_test_precision": 0.787907600402832, + "electricity_test_recall": 0.7888384461402893, + "electricity_test_f1-score": 0.7875658273696899, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8647524118423462, + "electricity_test_score": 0.7888384461402893 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.571717381477356, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.6995040774345398, + "electricity_val_precision": 0.7549394965171814, + "electricity_val_recall": 0.6995040774345398, + "electricity_val_f1-score": 0.6964490413665771, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.842940092086792, + "electricity_val_score": 0.6995040774345398 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.5833672881126404, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.690731942653656, + "electricity_test_precision": 0.7448217868804932, + "electricity_test_recall": 0.690731942653656, + "electricity_test_f1-score": 0.6875823140144348, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8353418707847595, + "electricity_test_score": 0.690731942653656 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.4550061523914337, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.78291255235672, + "electricity_val_precision": 0.7862043380737305, + "electricity_val_recall": 0.78291255235672, + "electricity_val_f1-score": 0.7837667465209961, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8672930002212524, + "electricity_val_score": 0.78291255235672 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.45936721563339233, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.778983473777771, + "electricity_test_precision": 0.7815658450126648, + "electricity_test_recall": 0.778983473777771, + "electricity_test_f1-score": 0.7797327041625977, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8643395304679871, + "electricity_test_score": 0.778983473777771 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.49732921123504636, + "std": 0.05507091941901319, + "n": 5, + "ci95": 0.06836861573328218 + }, + "lr": { + "mean": 0.0009999994654208422, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.7569704294204712, + "std": 0.03927973630567384, + "n": 5, + "ci95": 0.04876441551945655 + }, + "electricity_test_precision": { + "mean": 0.7736307263374329, + "std": 0.016737692610561187, + "n": 5, + "ci95": 0.020779258571052193 + }, + "electricity_test_recall": { + "mean": 0.7569704294204712, + "std": 0.03927973630567384, + "n": 5, + "ci95": 0.04876441551945655 + }, + "electricity_test_f1-score": { + "mean": 0.7531283497810364, + "std": 0.04153544891495393, + "n": 5, + "ci95": 0.05156480364109276 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8539737701416016, + "std": 0.012329034961561139, + "n": 5, + "ci95": 0.015306064662471532 + }, + "electricity_test_score": { + "mean": 0.7569704294204712, + "std": 0.03927973630567384, + "n": 5, + "ci95": 0.04876441551945655 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/downstream_results/resnet_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/downstream_results/resnet_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..c691ed60d8f8930ac29d461b9b4e3aabe4372bee --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/downstream_results/resnet_electricity_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "electricity", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpr9mtrvxr", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpr9mtrvxr/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.4002872109413147, + "lr": 0.0, + "electricity_val_accuracy": 0.8189810514450073, + "electricity_val_precision": 0.8183836936950684, + "electricity_val_recall": 0.8189810514450073, + "electricity_val_f1-score": 0.8184431195259094, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8961666226387024, + "electricity_val_score": 0.8189810514450073 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4078831672668457, + "lr": 0.0, + "electricity_test_accuracy": 0.8093234300613403, + "electricity_test_precision": 0.8087258338928223, + "electricity_test_recall": 0.8093234300613403, + "electricity_test_f1-score": 0.8088589310646057, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8922305107116699, + "electricity_test_score": 0.8093234300613403 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.40211179852485657, + "lr": 0.0, + "electricity_val_accuracy": 0.8083859086036682, + "electricity_val_precision": 0.8125491142272949, + "electricity_val_recall": 0.8083859086036682, + "electricity_val_f1-score": 0.8092515468597412, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9013146162033081, + "electricity_val_score": 0.8083859086036682 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.40903201699256897, + "lr": 0.0, + "electricity_test_accuracy": 0.8140848278999329, + "electricity_test_precision": 0.8187415599822998, + "electricity_test_recall": 0.8140848278999329, + "electricity_test_f1-score": 0.8149641156196594, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8984622359275818, + "electricity_test_score": 0.8140848278999329 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.39618009328842163, + "lr": 0.0, + "electricity_val_accuracy": 0.8178539276123047, + "electricity_val_precision": 0.8199416399002075, + "electricity_val_recall": 0.8178539276123047, + "electricity_val_f1-score": 0.8184287548065186, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9009413123130798, + "electricity_val_score": 0.8178539276123047 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.399535208940506, + "lr": 0.0, + "electricity_test_accuracy": 0.8174067139625549, + "electricity_test_precision": 0.8203089833259583, + "electricity_test_recall": 0.8174067139625549, + "electricity_test_f1-score": 0.8180980086326599, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8999048471450806, + "electricity_test_score": 0.8174067139625549 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.3925440013408661, + "lr": 0.0, + "electricity_val_accuracy": 0.8160504698753357, + "electricity_val_precision": 0.8161037564277649, + "electricity_val_recall": 0.8160504698753357, + "electricity_val_f1-score": 0.8160764575004578, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.899478554725647, + "electricity_val_score": 0.8160504698753357 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.40357688069343567, + "lr": 0.0, + "electricity_test_accuracy": 0.8128668069839478, + "electricity_test_precision": 0.8131201267242432, + "electricity_test_recall": 0.8128668069839478, + "electricity_test_f1-score": 0.8129804134368896, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8943611979484558, + "electricity_test_score": 0.8128668069839478 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.37866079807281494, + "lr": 0.0, + "electricity_val_accuracy": 0.8228133320808411, + "electricity_val_precision": 0.8225879669189453, + "electricity_val_recall": 0.8228133320808411, + "electricity_val_f1-score": 0.8226836919784546, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9069890975952148, + "electricity_val_score": 0.8228133320808411 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3883085548877716, + "lr": 0.0, + "electricity_test_accuracy": 0.8222787976264954, + "electricity_test_precision": 0.8222246170043945, + "electricity_test_recall": 0.8222787976264954, + "electricity_test_f1-score": 0.8222509026527405, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9028205871582031, + "electricity_test_score": 0.8222787976264954 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.4016671657562256, + "std": 0.008359492267084372, + "n": 5, + "ci95": 0.010378016575047704 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.8151921153068542, + "std": 0.0049040831001921975, + "n": 5, + "ci95": 0.006088247237168271 + }, + "electricity_test_precision": { + "mean": 0.8166242241859436, + "std": 0.005569744804723517, + "n": 5, + "ci95": 0.006914642905982019 + }, + "electricity_test_recall": { + "mean": 0.8151921153068542, + "std": 0.0049040831001921975, + "n": 5, + "ci95": 0.006088247237168271 + }, + "electricity_test_f1-score": { + "mean": 0.815430474281311, + "std": 0.0050754175068933, + "n": 5, + "ci95": 0.00630095289629321 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8975558757781983, + "std": 0.004260921867543, + "n": 5, + "ci95": 0.005289785115354562 + }, + "electricity_test_score": { + "mean": 0.8151921153068542, + "std": 0.0049040831001921975, + "n": 5, + "ci95": 0.006088247237168271 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..e0a30f747b4707004b85d2e9c9d5c6f7f7fb4b8d --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fbf09a3b6af5038bd3c3c2af1a155af0ce2eac8c1a2a42c43a0499d238942a88 +size 92799104 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..ccc49c062a2e438abd1f79a42ffe23a5838c68b2 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:329981e1323efcb7a9b21ce2f0ee263fd57b19888b8262c5d34234224168a944 +size 8332910 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..899e344b7b517389011eb0af5d86656ed0aac011 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260916_000611/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: electricity +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.0005 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.6 +mask_max_trgt_share: 0.9 +mask_min_ctx_share: 0.15 +mask_min_trgt_share: 0.15 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 64 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 6 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 128 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.001 +recon_y_weight: 0.25 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 2.5e-05 +sw_s_y_weight: 1.0e-05 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 40 +vtjepa_kl_x_end_weight: 1.0e-06 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-06 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 40 +vtjepa_kl_z_end_weight: 1.0e-05 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 40 +wjepa_joint_end_weight: 0.00025 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..f09ee7dd51359f1392ac2dcdcf86e77f2b9aa8f4 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-15T23:31:06.533975+00:00", + "spec": { + "dataset": "electricity", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 5e-05, + "sw_z_weight": 0.0, + "sw_s_y_weight": 1e-05, + "lambda_joint": 0.00025, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..0ca0f04556fb8a2e4ef0169ee78a19705114e7c1 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "electricity", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 5e-05, + "sw_z_weight": 0.0, + "sw_s_y_weight": 1e-05, + "lambda_joint": 0.00025, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "electricity", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.00025", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "5e-05", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "1e-05", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..e789b1f9c19db368a7c140a263524132f61c0764 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "electricity", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.00025, + "s_x": 5e-05, + "z": 0.0, + "s_y": 1e-05 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 40 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "electricity", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "64", + "--model_num_layers", + "6", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.15", + "--mask_max_ctx_share", + "0.6", + "--mask_min_trgt_share", + "0.15", + "--mask_max_trgt_share", + "0.9", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "128", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.0005", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "1e-06", + "--vtjepa_kl_x_annealing_epochs", + "40", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-05", + "--vtjepa_kl_z_annealing_epochs", + "40", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-06", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.001", + "--recon_y_weight", + "0.25", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.00025", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "40", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "5e-05", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "1e-05", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/downstream_results/autoint_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/downstream_results/autoint_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..619c8cc8ae2402a6fad684176c2118313cbd27b2 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/downstream_results/autoint_electricity_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "electricity", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpj6t9z2ll", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpj6t9z2ll/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.38083615899086, + "lr": 0.0, + "electricity_val_accuracy": 0.8225879073143005, + "electricity_val_precision": 0.8220837712287903, + "electricity_val_recall": 0.8225879073143005, + "electricity_val_f1-score": 0.8221973180770874, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9075164198875427, + "electricity_val_score": 0.8225879073143005 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3831469416618347, + "lr": 0.0, + "electricity_test_accuracy": 0.826154351234436, + "electricity_test_precision": 0.8257611393928528, + "electricity_test_recall": 0.826154351234436, + "electricity_test_f1-score": 0.8258849382400513, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9055925607681274, + "electricity_test_score": 0.826154351234436 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.39049676060676575, + "lr": 0.0, + "electricity_val_accuracy": 0.8221370577812195, + "electricity_val_precision": 0.8215466737747192, + "electricity_val_recall": 0.8221370577812195, + "electricity_val_f1-score": 0.8215613961219788, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9021830558776855, + "electricity_val_score": 0.8221370577812195 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.39296814799308777, + "lr": 0.0, + "electricity_test_accuracy": 0.8236075639724731, + "electricity_test_precision": 0.8230963349342346, + "electricity_test_recall": 0.8236075639724731, + "electricity_test_f1-score": 0.8231960535049438, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9012018442153931, + "electricity_test_score": 0.8236075639724731 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.37689918279647827, + "lr": 0.0, + "electricity_val_accuracy": 0.8264201879501343, + "electricity_val_precision": 0.8259297013282776, + "electricity_val_recall": 0.8264201879501343, + "electricity_val_f1-score": 0.8260307908058167, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9090912342071533, + "electricity_val_score": 0.8264201879501343 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3787509799003601, + "lr": 0.0, + "electricity_test_accuracy": 0.8295869827270508, + "electricity_test_precision": 0.8290953040122986, + "electricity_test_recall": 0.8295869827270508, + "electricity_test_f1-score": 0.8291682600975037, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9082685112953186, + "electricity_test_score": 0.8295869827270508 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.39570626616477966, + "lr": 0.0, + "electricity_val_accuracy": 0.8155996203422546, + "electricity_val_precision": 0.8151055574417114, + "electricity_val_recall": 0.8155996203422546, + "electricity_val_f1-score": 0.8152470588684082, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8984375596046448, + "electricity_val_score": 0.8155996203422546 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3995775878429413, + "lr": 0.0, + "electricity_test_accuracy": 0.8150813579559326, + "electricity_test_precision": 0.8144586682319641, + "electricity_test_recall": 0.8150813579559326, + "electricity_test_f1-score": 0.8145215511322021, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8969169855117798, + "electricity_test_score": 0.8150813579559326 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.39763420820236206, + "lr": 0.0, + "electricity_val_accuracy": 0.8162758946418762, + "electricity_val_precision": 0.816234827041626, + "electricity_val_recall": 0.8162758946418762, + "electricity_val_f1-score": 0.8146142959594727, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8983796834945679, + "electricity_val_score": 0.8162758946418762 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.399333119392395, + "lr": 0.0, + "electricity_test_accuracy": 0.8195105791091919, + "electricity_test_precision": 0.8192992210388184, + "electricity_test_recall": 0.8195105791091919, + "electricity_test_f1-score": 0.8180992603302002, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8965386748313904, + "electricity_test_score": 0.8195105791091919 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.3907553553581238, + "std": 0.009464397262821271, + "n": 5, + "ci95": 0.011749717390510018 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.8227881669998169, + "std": 0.005664891680622579, + "n": 5, + "ci95": 0.007032764416666647 + }, + "electricity_test_precision": { + "mean": 0.8223421335220337, + "std": 0.005684893885238435, + "n": 5, + "ci95": 0.007057596452442099 + }, + "electricity_test_recall": { + "mean": 0.8227881669998169, + "std": 0.005664891680622579, + "n": 5, + "ci95": 0.007032764416666647 + }, + "electricity_test_f1-score": { + "mean": 0.8221740126609802, + "std": 0.005892766908466559, + "n": 5, + "ci95": 0.007315663522981918 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.9017037153244019, + "std": 0.005197642860674944, + "n": 5, + "ci95": 0.006452691387927628 + }, + "electricity_test_score": { + "mean": 0.8227881669998169, + "std": 0.005664891680622579, + "n": 5, + "ci95": 0.007032764416666647 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/downstream_results/dcnv2_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/downstream_results/dcnv2_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..5de7b532c048d9ecfd8d08b4418f0218337339ef --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/downstream_results/dcnv2_electricity_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "electricity", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpvoeug6yj", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpvoeug6yj/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.36828869581222534, + "lr": 0.0, + "electricity_val_accuracy": 0.8331830501556396, + "electricity_val_precision": 0.832675039768219, + "electricity_val_recall": 0.8331830501556396, + "electricity_val_f1-score": 0.8325451612472534, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9150740504264832, + "electricity_val_score": 0.8331830501556396 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.37148943543434143, + "lr": 0.0, + "electricity_test_accuracy": 0.8351234793663025, + "electricity_test_precision": 0.8347146511077881, + "electricity_test_recall": 0.8351234793663025, + "electricity_test_f1-score": 0.8348125219345093, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.913567841053009, + "electricity_test_score": 0.8351234793663025 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.37160301208496094, + "lr": 0.0, + "electricity_val_accuracy": 0.8325067758560181, + "electricity_val_precision": 0.8320543169975281, + "electricity_val_recall": 0.8325067758560181, + "electricity_val_f1-score": 0.831684410572052, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9119824767112732, + "electricity_val_score": 0.8325067758560181 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.37350720167160034, + "lr": 0.0, + "electricity_test_accuracy": 0.834126889705658, + "electricity_test_precision": 0.8336272835731506, + "electricity_test_recall": 0.834126889705658, + "electricity_test_f1-score": 0.8335598111152649, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9122480750083923, + "electricity_test_score": 0.834126889705658 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.3726440668106079, + "lr": 0.0, + "electricity_val_accuracy": 0.8345355987548828, + "electricity_val_precision": 0.8340369462966919, + "electricity_val_recall": 0.8345355987548828, + "electricity_val_f1-score": 0.8340221047401428, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9129883646965027, + "electricity_val_score": 0.8345355987548828 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.37364310026168823, + "lr": 0.0, + "electricity_test_accuracy": 0.8334625363349915, + "electricity_test_precision": 0.8330397605895996, + "electricity_test_recall": 0.8334625363349915, + "electricity_test_f1-score": 0.8331384062767029, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.913081705570221, + "electricity_test_score": 0.8334625363349915 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.3699244558811188, + "lr": 0.0, + "electricity_val_accuracy": 0.8279982209205627, + "electricity_val_precision": 0.8274524807929993, + "electricity_val_recall": 0.8279982209205627, + "electricity_val_f1-score": 0.8272846937179565, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.912844717502594, + "electricity_val_score": 0.8279982209205627 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3732665479183197, + "lr": 0.0, + "electricity_test_accuracy": 0.8370058536529541, + "electricity_test_precision": 0.8366317749023438, + "electricity_test_recall": 0.8370058536529541, + "electricity_test_f1-score": 0.8367340564727783, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9129190444946289, + "electricity_test_score": 0.8370058536529541 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.3687976896762848, + "lr": 0.0, + "electricity_val_accuracy": 0.829801619052887, + "electricity_val_precision": 0.8292626142501831, + "electricity_val_recall": 0.829801619052887, + "electricity_val_f1-score": 0.8292050361633301, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9137072563171387, + "electricity_val_score": 0.829801619052887 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.37557747960090637, + "lr": 0.0, + "electricity_test_accuracy": 0.8304728269577026, + "electricity_test_precision": 0.8299964070320129, + "electricity_test_recall": 0.8304728269577026, + "electricity_test_f1-score": 0.8300773501396179, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9117783904075623, + "electricity_test_score": 0.8304728269577026 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.3734967529773712, + "std": 0.0014520044828179868, + "n": 5, + "ci95": 0.001802612659750014 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.8340383172035217, + "std": 0.0023997884744270666, + "n": 5, + "ci95": 0.0029792532570759745 + }, + "electricity_test_precision": { + "mean": 0.833601975440979, + "std": 0.0024359075058381623, + "n": 5, + "ci95": 0.003024093768279624 + }, + "electricity_test_recall": { + "mean": 0.8340383172035217, + "std": 0.0023997884744270666, + "n": 5, + "ci95": 0.0029792532570759745 + }, + "electricity_test_f1-score": { + "mean": 0.8336644291877746, + "std": 0.0024441516813475763, + "n": 5, + "ci95": 0.003034328623142902 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.9127190113067627, + "std": 0.0007068629669606207, + "n": 5, + "ci95": 0.0008775455916491105 + }, + "electricity_test_score": { + "mean": 0.8340383172035217, + "std": 0.0023997884744270666, + "n": 5, + "ci95": 0.0029792532570759745 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/downstream_results/fttransformer_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/downstream_results/fttransformer_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..2205b9089002f7ca04369d44c796f3fb31227edc --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/downstream_results/fttransformer_electricity_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "electricity", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpqwnd5anw", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpqwnd5anw/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.4253382086753845, + "lr": 0.0, + "electricity_val_accuracy": 0.8032010793685913, + "electricity_val_precision": 0.8025455474853516, + "electricity_val_recall": 0.8032010793685913, + "electricity_val_f1-score": 0.8018348813056946, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8801766633987427, + "electricity_val_score": 0.8032010793685913 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4293154776096344, + "lr": 0.0, + "electricity_test_accuracy": 0.8034547567367554, + "electricity_test_precision": 0.802844762802124, + "electricity_test_recall": 0.8034547567367554, + "electricity_test_f1-score": 0.8020485639572144, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8774914741516113, + "electricity_test_score": 0.8034547567367554 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.4262220859527588, + "lr": 0.0, + "electricity_val_accuracy": 0.8011722564697266, + "electricity_val_precision": 0.8005000352859497, + "electricity_val_recall": 0.8011722564697266, + "electricity_val_f1-score": 0.7997601628303528, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8806599974632263, + "electricity_val_score": 0.8011722564697266 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4318920969963074, + "lr": 0.0, + "electricity_test_accuracy": 0.7970324158668518, + "electricity_test_precision": 0.7962751984596252, + "electricity_test_recall": 0.7970324158668518, + "electricity_test_f1-score": 0.7956543564796448, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8763056397438049, + "electricity_test_score": 0.7970324158668518 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.42772355675697327, + "lr": 0.0, + "electricity_val_accuracy": 0.798692524433136, + "electricity_val_precision": 0.7980766892433167, + "electricity_val_recall": 0.798692524433136, + "electricity_val_f1-score": 0.7970987558364868, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8773455023765564, + "electricity_val_score": 0.798692524433136 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.43564078211784363, + "lr": 0.0, + "electricity_test_accuracy": 0.7938212752342224, + "electricity_test_precision": 0.7933134436607361, + "electricity_test_recall": 0.7938212752342224, + "electricity_test_f1-score": 0.7919358015060425, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8736765384674072, + "electricity_test_score": 0.7938212752342224 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.43353602290153503, + "lr": 0.0, + "electricity_val_accuracy": 0.7984670996665955, + "electricity_val_precision": 0.7988982200622559, + "electricity_val_recall": 0.7984670996665955, + "electricity_val_f1-score": 0.7958927154541016, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8743888139724731, + "electricity_val_score": 0.7984670996665955 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4379521310329437, + "lr": 0.0, + "electricity_test_accuracy": 0.7916066646575928, + "electricity_test_precision": 0.7918879985809326, + "electricity_test_recall": 0.7916066646575928, + "electricity_test_f1-score": 0.7889108061790466, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8720296621322632, + "electricity_test_score": 0.7916066646575928 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.3902754783630371, + "lr": 0.0, + "electricity_val_accuracy": 0.8237150311470032, + "electricity_val_precision": 0.823143482208252, + "electricity_val_recall": 0.8237150311470032, + "electricity_val_f1-score": 0.822908878326416, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9009010195732117, + "electricity_val_score": 0.8237150311470032 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3970939815044403, + "lr": 0.0, + "electricity_test_accuracy": 0.8150813579559326, + "electricity_test_precision": 0.8144314885139465, + "electricity_test_recall": 0.8150813579559326, + "electricity_test_f1-score": 0.8142136931419373, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8975225687026978, + "electricity_test_score": 0.8150813579559326 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.4263788938522339, + "std": 0.01670589280295734, + "n": 5, + "ci95": 0.020739780224778043 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.800199294090271, + "std": 0.00944213144265759, + "n": 5, + "ci95": 0.011722075155391795 + }, + "electricity_test_precision": { + "mean": 0.7997505784034729, + "std": 0.009225484129311703, + "n": 5, + "ci95": 0.011453115111287665 + }, + "electricity_test_recall": { + "mean": 0.800199294090271, + "std": 0.00944213144265759, + "n": 5, + "ci95": 0.011722075155391795 + }, + "electricity_test_f1-score": { + "mean": 0.798552644252777, + "std": 0.010032916179749019, + "n": 5, + "ci95": 0.012455513694232444 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8794051766395569, + "std": 0.010352835250606068, + "n": 5, + "ci95": 0.012852682004693278 + }, + "electricity_test_score": { + "mean": 0.800199294090271, + "std": 0.00944213144265759, + "n": 5, + "ci95": 0.011722075155391795 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/downstream_results/mlp_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/downstream_results/mlp_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..a5eeddb01d0d2e5700ac386dc6acfeb829f45b8a --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/downstream_results/mlp_electricity_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "electricity", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmphgnj9f2k", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmphgnj9f2k/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.45377448201179504, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7842651009559631, + "electricity_val_precision": 0.7901862859725952, + "electricity_val_recall": 0.7842651009559631, + "electricity_val_f1-score": 0.7781946659088135, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8673897981643677, + "electricity_val_score": 0.7842651009559631 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4639207720756531, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7785406112670898, + "electricity_test_precision": 0.783967912197113, + "electricity_test_recall": 0.7785406112670898, + "electricity_test_f1-score": 0.7723544239997864, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8621642589569092, + "electricity_test_score": 0.7785406112670898 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.4568510353565216, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7824617028236389, + "electricity_val_precision": 0.7891604900360107, + "electricity_val_recall": 0.7824617028236389, + "electricity_val_f1-score": 0.7759427428245544, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8643500804901123, + "electricity_val_score": 0.7824617028236389 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.46997353434562683, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.775218665599823, + "electricity_test_precision": 0.781640887260437, + "electricity_test_recall": 0.775218665599823, + "electricity_test_f1-score": 0.7683533430099487, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8589786887168884, + "electricity_test_score": 0.775218665599823 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.446729838848114, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7880973815917969, + "electricity_val_precision": 0.7878503799438477, + "electricity_val_recall": 0.7880973815917969, + "electricity_val_f1-score": 0.7856864333152771, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8661814332008362, + "electricity_val_score": 0.7880973815917969 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.45572444796562195, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7847414612770081, + "electricity_test_precision": 0.7845763564109802, + "electricity_test_recall": 0.7847414612770081, + "electricity_test_f1-score": 0.7821466326713562, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8619311451911926, + "electricity_test_score": 0.7847414612770081 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.45924338698387146, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7840396761894226, + "electricity_val_precision": 0.7830287218093872, + "electricity_val_recall": 0.7840396761894226, + "electricity_val_f1-score": 0.782926619052887, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8588030934333801, + "electricity_val_score": 0.7840396761894226 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4670161306858063, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7766581773757935, + "electricity_test_precision": 0.7755798697471619, + "electricity_test_recall": 0.7766581773757935, + "electricity_test_f1-score": 0.7755626440048218, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8529821634292603, + "electricity_test_score": 0.7766581773757935 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.47163620591163635, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7790802717208862, + "electricity_val_precision": 0.787118673324585, + "electricity_val_recall": 0.7790802717208862, + "electricity_val_f1-score": 0.771767258644104, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8560341596603394, + "electricity_val_score": 0.7790802717208862 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.48404136300086975, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7685748934745789, + "electricity_test_precision": 0.7760655283927917, + "electricity_test_recall": 0.7685748934745789, + "electricity_test_f1-score": 0.7607215046882629, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8485464453697205, + "electricity_test_score": 0.7685748934745789 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.4681352496147156, + "std": 0.010361297933687471, + "n": 5, + "ci95": 0.012863188129046547 + }, + "lr": { + "mean": 0.0009999994654208422, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.7767467617988586, + "std": 0.005836425977502845, + "n": 5, + "ci95": 0.00724571823244109 + }, + "electricity_test_precision": { + "mean": 0.7803661108016968, + "std": 0.004293230394765992, + "n": 5, + "ci95": 0.005329895019200736 + }, + "electricity_test_recall": { + "mean": 0.7767467617988586, + "std": 0.005836425977502845, + "n": 5, + "ci95": 0.00724571823244109 + }, + "electricity_test_f1-score": { + "mean": 0.7718277096748352, + "std": 0.008001973930902734, + "n": 5, + "ci95": 0.009934170094874979 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8569205403327942, + "std": 0.00596814460095666, + "n": 5, + "ci95": 0.0074092422855499915 + }, + "electricity_test_score": { + "mean": 0.7767467617988586, + "std": 0.005836425977502845, + "n": 5, + "ci95": 0.00724571823244109 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/downstream_results/resnet_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/downstream_results/resnet_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..8f463ed1fb1dc97b962108d21842e23e65a633c1 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/downstream_results/resnet_electricity_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "electricity", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmppeg3nxof", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmppeg3nxof/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.4003864824771881, + "lr": 0.0, + "electricity_val_accuracy": 0.8151487708091736, + "electricity_val_precision": 0.8146806955337524, + "electricity_val_recall": 0.8151487708091736, + "electricity_val_f1-score": 0.8148254752159119, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8960868716239929, + "electricity_val_score": 0.8151487708091736 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.40950292348861694, + "lr": 0.0, + "electricity_test_accuracy": 0.8135311603546143, + "electricity_test_precision": 0.8133290410041809, + "electricity_test_recall": 0.8135311603546143, + "electricity_test_f1-score": 0.8134183883666992, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8915355801582336, + "electricity_test_score": 0.8135311603546143 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.40759947896003723, + "lr": 0.0, + "electricity_val_accuracy": 0.8095130920410156, + "electricity_val_precision": 0.8156034350395203, + "electricity_val_recall": 0.8095130920410156, + "electricity_val_f1-score": 0.8105146288871765, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9005982279777527, + "electricity_val_score": 0.8095130920410156 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.41420847177505493, + "lr": 0.0, + "electricity_test_accuracy": 0.8124238848686218, + "electricity_test_precision": 0.8188508152961731, + "electricity_test_recall": 0.8124238848686218, + "electricity_test_f1-score": 0.8134207129478455, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8979076147079468, + "electricity_test_score": 0.8124238848686218 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.39245733618736267, + "lr": 0.0, + "electricity_val_accuracy": 0.8196573257446289, + "electricity_val_precision": 0.8201331496238708, + "electricity_val_recall": 0.8196573257446289, + "electricity_val_f1-score": 0.8198521733283997, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9018487930297852, + "electricity_val_score": 0.8196573257446289 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3950406312942505, + "lr": 0.0, + "electricity_test_accuracy": 0.8216144442558289, + "electricity_test_precision": 0.8226320147514343, + "electricity_test_recall": 0.8216144442558289, + "electricity_test_f1-score": 0.8219646215438843, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9012686610221863, + "electricity_test_score": 0.8216144442558289 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.39851632714271545, + "lr": 0.0, + "electricity_val_accuracy": 0.808836817741394, + "electricity_val_precision": 0.8083773255348206, + "electricity_val_recall": 0.808836817741394, + "electricity_val_f1-score": 0.8085333108901978, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8958708643913269, + "electricity_val_score": 0.808836817741394 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4052407741546631, + "lr": 0.0, + "electricity_test_accuracy": 0.8145277500152588, + "electricity_test_precision": 0.8143652677536011, + "electricity_test_recall": 0.8145277500152588, + "electricity_test_f1-score": 0.8144391179084778, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8931828737258911, + "electricity_test_score": 0.8145277500152588 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.3846282958984375, + "lr": 0.0, + "electricity_val_accuracy": 0.8198827505111694, + "electricity_val_precision": 0.8193657398223877, + "electricity_val_recall": 0.8198827505111694, + "electricity_val_f1-score": 0.8194862008094788, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9040712714195251, + "electricity_val_score": 0.8198827505111694 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3941192328929901, + "lr": 0.0, + "electricity_test_accuracy": 0.8197320103645325, + "electricity_test_precision": 0.8194493651390076, + "electricity_test_recall": 0.8197320103645325, + "electricity_test_f1-score": 0.819563090801239, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8995412588119507, + "electricity_test_score": 0.8197320103645325 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.4036224067211151, + "std": 0.00884903308579223, + "n": 5, + "ci95": 0.010985764338714761 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.8163658499717712, + "std": 0.004056844318513874, + "n": 5, + "ci95": 0.005036429992967676 + }, + "electricity_test_precision": { + "mean": 0.8177253007888794, + "std": 0.003838357226489016, + "n": 5, + "ci95": 0.004765185928134204 + }, + "electricity_test_recall": { + "mean": 0.8163658499717712, + "std": 0.004056844318513874, + "n": 5, + "ci95": 0.005036429992967676 + }, + "electricity_test_f1-score": { + "mean": 0.8165611863136292, + "std": 0.003951312093876537, + "n": 5, + "ci95": 0.004905415435923301 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8966871976852417, + "std": 0.004166663487993788, + "n": 5, + "ci95": 0.005172766641738574 + }, + "electricity_test_score": { + "mean": 0.8163658499717712, + "std": 0.004056844318513874, + "n": 5, + "ci95": 0.005036429992967676 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..92db635ad3b60314c48a5902e435d5bc72f62432 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6e50a71be4f826e6aa5792366b091e7f5262a3a0c9b646df0688dae100b04585 +size 92799104 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..df078589942f675860e26cc4fdcdd27647c41a83 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a34c89c118819c2802944d33c90a279a78bb2c834bd007303d71f41086222b07 +size 8332910 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3788bf33925cd8931e59a0b02f66ce094204f312 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.00025_sx5e-05_sy1e-05/seed_42/wjepa_deterministic/electricity/20260915_221516/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: electricity +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.0005 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.6 +mask_max_trgt_share: 0.9 +mask_min_ctx_share: 0.15 +mask_min_trgt_share: 0.15 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 64 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 6 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 128 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.001 +recon_y_weight: 0.25 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 5.0e-05 +sw_s_y_weight: 1.0e-05 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 40 +vtjepa_kl_x_end_weight: 1.0e-06 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-06 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 40 +vtjepa_kl_z_end_weight: 1.0e-05 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 40 +wjepa_joint_end_weight: 0.00025 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..7ac48848590539077a923251c30adf9e0304df0f --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-15T22:53:55.609189+00:00", + "spec": { + "dataset": "electricity", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.0001, + "sw_z_weight": 0.0, + "sw_s_y_weight": 2e-05, + "lambda_joint": 0.0005, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..4809bdeccfef53c9e391ba0996fab30e0489ac78 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "electricity", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.0001, + "sw_z_weight": 0.0, + "sw_s_y_weight": 2e-05, + "lambda_joint": 0.0005, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "electricity", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.0005", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.0001", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "2e-05", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..478da54e0e43facf5bf11005691c07f023860cc1 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "electricity", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.0005, + "s_x": 0.0001, + "z": 0.0, + "s_y": 2e-05 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 40 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "electricity", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "64", + "--model_num_layers", + "6", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.15", + "--mask_max_ctx_share", + "0.6", + "--mask_min_trgt_share", + "0.15", + "--mask_max_trgt_share", + "0.9", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "128", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.0005", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "1e-06", + "--vtjepa_kl_x_annealing_epochs", + "40", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-05", + "--vtjepa_kl_z_annealing_epochs", + "40", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-06", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.001", + "--recon_y_weight", + "0.25", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.0005", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "40", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.0001", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "2e-05", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/downstream_results/autoint_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/downstream_results/autoint_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..14b06b45baefe7f3c8e20be210ef4c1d33063758 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/downstream_results/autoint_electricity_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "electricity", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp4gag0_nu", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp4gag0_nu/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.39373689889907837, + "lr": 0.0, + "electricity_val_accuracy": 0.8196573257446289, + "electricity_val_precision": 0.819051206111908, + "electricity_val_recall": 0.8196573257446289, + "electricity_val_f1-score": 0.8187980651855469, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9000295996665955, + "electricity_val_score": 0.8196573257446289 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.390974760055542, + "lr": 0.0, + "electricity_test_accuracy": 0.8254899978637695, + "electricity_test_precision": 0.824923038482666, + "electricity_test_recall": 0.8254899978637695, + "electricity_test_f1-score": 0.8248078227043152, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9010804295539856, + "electricity_test_score": 0.8254899978637695 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.39350008964538574, + "lr": 0.0, + "electricity_val_accuracy": 0.8165013790130615, + "electricity_val_precision": 0.8160676956176758, + "electricity_val_recall": 0.8165013790130615, + "electricity_val_f1-score": 0.8162100315093994, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.900027871131897, + "electricity_val_score": 0.8165013790130615 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.39516741037368774, + "lr": 0.0, + "electricity_test_accuracy": 0.8172959685325623, + "electricity_test_precision": 0.8168964982032776, + "electricity_test_recall": 0.8172959685325623, + "electricity_test_f1-score": 0.8170342445373535, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8986459374427795, + "electricity_test_score": 0.8172959685325623 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.40386682748794556, + "lr": 0.0, + "electricity_val_accuracy": 0.814472496509552, + "electricity_val_precision": 0.8138009309768677, + "electricity_val_recall": 0.814472496509552, + "electricity_val_f1-score": 0.8136505484580994, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8940901160240173, + "electricity_val_score": 0.814472496509552 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.40309640765190125, + "lr": 0.0, + "electricity_test_accuracy": 0.8185139894485474, + "electricity_test_precision": 0.8178873062133789, + "electricity_test_recall": 0.8185139894485474, + "electricity_test_f1-score": 0.8178412318229675, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8946102857589722, + "electricity_test_score": 0.8185139894485474 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.38110798597335815, + "lr": 0.0, + "electricity_val_accuracy": 0.8270964622497559, + "electricity_val_precision": 0.8265622854232788, + "electricity_val_recall": 0.8270964622497559, + "electricity_val_f1-score": 0.8266127705574036, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9078460335731506, + "electricity_val_score": 0.8270964622497559 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.38191449642181396, + "lr": 0.0, + "electricity_test_accuracy": 0.8247148990631104, + "electricity_test_precision": 0.824160099029541, + "electricity_test_recall": 0.8247148990631104, + "electricity_test_f1-score": 0.8241954445838928, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9063522219657898, + "electricity_test_score": 0.8247148990631104 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.38734662532806396, + "lr": 0.0, + "electricity_val_accuracy": 0.8221370577812195, + "electricity_val_precision": 0.8215556740760803, + "electricity_val_recall": 0.8221370577812195, + "electricity_val_f1-score": 0.8212981224060059, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9037337899208069, + "electricity_val_score": 0.8221370577812195 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3867725729942322, + "lr": 0.0, + "electricity_test_accuracy": 0.822389543056488, + "electricity_test_precision": 0.8217951059341431, + "electricity_test_recall": 0.822389543056488, + "electricity_test_f1-score": 0.8217272162437439, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9032497406005859, + "electricity_test_score": 0.822389543056488 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.3915851294994354, + "std": 0.008099331084436388, + "n": 5, + "ci95": 0.010055035587753067 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.8216808795928955, + "std": 0.0036562787860467618, + "n": 5, + "ci95": 0.004539141927793546 + }, + "electricity_test_precision": { + "mean": 0.8211324095726014, + "std": 0.0036210695377845675, + "n": 5, + "ci95": 0.004495430880473268 + }, + "electricity_test_recall": { + "mean": 0.8216808795928955, + "std": 0.0036562787860467618, + "n": 5, + "ci95": 0.004539141927793546 + }, + "electricity_test_f1-score": { + "mean": 0.8211211919784546, + "std": 0.003566189093729045, + "n": 5, + "ci95": 0.004427298733225903 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.9007877230644226, + "std": 0.0044681884688476505, + "n": 5, + "ci95": 0.005547099334336871 + }, + "electricity_test_score": { + "mean": 0.8216808795928955, + "std": 0.0036562787860467618, + "n": 5, + "ci95": 0.004539141927793546 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/downstream_results/dcnv2_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/downstream_results/dcnv2_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..5f3be40acc690661a3f7a277142f699d27570a15 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/downstream_results/dcnv2_electricity_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "electricity", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpx1fp850q", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpx1fp850q/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.3701808750629425, + "lr": 0.0, + "electricity_val_accuracy": 0.8338593244552612, + "electricity_val_precision": 0.8334012031555176, + "electricity_val_recall": 0.8338593244552612, + "electricity_val_f1-score": 0.833479642868042, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9127412438392639, + "electricity_val_score": 0.8338593244552612 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.37670472264289856, + "lr": 0.0, + "electricity_test_accuracy": 0.8281474709510803, + "electricity_test_precision": 0.8276653289794922, + "electricity_test_recall": 0.8281474709510803, + "electricity_test_f1-score": 0.8277571797370911, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9110413789749146, + "electricity_test_score": 0.8281474709510803 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.37654322385787964, + "lr": 0.0, + "electricity_val_accuracy": 0.8288999199867249, + "electricity_val_precision": 0.828376829624176, + "electricity_val_recall": 0.8288999199867249, + "electricity_val_f1-score": 0.8284212350845337, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9096097946166992, + "electricity_val_score": 0.8288999199867249 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3761906623840332, + "lr": 0.0, + "electricity_test_accuracy": 0.8295869827270508, + "electricity_test_precision": 0.8290987610816956, + "electricity_test_recall": 0.8295869827270508, + "electricity_test_f1-score": 0.8291753530502319, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9105860590934753, + "electricity_test_score": 0.8295869827270508 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.37085652351379395, + "lr": 0.0, + "electricity_val_accuracy": 0.8293507695198059, + "electricity_val_precision": 0.828810453414917, + "electricity_val_recall": 0.8293507695198059, + "electricity_val_f1-score": 0.8287832736968994, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9130625128746033, + "electricity_val_score": 0.8293507695198059 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3718201518058777, + "lr": 0.0, + "electricity_test_accuracy": 0.8334625363349915, + "electricity_test_precision": 0.8330815434455872, + "electricity_test_recall": 0.8334625363349915, + "electricity_test_f1-score": 0.8331913948059082, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9130263328552246, + "electricity_test_score": 0.8334625363349915 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.3730820119380951, + "lr": 0.0, + "electricity_val_accuracy": 0.8288999199867249, + "electricity_val_precision": 0.8284521102905273, + "electricity_val_recall": 0.8288999199867249, + "electricity_val_f1-score": 0.8279758095741272, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9108537435531616, + "electricity_val_score": 0.8288999199867249 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3754652142524719, + "lr": 0.0, + "electricity_test_accuracy": 0.8316908478736877, + "electricity_test_precision": 0.8311728239059448, + "electricity_test_recall": 0.8316908478736877, + "electricity_test_f1-score": 0.8311228156089783, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9103821516036987, + "electricity_test_score": 0.8316908478736877 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.3717864751815796, + "lr": 0.0, + "electricity_val_accuracy": 0.8250676393508911, + "electricity_val_precision": 0.8244907259941101, + "electricity_val_recall": 0.8250676393508911, + "electricity_val_f1-score": 0.8244465589523315, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9118346571922302, + "electricity_val_score": 0.8250676393508911 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.37519994378089905, + "lr": 0.0, + "electricity_test_accuracy": 0.8324659466743469, + "electricity_test_precision": 0.8319783210754395, + "electricity_test_recall": 0.8324659466743469, + "electricity_test_f1-score": 0.8320263624191284, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.911387026309967, + "electricity_test_score": 0.8324659466743469 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.3750761389732361, + "std": 0.0019145891752293503, + "n": 5, + "ci95": 0.0023768953376718956 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.8310707569122314, + "std": 0.0021678827645424947, + "n": 5, + "ci95": 0.0026913504486115432 + }, + "electricity_test_precision": { + "mean": 0.8305993556976319, + "std": 0.0021938530203372557, + "n": 5, + "ci95": 0.002723591610692342 + }, + "electricity_test_recall": { + "mean": 0.8310707569122314, + "std": 0.0021678827645424947, + "n": 5, + "ci95": 0.0026913504486115432 + }, + "electricity_test_f1-score": { + "mean": 0.8306546211242676, + "std": 0.0021862926150443027, + "n": 5, + "ci95": 0.0027142056325805746 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.9112845897674561, + "std": 0.00104948242726325, + "n": 5, + "ci95": 0.0013028956397561288 + }, + "electricity_test_score": { + "mean": 0.8310707569122314, + "std": 0.0021678827645424947, + "n": 5, + "ci95": 0.0026913504486115432 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/downstream_results/fttransformer_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/downstream_results/fttransformer_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..58574ae920508b71b5c3c857c2bf40cfb353184f --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/downstream_results/fttransformer_electricity_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "electricity", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp4z0c02xe", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp4z0c02xe/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.42955446243286133, + "lr": 0.0, + "electricity_val_accuracy": 0.799143373966217, + "electricity_val_precision": 0.7986562848091125, + "electricity_val_recall": 0.799143373966217, + "electricity_val_f1-score": 0.7973957061767578, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8775113224983215, + "electricity_val_score": 0.799143373966217 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4330693483352661, + "lr": 0.0, + "electricity_test_accuracy": 0.7976968288421631, + "electricity_test_precision": 0.7974163293838501, + "electricity_test_recall": 0.7976968288421631, + "electricity_test_f1-score": 0.7957019209861755, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8755244612693787, + "electricity_test_score": 0.7976968288421631 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.4367830753326416, + "lr": 0.0, + "electricity_val_accuracy": 0.7899008393287659, + "electricity_val_precision": 0.7893691658973694, + "electricity_val_recall": 0.7899008393287659, + "electricity_val_f1-score": 0.7878661751747131, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.871829092502594, + "electricity_val_score": 0.7899008393287659 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.44203898310661316, + "lr": 0.0, + "electricity_test_accuracy": 0.7901672124862671, + "electricity_test_precision": 0.789589524269104, + "electricity_test_recall": 0.7901672124862671, + "electricity_test_f1-score": 0.7882304787635803, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8697992563247681, + "electricity_test_score": 0.7901672124862671 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.4398593008518219, + "lr": 0.0, + "electricity_val_accuracy": 0.7858430743217468, + "electricity_val_precision": 0.7855836749076843, + "electricity_val_recall": 0.7858430743217468, + "electricity_val_f1-score": 0.7833530306816101, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8697785139083862, + "electricity_val_score": 0.7858430743217468 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4445708096027374, + "lr": 0.0, + "electricity_test_accuracy": 0.7869560122489929, + "electricity_test_precision": 0.7868457436561584, + "electricity_test_recall": 0.7869560122489929, + "electricity_test_f1-score": 0.7844011187553406, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.867475152015686, + "electricity_test_score": 0.7869560122489929 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.4355636239051819, + "lr": 0.0, + "electricity_val_accuracy": 0.7946348190307617, + "electricity_val_precision": 0.7942522764205933, + "electricity_val_recall": 0.7946348190307617, + "electricity_val_f1-score": 0.792609691619873, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8727938532829285, + "electricity_val_score": 0.7946348190307617 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.44228360056877136, + "lr": 0.0, + "electricity_test_accuracy": 0.7886170148849487, + "electricity_test_precision": 0.7883430123329163, + "electricity_test_recall": 0.7886170148849487, + "electricity_test_f1-score": 0.786282479763031, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8692082166671753, + "electricity_test_score": 0.7886170148849487 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.4153716266155243, + "lr": 0.0, + "electricity_val_accuracy": 0.8090622425079346, + "electricity_val_precision": 0.8084056377410889, + "electricity_val_recall": 0.8090622425079346, + "electricity_val_f1-score": 0.8079156279563904, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.886652410030365, + "electricity_val_score": 0.8090622425079346 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4183814227581024, + "lr": 0.0, + "electricity_test_accuracy": 0.8083269000053406, + "electricity_test_precision": 0.8076409101486206, + "electricity_test_recall": 0.8083269000053406, + "electricity_test_f1-score": 0.8072571754455566, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8844841718673706, + "electricity_test_score": 0.8083269000053406 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.4360688328742981, + "std": 0.010821171652159452, + "n": 5, + "ci95": 0.013434105227866429 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.7943527936935425, + "std": 0.008825939647013344, + "n": 5, + "ci95": 0.010957094644101151 + }, + "electricity_test_precision": { + "mean": 0.7939671039581299, + "std": 0.008665701282435281, + "n": 5, + "ci95": 0.010758164332257025 + }, + "electricity_test_recall": { + "mean": 0.7943527936935425, + "std": 0.008825939647013344, + "n": 5, + "ci95": 0.010957094644101151 + }, + "electricity_test_f1-score": { + "mean": 0.7923746347427368, + "std": 0.009359836934937242, + "n": 5, + "ci95": 0.011619909409211253 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8732982516288758, + "std": 0.006945514095023255, + "n": 5, + "ci95": 0.008622612246942017 + }, + "electricity_test_score": { + "mean": 0.7943527936935425, + "std": 0.008825939647013344, + "n": 5, + "ci95": 0.010957094644101151 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/downstream_results/mlp_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/downstream_results/mlp_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..85791788ad06a539dd1777664a20dd6d1080167b --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/downstream_results/mlp_electricity_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "electricity", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp3owtuidi", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp3owtuidi/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.4659750759601593, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7759242653846741, + "electricity_val_precision": 0.774796187877655, + "electricity_val_recall": 0.7759242653846741, + "electricity_val_f1-score": 0.7745442390441895, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8554174900054932, + "electricity_val_score": 0.7759242653846741 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4736913740634918, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7727826237678528, + "electricity_test_precision": 0.7716465592384338, + "electricity_test_recall": 0.7727826237678528, + "electricity_test_f1-score": 0.7711271047592163, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8495428562164307, + "electricity_test_score": 0.7727826237678528 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.4483082592487335, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7822362780570984, + "electricity_val_precision": 0.7811973094940186, + "electricity_val_recall": 0.7822362780570984, + "electricity_val_f1-score": 0.7809396982192993, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8649780750274658, + "electricity_val_score": 0.7822362780570984 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4587436616420746, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.779758632183075, + "electricity_test_precision": 0.778696596622467, + "electricity_test_recall": 0.779758632183075, + "electricity_test_f1-score": 0.7784860730171204, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8592883348464966, + "electricity_test_score": 0.779758632183075 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.4523443579673767, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7838142514228821, + "electricity_val_precision": 0.7847293019294739, + "electricity_val_recall": 0.7838142514228821, + "electricity_val_f1-score": 0.7803041338920593, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8643068671226501, + "electricity_val_score": 0.7838142514228821 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.465254008769989, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.778983473777771, + "electricity_test_precision": 0.7795163989067078, + "electricity_test_recall": 0.778983473777771, + "electricity_test_f1-score": 0.7755544781684875, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8566256165504456, + "electricity_test_score": 0.778983473777771 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.6022996306419373, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7844905257225037, + "electricity_val_precision": 0.7835615873336792, + "electricity_val_recall": 0.7844905257225037, + "electricity_val_f1-score": 0.7836664319038391, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.86118483543396, + "electricity_val_score": 0.7844905257225037 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.5377166867256165, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7831912040710449, + "electricity_test_precision": 0.7822110056877136, + "electricity_test_recall": 0.7831912040710449, + "electricity_test_f1-score": 0.7822346091270447, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8566612005233765, + "electricity_test_score": 0.7831912040710449 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.45763084292411804, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7793056964874268, + "electricity_val_precision": 0.7782553434371948, + "electricity_val_recall": 0.7793056964874268, + "electricity_val_f1-score": 0.7777500748634338, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8597719073295593, + "electricity_val_score": 0.7793056964874268 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4718836843967438, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7695714831352234, + "electricity_test_precision": 0.7684947848320007, + "electricity_test_recall": 0.7695714831352234, + "electricity_test_f1-score": 0.7674967646598816, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8503607511520386, + "electricity_test_score": 0.7695714831352234 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.48145788311958315, + "std": 0.031998196545085705, + "n": 5, + "ci95": 0.03972463918940329 + }, + "lr": { + "mean": 0.0009999994654208422, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.7768574833869935, + "std": 0.005539467433038164, + "n": 5, + "ci95": 0.00687705461052576 + }, + "electricity_test_precision": { + "mean": 0.7761130690574646, + "std": 0.005775596601041739, + "n": 5, + "ci95": 0.007170200694175172 + }, + "electricity_test_recall": { + "mean": 0.7768574833869935, + "std": 0.005539467433038164, + "n": 5, + "ci95": 0.00687705461052576 + }, + "electricity_test_f1-score": { + "mean": 0.7749798059463501, + "std": 0.005832999645495764, + "n": 5, + "ci95": 0.007241464561377702 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8544957518577576, + "std": 0.004296037398193589, + "n": 5, + "ci95": 0.0053333798155456675 + }, + "electricity_test_score": { + "mean": 0.7768574833869935, + "std": 0.005539467433038164, + "n": 5, + "ci95": 0.00687705461052576 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/downstream_results/resnet_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/downstream_results/resnet_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..de55519b35e576d2f9708af4701e0f7d7bc6c059 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/downstream_results/resnet_electricity_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "electricity", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpbg1sjts0", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpbg1sjts0/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.4188918471336365, + "lr": 0.0, + "electricity_val_accuracy": 0.8081604838371277, + "electricity_val_precision": 0.8074683547019958, + "electricity_val_recall": 0.8081604838371277, + "electricity_val_f1-score": 0.8070668578147888, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8845577836036682, + "electricity_val_score": 0.8081604838371277 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4219503402709961, + "lr": 0.0, + "electricity_test_accuracy": 0.8006865382194519, + "electricity_test_precision": 0.7998826503753662, + "electricity_test_recall": 0.8006865382194519, + "electricity_test_f1-score": 0.7997978329658508, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.882916271686554, + "electricity_test_score": 0.8006865382194519 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.42871996760368347, + "lr": 0.0, + "electricity_val_accuracy": 0.8004959225654602, + "electricity_val_precision": 0.8098951578140259, + "electricity_val_recall": 0.8004959225654602, + "electricity_val_f1-score": 0.801662266254425, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8947359323501587, + "electricity_val_score": 0.8004959225654602 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4322846233844757, + "lr": 0.0, + "electricity_test_accuracy": 0.7993577718734741, + "electricity_test_precision": 0.8101338744163513, + "electricity_test_recall": 0.7993577718734741, + "electricity_test_f1-score": 0.8005353808403015, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8932556509971619, + "electricity_test_score": 0.7993577718734741 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.39474326372146606, + "lr": 0.0, + "electricity_val_accuracy": 0.8133453726768494, + "electricity_val_precision": 0.814716100692749, + "electricity_val_recall": 0.8133453726768494, + "electricity_val_f1-score": 0.813790500164032, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8999252915382385, + "electricity_val_score": 0.8133453726768494 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3961979150772095, + "lr": 0.0, + "electricity_test_accuracy": 0.8208393454551697, + "electricity_test_precision": 0.8222321271896362, + "electricity_test_recall": 0.8208393454551697, + "electricity_test_f1-score": 0.8212764859199524, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9000495672225952, + "electricity_test_score": 0.8208393454551697 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.39740923047065735, + "lr": 0.0, + "electricity_val_accuracy": 0.814021646976471, + "electricity_val_precision": 0.8133848309516907, + "electricity_val_recall": 0.814021646976471, + "electricity_val_f1-score": 0.8134526610374451, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8970203995704651, + "electricity_val_score": 0.814021646976471 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4050596058368683, + "lr": 0.0, + "electricity_test_accuracy": 0.8119809627532959, + "electricity_test_precision": 0.811488926410675, + "electricity_test_recall": 0.8119809627532959, + "electricity_test_f1-score": 0.8116375803947449, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.893227756023407, + "electricity_test_score": 0.8119809627532959 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.3888181746006012, + "lr": 0.0, + "electricity_val_accuracy": 0.8171776533126831, + "electricity_val_precision": 0.8165359497070312, + "electricity_val_recall": 0.8171776533126831, + "electricity_val_f1-score": 0.8163502812385559, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9006646275520325, + "electricity_val_score": 0.8171776533126831 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.39454686641693115, + "lr": 0.0, + "electricity_test_accuracy": 0.8209500312805176, + "electricity_test_precision": 0.8203457593917847, + "electricity_test_recall": 0.8209500312805176, + "electricity_test_f1-score": 0.8203102350234985, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8985642194747925, + "electricity_test_score": 0.8209500312805176 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.4100078701972961, + "std": 0.01653117617200666, + "n": 5, + "ci95": 0.02052287565282429 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.8107629299163819, + "std": 0.010469074552210468, + "n": 5, + "ci95": 0.012996989022414007 + }, + "electricity_test_precision": { + "mean": 0.8128166675567627, + "std": 0.00896642598798955, + "n": 5, + "ci95": 0.01113150351112764 + }, + "electricity_test_recall": { + "mean": 0.8107629299163819, + "std": 0.010469074552210468, + "n": 5, + "ci95": 0.012996989022414007 + }, + "electricity_test_f1-score": { + "mean": 0.8107115030288696, + "std": 0.010335301104978408, + "n": 5, + "ci95": 0.012830913977624259 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8936026930809021, + "std": 0.006720110649733638, + "n": 5, + "ci95": 0.00834278177201003 + }, + "electricity_test_score": { + "mean": 0.8107629299163819, + "std": 0.010469074552210468, + "n": 5, + "ci95": 0.012996989022414007 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..4e6a6c77cb6bff02ab27810d5c6ff7768e8536a7 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dc49c4a8773379eac7ac977f14243a21c008f44fcbb7774a81c5839b52754fd5 +size 92799104 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..70148db19e188ec0a99658009b1e9277f1c67f4c --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c78d282cd57e2dbc6ec99ea3ebb07b78b3767b16908d8611581517cc9511f876 +size 8331758 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4060b14fb544f4e236b7b153ef18be7a6ef8b12e --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.0005_sx0.0001_sy2e-05/seed_42/wjepa_deterministic/electricity/20260915_214426/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: electricity +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.0005 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.6 +mask_max_trgt_share: 0.9 +mask_min_ctx_share: 0.15 +mask_min_trgt_share: 0.15 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 64 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 6 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 128 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.001 +recon_y_weight: 0.25 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.0001 +sw_s_y_weight: 2.0e-05 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 40 +vtjepa_kl_x_end_weight: 1.0e-06 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-06 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 40 +vtjepa_kl_z_end_weight: 1.0e-05 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 40 +wjepa_joint_end_weight: 0.0005 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..8c7966ea75b527741ebb5fe4c709350bc5ecf9f6 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-16T01:56:44.868466+00:00", + "spec": { + "dataset": "electricity", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.0002, + "sw_z_weight": 0.0, + "sw_s_y_weight": 4e-05, + "lambda_joint": 0.001, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..8ca4ceef53848b8b1a01fc1fd40b86a45a48662a --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "electricity", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 0.0002, + "sw_z_weight": 0.0, + "sw_s_y_weight": 4e-05, + "lambda_joint": 0.001, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "electricity", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.001", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "0.0002", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "4e-05", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..959b4d73db5726481ea1b3349e943ae73bd94141 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "electricity", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.001, + "s_x": 0.0002, + "z": 0.0, + "s_y": 4e-05 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 40 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "electricity", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "64", + "--model_num_layers", + "6", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.15", + "--mask_max_ctx_share", + "0.6", + "--mask_min_trgt_share", + "0.15", + "--mask_max_trgt_share", + "0.9", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "128", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.0005", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "1e-06", + "--vtjepa_kl_x_annealing_epochs", + "40", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-05", + "--vtjepa_kl_z_annealing_epochs", + "40", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-06", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.001", + "--recon_y_weight", + "0.25", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.001", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "40", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "0.0002", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "4e-05", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/downstream_results/autoint_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/downstream_results/autoint_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..60a7c660008778f5c20a3d1485a568ad5ea5806b --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/downstream_results/autoint_electricity_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "electricity", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpefo5m6rz", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpefo5m6rz/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.404644250869751, + "lr": 0.0, + "electricity_val_accuracy": 0.8119927644729614, + "electricity_val_precision": 0.8113160133361816, + "electricity_val_recall": 0.8119927644729614, + "electricity_val_f1-score": 0.8113422393798828, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8937628269195557, + "electricity_val_score": 0.8119927644729614 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.40166807174682617, + "lr": 0.0, + "electricity_test_accuracy": 0.8158565163612366, + "electricity_test_precision": 0.8152959942817688, + "electricity_test_recall": 0.8158565163612366, + "electricity_test_f1-score": 0.8154116868972778, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8959375023841858, + "electricity_test_score": 0.8158565163612366 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.40089136362075806, + "lr": 0.0, + "electricity_val_accuracy": 0.8162758946418762, + "electricity_val_precision": 0.815701961517334, + "electricity_val_recall": 0.8162758946418762, + "electricity_val_f1-score": 0.8152099847793579, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8961867094039917, + "electricity_val_score": 0.8162758946418762 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4036189615726471, + "lr": 0.0, + "electricity_test_accuracy": 0.8144170045852661, + "electricity_test_precision": 0.813757061958313, + "electricity_test_recall": 0.8144170045852661, + "electricity_test_f1-score": 0.8135635852813721, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8944199085235596, + "electricity_test_score": 0.8144170045852661 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.40553128719329834, + "lr": 0.0, + "electricity_val_accuracy": 0.8090622425079346, + "electricity_val_precision": 0.8083894848823547, + "electricity_val_recall": 0.8090622425079346, + "electricity_val_f1-score": 0.8079544305801392, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8936381340026855, + "electricity_val_score": 0.8090622425079346 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4074324369430542, + "lr": 0.0, + "electricity_test_accuracy": 0.8143062591552734, + "electricity_test_precision": 0.8136658668518066, + "electricity_test_recall": 0.8143062591552734, + "electricity_test_f1-score": 0.8133599758148193, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8925489783287048, + "electricity_test_score": 0.8143062591552734 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.3791830539703369, + "lr": 0.0, + "electricity_val_accuracy": 0.8286744952201843, + "electricity_val_precision": 0.8281388878822327, + "electricity_val_recall": 0.8286744952201843, + "electricity_val_f1-score": 0.8281578421592712, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9078845381736755, + "electricity_val_score": 0.8286744952201843 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3818378746509552, + "lr": 0.0, + "electricity_test_accuracy": 0.8272616267204285, + "electricity_test_precision": 0.8267131447792053, + "electricity_test_recall": 0.8272616267204285, + "electricity_test_f1-score": 0.8267013430595398, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9055904150009155, + "electricity_test_score": 0.8272616267204285 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.40337681770324707, + "lr": 0.0, + "electricity_val_accuracy": 0.8110910654067993, + "electricity_val_precision": 0.8104211091995239, + "electricity_val_recall": 0.8110910654067993, + "electricity_val_f1-score": 0.8100613951683044, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8933847546577454, + "electricity_val_score": 0.8110910654067993 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4052073359489441, + "lr": 0.0, + "electricity_test_accuracy": 0.8157457709312439, + "electricity_test_precision": 0.8152078986167908, + "electricity_test_recall": 0.8157457709312439, + "electricity_test_f1-score": 0.8146214485168457, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8929359316825867, + "electricity_test_score": 0.8157457709312439 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.39995293617248534, + "std": 0.01034515025252911, + "n": 5, + "ci95": 0.012843141349008252 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.8175174355506897, + "std": 0.005494793800588441, + "n": 5, + "ci95": 0.006821593862047489 + }, + "electricity_test_precision": { + "mean": 0.8169279932975769, + "std": 0.005524210588930376, + "n": 5, + "ci95": 0.006858113773453993 + }, + "electricity_test_recall": { + "mean": 0.8175174355506897, + "std": 0.005494793800588441, + "n": 5, + "ci95": 0.006821593862047489 + }, + "electricity_test_f1-score": { + "mean": 0.8167316079139709, + "std": 0.005634599663176238, + "n": 5, + "ci95": 0.006995157939011586 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8962865471839905, + "std": 0.005370385779949807, + "n": 5, + "ci95": 0.006667145666032 + }, + "electricity_test_score": { + "mean": 0.8175174355506897, + "std": 0.005494793800588441, + "n": 5, + "ci95": 0.006821593862047489 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/downstream_results/dcnv2_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/downstream_results/dcnv2_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..b85764f1d9245b2f2e002db58ccc0157e6917151 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/downstream_results/dcnv2_electricity_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "electricity", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmph6zzw1xd", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmph6zzw1xd/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.37600094079971313, + "lr": 0.0, + "electricity_val_accuracy": 0.8261947631835938, + "electricity_val_precision": 0.8256273865699768, + "electricity_val_recall": 0.8261947631835938, + "electricity_val_f1-score": 0.8255855441093445, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9096471071243286, + "electricity_val_score": 0.8261947631835938 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.37932315468788147, + "lr": 0.0, + "electricity_test_accuracy": 0.8302513360977173, + "electricity_test_precision": 0.829736590385437, + "electricity_test_recall": 0.8302513360977173, + "electricity_test_f1-score": 0.829762876033783, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9088711738586426, + "electricity_test_score": 0.8302513360977173 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.3703933358192444, + "lr": 0.0, + "electricity_val_accuracy": 0.8288999199867249, + "electricity_val_precision": 0.8283921480178833, + "electricity_val_recall": 0.8288999199867249, + "electricity_val_f1-score": 0.8280928730964661, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9129377603530884, + "electricity_val_score": 0.8288999199867249 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3737042546272278, + "lr": 0.0, + "electricity_test_accuracy": 0.8300299048423767, + "electricity_test_precision": 0.8295083045959473, + "electricity_test_recall": 0.8300299048423767, + "electricity_test_f1-score": 0.8295189738273621, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9118359088897705, + "electricity_test_score": 0.8300299048423767 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.36995890736579895, + "lr": 0.0, + "electricity_val_accuracy": 0.8309287428855896, + "electricity_val_precision": 0.8304126858711243, + "electricity_val_recall": 0.8309287428855896, + "electricity_val_f1-score": 0.830219566822052, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.913167417049408, + "electricity_val_score": 0.8309287428855896 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.375161349773407, + "lr": 0.0, + "electricity_test_accuracy": 0.8312479257583618, + "electricity_test_precision": 0.8307352066040039, + "electricity_test_recall": 0.8312479257583618, + "electricity_test_f1-score": 0.8307442665100098, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9120092988014221, + "electricity_test_score": 0.8312479257583618 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.37768295407295227, + "lr": 0.0, + "electricity_val_accuracy": 0.8225879073143005, + "electricity_val_precision": 0.8220114707946777, + "electricity_val_recall": 0.8225879073143005, + "electricity_val_f1-score": 0.8217511177062988, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9086050987243652, + "electricity_val_score": 0.8225879073143005 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3817760944366455, + "lr": 0.0, + "electricity_test_accuracy": 0.8288118839263916, + "electricity_test_precision": 0.828272819519043, + "electricity_test_recall": 0.8288118839263916, + "electricity_test_f1-score": 0.8282416462898254, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9068576693534851, + "electricity_test_score": 0.8288118839263916 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.3673419952392578, + "lr": 0.0, + "electricity_val_accuracy": 0.8363390564918518, + "electricity_val_precision": 0.8358507752418518, + "electricity_val_recall": 0.8363390564918518, + "electricity_val_f1-score": 0.8358020782470703, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9144577980041504, + "electricity_val_score": 0.8363390564918518 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3759532570838928, + "lr": 0.0, + "electricity_test_accuracy": 0.8301406502723694, + "electricity_test_precision": 0.8298007249832153, + "electricity_test_recall": 0.8301406502723694, + "electricity_test_f1-score": 0.8299169540405273, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9119086861610413, + "electricity_test_score": 0.8301406502723694 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.3771836221218109, + "std": 0.003293090619152877, + "n": 5, + "ci95": 0.004088256551569548 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.8300963401794433, + "std": 0.0008669454977973773, + "n": 5, + "ci95": 0.0010762824413667652 + }, + "electricity_test_precision": { + "mean": 0.8296107292175293, + "std": 0.0008827346687522368, + "n": 5, + "ci95": 0.0010958841435563824 + }, + "electricity_test_recall": { + "mean": 0.8300963401794433, + "std": 0.0008669454977973773, + "n": 5, + "ci95": 0.0010762824413667652 + }, + "electricity_test_f1-score": { + "mean": 0.8296369433403015, + "std": 0.0009057041877389639, + "n": 5, + "ci95": 0.0011243999960925159 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.9102965474128724, + "std": 0.0023323654567475828, + "n": 5, + "ci95": 0.0028955499444031987 + }, + "electricity_test_score": { + "mean": 0.8300963401794433, + "std": 0.0008669454977973773, + "n": 5, + "ci95": 0.0010762824413667652 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/downstream_results/fttransformer_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/downstream_results/fttransformer_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..ba406c0515bc8673fddcc8da8607611976602084 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/downstream_results/fttransformer_electricity_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "electricity", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmphue75xkh", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmphue75xkh/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.43786969780921936, + "lr": 0.0, + "electricity_val_accuracy": 0.790802538394928, + "electricity_val_precision": 0.7899205684661865, + "electricity_val_recall": 0.790802538394928, + "electricity_val_f1-score": 0.7894054651260376, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8716440200805664, + "electricity_val_score": 0.790802538394928 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.44138848781585693, + "lr": 0.0, + "electricity_test_accuracy": 0.7913852334022522, + "electricity_test_precision": 0.7906611561775208, + "electricity_test_recall": 0.7913852334022522, + "electricity_test_f1-score": 0.7897088527679443, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8696466088294983, + "electricity_test_score": 0.7913852334022522 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.43824800848960876, + "lr": 0.0, + "electricity_val_accuracy": 0.7944093942642212, + "electricity_val_precision": 0.7938113808631897, + "electricity_val_recall": 0.7944093942642212, + "electricity_val_f1-score": 0.7926322221755981, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8717016577720642, + "electricity_val_score": 0.7944093942642212 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4431895911693573, + "lr": 0.0, + "electricity_test_accuracy": 0.7890598773956299, + "electricity_test_precision": 0.788698673248291, + "electricity_test_recall": 0.7890598773956299, + "electricity_test_f1-score": 0.786831259727478, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.867773711681366, + "electricity_test_score": 0.7890598773956299 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.43396085500717163, + "lr": 0.0, + "electricity_val_accuracy": 0.7975654006004333, + "electricity_val_precision": 0.7974778413772583, + "electricity_val_recall": 0.7975654006004333, + "electricity_val_f1-score": 0.7953619956970215, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8744989633560181, + "electricity_val_score": 0.7975654006004333 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4418977200984955, + "lr": 0.0, + "electricity_test_accuracy": 0.7923818230628967, + "electricity_test_precision": 0.7923421263694763, + "electricity_test_recall": 0.7923818230628967, + "electricity_test_f1-score": 0.7899752259254456, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8689972162246704, + "electricity_test_score": 0.7923818230628967 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.4241747558116913, + "lr": 0.0, + "electricity_val_accuracy": 0.799143373966217, + "electricity_val_precision": 0.7983817458152771, + "electricity_val_recall": 0.799143373966217, + "electricity_val_f1-score": 0.7978335618972778, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8805555105209351, + "electricity_val_score": 0.799143373966217 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.43167173862457275, + "lr": 0.0, + "electricity_test_accuracy": 0.7930461764335632, + "electricity_test_precision": 0.7922996282577515, + "electricity_test_recall": 0.7930461764335632, + "electricity_test_f1-score": 0.7914779186248779, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.875771164894104, + "electricity_test_score": 0.7930461764335632 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.42530372738838196, + "lr": 0.0, + "electricity_val_accuracy": 0.8027502298355103, + "electricity_val_precision": 0.8020356297492981, + "electricity_val_recall": 0.8027502298355103, + "electricity_val_f1-score": 0.8014844655990601, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8806633353233337, + "electricity_val_score": 0.8027502298355103 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4308037757873535, + "lr": 0.0, + "electricity_test_accuracy": 0.7958143949508667, + "electricity_test_precision": 0.7950441837310791, + "electricity_test_recall": 0.7958143949508667, + "electricity_test_f1-score": 0.7944121956825256, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8773896098136902, + "electricity_test_score": 0.7958143949508667 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.4377902626991272, + "std": 0.0060253322403626556, + "n": 5, + "ci95": 0.007480238734937253 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.7923375010490418, + "std": 0.002461839095605835, + "n": 5, + "ci95": 0.003056286927843382 + }, + "electricity_test_precision": { + "mean": 0.7918091535568237, + "std": 0.0023443032766834997, + "n": 5, + "ci95": 0.0029103703293268984 + }, + "electricity_test_recall": { + "mean": 0.7923375010490418, + "std": 0.002461839095605835, + "n": 5, + "ci95": 0.003056286927843382 + }, + "electricity_test_f1-score": { + "mean": 0.7904810905456543, + "std": 0.0027668018097288566, + "n": 5, + "ci95": 0.0034348874457722192 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8719156622886658, + "std": 0.004348870189761447, + "n": 5, + "ci95": 0.005398969874018023 + }, + "electricity_test_score": { + "mean": 0.7923375010490418, + "std": 0.002461839095605835, + "n": 5, + "ci95": 0.003056286927843382 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/downstream_results/mlp_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/downstream_results/mlp_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..ea07ebed8c3d3329006698738838ec438f444088 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/downstream_results/mlp_electricity_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "electricity", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpkhzioelk", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpkhzioelk/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.4537869095802307, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7860685586929321, + "electricity_val_precision": 0.7872244715690613, + "electricity_val_recall": 0.7860685586929321, + "electricity_val_f1-score": 0.7825033664703369, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8630810379981995, + "electricity_val_score": 0.7860685586929321 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4640178680419922, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7785406112670898, + "electricity_test_precision": 0.779639720916748, + "electricity_test_recall": 0.7785406112670898, + "electricity_test_f1-score": 0.7746859788894653, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.857112467288971, + "electricity_test_score": 0.7785406112670898 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.45973119139671326, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7804328203201294, + "electricity_val_precision": 0.7838467359542847, + "electricity_val_recall": 0.7804328203201294, + "electricity_val_f1-score": 0.7752981781959534, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8611858487129211, + "electricity_val_score": 0.7804328203201294 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4699283838272095, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.775993824005127, + "electricity_test_precision": 0.7793232798576355, + "electricity_test_recall": 0.775993824005127, + "electricity_test_f1-score": 0.7706780433654785, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8553431034088135, + "electricity_test_score": 0.775993824005127 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.45084697008132935, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7880973815917969, + "electricity_val_precision": 0.79059237241745, + "electricity_val_recall": 0.7880973815917969, + "electricity_val_f1-score": 0.7838549017906189, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8652268648147583, + "electricity_val_score": 0.7880973815917969 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.46176084876060486, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7780976891517639, + "electricity_test_precision": 0.7810484766960144, + "electricity_test_recall": 0.7780976891517639, + "electricity_test_f1-score": 0.7731097340583801, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.859870433807373, + "electricity_test_score": 0.7780976891517639 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.46266278624534607, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7763751149177551, + "electricity_val_precision": 0.7772219777107239, + "electricity_val_recall": 0.7763751149177551, + "electricity_val_f1-score": 0.776714563369751, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8581639528274536, + "electricity_val_score": 0.7763751149177551 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4694019854068756, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7724504470825195, + "electricity_test_precision": 0.7733598947525024, + "electricity_test_recall": 0.7724504470825195, + "electricity_test_f1-score": 0.7728118300437927, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8540533781051636, + "electricity_test_score": 0.7724504470825195 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.47277605533599854, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7711902856826782, + "electricity_val_precision": 0.784538209438324, + "electricity_val_recall": 0.7711902856826782, + "electricity_val_f1-score": 0.7611512541770935, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8615245819091797, + "electricity_val_score": 0.7711902856826782 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4854964017868042, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7664710283279419, + "electricity_test_precision": 0.7811596393585205, + "electricity_test_recall": 0.7664710283279419, + "electricity_test_f1-score": 0.7555137276649475, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8552334308624268, + "electricity_test_score": 0.7664710283279419 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.4701210975646973, + "std": 0.009274931421250916, + "n": 5, + "ci95": 0.011514502190662924 + }, + "lr": { + "mean": 0.0009999994654208422, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.7743107199668884, + "std": 0.0049996833754833245, + "n": 5, + "ci95": 0.0062069316273024675 + }, + "electricity_test_precision": { + "mean": 0.7789062023162842, + "std": 0.0032070566266500494, + "n": 5, + "ci95": 0.0039814483661337495 + }, + "electricity_test_recall": { + "mean": 0.7743107199668884, + "std": 0.0049996833754833245, + "n": 5, + "ci95": 0.0062069316273024675 + }, + "electricity_test_f1-score": { + "mean": 0.7693598628044128, + "std": 0.007870803375235038, + "n": 5, + "ci95": 0.009771326448707894 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8563225626945495, + "std": 0.002264182305069539, + "n": 5, + "ci95": 0.0028109029520206664 + }, + "electricity_test_score": { + "mean": 0.7743107199668884, + "std": 0.0049996833754833245, + "n": 5, + "ci95": 0.0062069316273024675 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/downstream_results/resnet_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/downstream_results/resnet_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..50cb352ef5a97f6b2be8f993c84f998c7f2cecee --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/downstream_results/resnet_electricity_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "electricity", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpzvvwj8nd", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpzvvwj8nd/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.41599124670028687, + "lr": 0.0, + "electricity_val_accuracy": 0.806582510471344, + "electricity_val_precision": 0.8058660626411438, + "electricity_val_recall": 0.806582510471344, + "electricity_val_f1-score": 0.8054895997047424, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8875266313552856, + "electricity_val_score": 0.806582510471344 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4162741005420685, + "lr": 0.0, + "electricity_test_accuracy": 0.8062230348587036, + "electricity_test_precision": 0.8054976463317871, + "electricity_test_recall": 0.8062230348587036, + "electricity_test_f1-score": 0.8051841259002686, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8876495361328125, + "electricity_test_score": 0.8062230348587036 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.4219616949558258, + "lr": 0.0, + "electricity_val_accuracy": 0.8036519289016724, + "electricity_val_precision": 0.811510443687439, + "electricity_val_recall": 0.8036519289016724, + "electricity_val_f1-score": 0.8047614097595215, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.895802915096283, + "electricity_val_score": 0.8036519289016724 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4367612600326538, + "lr": 0.0, + "electricity_test_accuracy": 0.7992470264434814, + "electricity_test_precision": 0.8074566125869751, + "electricity_test_recall": 0.7992470264434814, + "electricity_test_f1-score": 0.8003823161125183, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8887995481491089, + "electricity_test_score": 0.7992470264434814 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.4019428491592407, + "lr": 0.0, + "electricity_val_accuracy": 0.8142470717430115, + "electricity_val_precision": 0.8157529234886169, + "electricity_val_recall": 0.8142470717430115, + "electricity_val_f1-score": 0.8147201538085938, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.896934449672699, + "electricity_val_score": 0.8142470717430115 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.40368959307670593, + "lr": 0.0, + "electricity_test_accuracy": 0.8133097290992737, + "electricity_test_precision": 0.8153637051582336, + "electricity_test_recall": 0.8133097290992737, + "electricity_test_f1-score": 0.8138866424560547, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8966759443283081, + "electricity_test_score": 0.8133097290992737 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.40096232295036316, + "lr": 0.0009755285573191941, + "electricity_val_accuracy": 0.8101893663406372, + "electricity_val_precision": 0.8119125366210938, + "electricity_val_recall": 0.8101893663406372, + "electricity_val_f1-score": 0.8072807788848877, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8953582048416138, + "electricity_val_score": 0.8101893663406372 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.40565043687820435, + "lr": 0.0009755277424119413, + "electricity_test_accuracy": 0.8141955733299255, + "electricity_test_precision": 0.8153116703033447, + "electricity_test_recall": 0.8141955733299255, + "electricity_test_f1-score": 0.8117561936378479, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8938940763473511, + "electricity_test_score": 0.8141955733299255 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.38813501596450806, + "lr": 0.0, + "electricity_val_accuracy": 0.8196573257446289, + "electricity_val_precision": 0.8190400004386902, + "electricity_val_recall": 0.8196573257446289, + "electricity_val_f1-score": 0.8190333247184753, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9017026424407959, + "electricity_val_score": 0.8196573257446289 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.39023786783218384, + "lr": 0.0, + "electricity_test_accuracy": 0.8188462257385254, + "electricity_test_precision": 0.8183874487876892, + "electricity_test_recall": 0.8188462257385254, + "electricity_test_f1-score": 0.818522572517395, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9018798470497131, + "electricity_test_score": 0.8188462257385254 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.4105226516723633, + "std": 0.017345354961779063, + "n": 5, + "ci95": 0.021533650076120372 + }, + "lr": { + "mean": 0.00019510554848238825, + "std": 0.00043626926919400105, + "n": 5, + "ci95": 0.0005416130025871097 + }, + "electricity_test_accuracy": { + "mean": 0.8103643178939819, + "std": 0.007681814896895927, + "n": 5, + "ci95": 0.009536703878576561 + }, + "electricity_test_precision": { + "mean": 0.8124034166336059, + "std": 0.005594426291295208, + "n": 5, + "ci95": 0.006945284106255198 + }, + "electricity_test_recall": { + "mean": 0.8103643178939819, + "std": 0.007681814896895927, + "n": 5, + "ci95": 0.009536703878576561 + }, + "electricity_test_f1-score": { + "mean": 0.8099463701248169, + "std": 0.007185098361809486, + "n": 5, + "ci95": 0.00892004771459816 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8937797904014587, + "std": 0.005839403155304271, + "n": 5, + "ci95": 0.007249414294305003 + }, + "electricity_test_score": { + "mean": 0.8103643178939819, + "std": 0.007681814896895927, + "n": 5, + "ci95": 0.009536703878576561 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..bfbea6c35ad94e1650d5ff51161a0afb6d52807c --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:88c3998dd50f55c5909b37a0b20831dcd9147996c123ef26796e5b326797a93a +size 92799104 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..578a6d245c87cc23c60be9578810c790b7729014 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:24dbb6809fc9780a059ac1ad16c0cc8d9d972ea6a3aeb67743fbe6d3df85c8cc +size 8331758 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9847a1583841a5784beaea258f16c22ecc1b869a --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f0.001_sx0.0002_sy4e-05/seed_42/wjepa_deterministic/electricity/20260916_004617/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: electricity +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.0005 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.6 +mask_max_trgt_share: 0.9 +mask_min_ctx_share: 0.15 +mask_min_trgt_share: 0.15 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 64 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 6 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 128 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.001 +recon_y_weight: 0.25 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 0.0002 +sw_s_y_weight: 4.0e-05 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 40 +vtjepa_kl_x_end_weight: 1.0e-06 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-06 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 40 +vtjepa_kl_z_end_weight: 1.0e-05 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 40 +wjepa_joint_end_weight: 0.001 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..e693278c0b62deeaea1e3718f4e9e53bbe84655c --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-16T00:46:05.083942+00:00", + "spec": { + "dataset": "electricity", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 1.25e-05, + "sw_z_weight": 0.0, + "sw_s_y_weight": 2.5e-06, + "lambda_joint": 6.25e-05, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..c36fc10f672c8ee7a705f597494a4616459e72f6 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "electricity", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 1.25e-05, + "sw_z_weight": 0.0, + "sw_s_y_weight": 2.5e-06, + "lambda_joint": 6.25e-05, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "electricity", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "6.25e-05", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "1.25e-05", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "2.5e-06", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..093f84a8ad29bb1c93001429e0b75055efdf54da --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "electricity", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 6.25e-05, + "s_x": 1.25e-05, + "z": 0.0, + "s_y": 2.5e-06 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 40 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "electricity", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "64", + "--model_num_layers", + "6", + "--model_num_heads", + "4", + "--model_dropout_prob", + "0.001", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.15", + "--mask_max_ctx_share", + "0.6", + "--mask_min_trgt_share", + "0.15", + "--mask_max_trgt_share", + "0.9", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "128", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.0005", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "1e-06", + "--vtjepa_kl_x_annealing_epochs", + "40", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-05", + "--vtjepa_kl_z_annealing_epochs", + "40", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-06", + "--vtjepa_kl_y_annealing_epochs", + "15", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.001", + "--recon_y_weight", + "0.25", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "mean", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "6.25e-05", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "40", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "1.25e-05", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "2.5e-06", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/downstream_results/autoint_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/downstream_results/autoint_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..13a4faf7e77f701215f23fc49736843c371d7447 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/downstream_results/autoint_electricity_downstream_results.json @@ -0,0 +1,226 @@ +{ + "dataset": "electricity", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp04d1dq9u", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp04d1dq9u/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.38905662298202515, + "lr": 0.0, + "electricity_val_accuracy": 0.8185302019119263, + "electricity_val_precision": 0.817913830280304, + "electricity_val_recall": 0.8185302019119263, + "electricity_val_f1-score": 0.8179429173469543, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9024119973182678, + "electricity_val_score": 0.8185302019119263 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3939928710460663, + "lr": 0.0, + "electricity_test_accuracy": 0.8220573663711548, + "electricity_test_precision": 0.821478009223938, + "electricity_test_recall": 0.8220573663711548, + "electricity_test_f1-score": 0.8215071558952332, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9005591869354248, + "electricity_test_score": 0.8220573663711548 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.4040851593017578, + "lr": 0.0, + "electricity_val_accuracy": 0.8090622425079346, + "electricity_val_precision": 0.8083510994911194, + "electricity_val_recall": 0.8090622425079346, + "electricity_val_f1-score": 0.8083583116531372, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8943294286727905, + "electricity_val_score": 0.8090622425079346 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4054798185825348, + "lr": 0.0, + "electricity_test_accuracy": 0.8136419057846069, + "electricity_test_precision": 0.8129703402519226, + "electricity_test_recall": 0.8136419057846069, + "electricity_test_f1-score": 0.8129176497459412, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8931689858436584, + "electricity_test_score": 0.8136419057846069 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.39698395133018494, + "lr": 0.0, + "electricity_val_accuracy": 0.8106402158737183, + "electricity_val_precision": 0.8099404573440552, + "electricity_val_recall": 0.8106402158737183, + "electricity_val_f1-score": 0.8099334836006165, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8982093930244446, + "electricity_val_score": 0.8106402158737183 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3987520933151245, + "lr": 0.0, + "electricity_test_accuracy": 0.8179603815078735, + "electricity_test_precision": 0.8173872828483582, + "electricity_test_recall": 0.8179603815078735, + "electricity_test_f1-score": 0.8174788355827332, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8975237011909485, + "electricity_test_score": 0.8179603815078735 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.3926725387573242, + "lr": 0.0, + "electricity_val_accuracy": 0.8153741955757141, + "electricity_val_precision": 0.8147094249725342, + "electricity_val_recall": 0.8153741955757141, + "electricity_val_f1-score": 0.8146255612373352, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8999230265617371, + "electricity_val_score": 0.8153741955757141 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3986698389053345, + "lr": 0.0, + "electricity_test_accuracy": 0.815302848815918, + "electricity_test_precision": 0.8147023320198059, + "electricity_test_recall": 0.815302848815918, + "electricity_test_f1-score": 0.8147909641265869, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8968901038169861, + "electricity_test_score": 0.815302848815918 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.38629573583602905, + "lr": 0.0, + "electricity_val_accuracy": 0.8228133320808411, + "electricity_val_precision": 0.8223989605903625, + "electricity_val_recall": 0.8228133320808411, + "electricity_val_f1-score": 0.8216855525970459, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9044284224510193, + "electricity_val_score": 0.8228133320808411 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.389995276927948, + "lr": 0.0, + "electricity_test_accuracy": 0.822389543056488, + "electricity_test_precision": 0.8218015432357788, + "electricity_test_recall": 0.822389543056488, + "electricity_test_f1-score": 0.8216304779052734, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9021990299224854, + "electricity_test_score": 0.822389543056488 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.3973779797554016, + "std": 0.005812919716709318, + "n": 5, + "ci95": 0.0072165360337693866 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.8182704091072083, + "std": 0.003925362531448218, + "n": 5, + "ci95": 0.004873199963931453 + }, + "electricity_test_precision": { + "mean": 0.8176679015159607, + "std": 0.003954249161241952, + "n": 5, + "ci95": 0.004909061702087136 + }, + "electricity_test_recall": { + "mean": 0.8182704091072083, + "std": 0.003925362531448218, + "n": 5, + "ci95": 0.004873199963931453 + }, + "electricity_test_f1-score": { + "mean": 0.8176650166511535, + "std": 0.0039153304702711405, + "n": 5, + "ci95": 0.004860745511693062 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8980682015419006, + "std": 0.003498408948369725, + "n": 5, + "ci95": 0.004343152059059112 + }, + "electricity_test_score": { + "mean": 0.8182704091072083, + "std": 0.003925362531448218, + "n": 5, + "ci95": 0.004873199963931453 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/downstream_results/dcnv2_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/downstream_results/dcnv2_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..f22ee67546f8476a7f499263de3d36ffa832ab63 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/downstream_results/dcnv2_electricity_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "electricity", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpblid9xq_", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpblid9xq_/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.3722821772098541, + "lr": 0.0, + "electricity_val_accuracy": 0.829801619052887, + "electricity_val_precision": 0.8292633295059204, + "electricity_val_recall": 0.829801619052887, + "electricity_val_f1-score": 0.8292203545570374, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9130862355232239, + "electricity_val_score": 0.829801619052887 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.37216877937316895, + "lr": 0.0, + "electricity_test_accuracy": 0.8314693570137024, + "electricity_test_precision": 0.8310815095901489, + "electricity_test_recall": 0.8314693570137024, + "electricity_test_f1-score": 0.8311949968338013, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9135711193084717, + "electricity_test_score": 0.8314693570137024 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.3699020743370056, + "lr": 0.0, + "electricity_val_accuracy": 0.8352118730545044, + "electricity_val_precision": 0.8347164392471313, + "electricity_val_recall": 0.8352118730545044, + "electricity_val_f1-score": 0.834619402885437, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9130039811134338, + "electricity_val_score": 0.8352118730545044 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3702132999897003, + "lr": 0.0, + "electricity_test_accuracy": 0.8371165990829468, + "electricity_test_precision": 0.8366872668266296, + "electricity_test_recall": 0.8371165990829468, + "electricity_test_f1-score": 0.836763322353363, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.913454532623291, + "electricity_test_score": 0.8371165990829468 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.3800623416900635, + "lr": 0.0, + "electricity_val_accuracy": 0.8275473117828369, + "electricity_val_precision": 0.8270005583763123, + "electricity_val_recall": 0.8275473117828369, + "electricity_val_f1-score": 0.8268158435821533, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9096480011940002, + "electricity_val_score": 0.8275473117828369 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.379536896944046, + "lr": 0.0, + "electricity_test_accuracy": 0.8305835723876953, + "electricity_test_precision": 0.8300699591636658, + "electricity_test_recall": 0.8305835723876953, + "electricity_test_f1-score": 0.8300923705101013, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9107470512390137, + "electricity_test_score": 0.8305835723876953 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.3760015070438385, + "lr": 0.0, + "electricity_val_accuracy": 0.8257439136505127, + "electricity_val_precision": 0.8251768946647644, + "electricity_val_recall": 0.8257439136505127, + "electricity_val_f1-score": 0.8250210881233215, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9103026986122131, + "electricity_val_score": 0.8257439136505127 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3792365789413452, + "lr": 0.0, + "electricity_test_accuracy": 0.8306942582130432, + "electricity_test_precision": 0.8301681280136108, + "electricity_test_recall": 0.8306942582130432, + "electricity_test_f1-score": 0.8301042318344116, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9091469049453735, + "electricity_test_score": 0.8306942582130432 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.3685497045516968, + "lr": 0.0, + "electricity_val_accuracy": 0.8311541676521301, + "electricity_val_precision": 0.8307291865348816, + "electricity_val_recall": 0.8311541676521301, + "electricity_val_f1-score": 0.8308382630348206, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9138367772102356, + "electricity_val_score": 0.8311541676521301 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.37744617462158203, + "lr": 0.0, + "electricity_test_accuracy": 0.8326874375343323, + "electricity_test_precision": 0.8323187232017517, + "electricity_test_recall": 0.8326874375343323, + "electricity_test_f1-score": 0.832431435585022, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.9106313586235046, + "electricity_test_score": 0.8326874375343323 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.37572034597396853, + "std": 0.00426759946872212, + "n": 5, + "ci95": 0.005298075123109139 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.832510244846344, + "std": 0.0027084690462108253, + "n": 5, + "ci95": 0.003362469364946646 + }, + "electricity_test_precision": { + "mean": 0.8320651173591613, + "std": 0.0027374847336770793, + "n": 5, + "ci95": 0.003398491323678144 + }, + "electricity_test_recall": { + "mean": 0.832510244846344, + "std": 0.0027084690462108253, + "n": 5, + "ci95": 0.003362469364946646 + }, + "electricity_test_f1-score": { + "mean": 0.8321172714233398, + "std": 0.002769820882222796, + "n": 5, + "ci95": 0.0034386355184281084 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.9115101933479309, + "std": 0.0019344081431707343, + "n": 5, + "ci95": 0.002401499891540065 + }, + "electricity_test_score": { + "mean": 0.832510244846344, + "std": 0.0027084690462108253, + "n": 5, + "ci95": 0.003362469364946646 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/downstream_results/fttransformer_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/downstream_results/fttransformer_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..53cb5b1931a220cbd325461785987a62fe7a1bdc --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/downstream_results/fttransformer_electricity_downstream_results.json @@ -0,0 +1,225 @@ +{ + "dataset": "electricity", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp6n705fqc", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp6n705fqc/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.43035516142845154, + "lr": 0.0, + "electricity_val_accuracy": 0.7980162501335144, + "electricity_val_precision": 0.7972181439399719, + "electricity_val_recall": 0.7980162501335144, + "electricity_val_f1-score": 0.7967514991760254, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8773476481437683, + "electricity_val_score": 0.7980162501335144 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.43873563408851624, + "lr": 0.0, + "electricity_test_accuracy": 0.7934890985488892, + "electricity_test_precision": 0.7927331924438477, + "electricity_test_recall": 0.7934890985488892, + "electricity_test_f1-score": 0.7919571399688721, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.871421754360199, + "electricity_test_score": 0.7934890985488892 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.4380297362804413, + "lr": 0.0, + "electricity_val_accuracy": 0.7966636419296265, + "electricity_val_precision": 0.7963434457778931, + "electricity_val_recall": 0.7966636419296265, + "electricity_val_f1-score": 0.7946465015411377, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8723262548446655, + "electricity_val_score": 0.7966636419296265 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.44406530261039734, + "lr": 0.0, + "electricity_test_accuracy": 0.7897242903709412, + "electricity_test_precision": 0.7894430160522461, + "electricity_test_recall": 0.7897242903709412, + "electricity_test_f1-score": 0.7874399423599243, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.867798388004303, + "electricity_test_score": 0.7897242903709412 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.43722057342529297, + "lr": 0.0, + "electricity_val_accuracy": 0.7928313612937927, + "electricity_val_precision": 0.7921154499053955, + "electricity_val_recall": 0.7928313612937927, + "electricity_val_f1-score": 0.7911683320999146, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8720449805259705, + "electricity_val_score": 0.7928313612937927 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4454279839992523, + "lr": 0.0, + "electricity_test_accuracy": 0.7881740927696228, + "electricity_test_precision": 0.7875502705574036, + "electricity_test_recall": 0.7881740927696228, + "electricity_test_f1-score": 0.7862189412117004, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8667097687721252, + "electricity_test_score": 0.7881740927696228 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.4373626708984375, + "lr": 0.0, + "electricity_val_accuracy": 0.7950856685638428, + "electricity_val_precision": 0.7942851185798645, + "electricity_val_recall": 0.7950856685638428, + "electricity_val_f1-score": 0.7936848402023315, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8721091151237488, + "electricity_val_score": 0.7950856685638428 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.44168680906295776, + "lr": 0.0, + "electricity_test_accuracy": 0.7886170148849487, + "electricity_test_precision": 0.7877634167671204, + "electricity_test_recall": 0.7886170148849487, + "electricity_test_f1-score": 0.7870712876319885, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8692559599876404, + "electricity_test_score": 0.7886170148849487 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.4379025101661682, + "lr": 0.0, + "electricity_val_accuracy": 0.7885482311248779, + "electricity_val_precision": 0.7882094383239746, + "electricity_val_recall": 0.7885482311248779, + "electricity_val_f1-score": 0.7862467169761658, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8715797066688538, + "electricity_val_score": 0.7885482311248779 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.44563066959381104, + "lr": 0.0, + "electricity_test_accuracy": 0.790610134601593, + "electricity_test_precision": 0.7904745936393738, + "electricity_test_recall": 0.790610134601593, + "electricity_test_f1-score": 0.7882214188575745, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8668137192726135, + "electricity_test_score": 0.790610134601593 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.4431092798709869, + "std": 0.0029068872990402725, + "n": 5, + "ci95": 0.003608798669510285 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.790122926235199, + "std": 0.0021090945605504157, + "n": 5, + "ci95": 0.002618366954404678 + }, + "electricity_test_precision": { + "mean": 0.7895928978919983, + "std": 0.002131937466456304, + "n": 5, + "ci95": 0.002646725621239865 + }, + "electricity_test_recall": { + "mean": 0.790122926235199, + "std": 0.0021090945605504157, + "n": 5, + "ci95": 0.002618366954404678 + }, + "electricity_test_f1-score": { + "mean": 0.788181746006012, + "std": 0.002229975707284255, + "n": 5, + "ci95": 0.0027684366601156574 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8683999180793762, + "std": 0.0019747640380179104, + "n": 5, + "ci95": 0.0024516003201598704 + }, + "electricity_test_score": { + "mean": 0.790122926235199, + "std": 0.0021090945605504157, + "n": 5, + "ci95": 0.002618366954404678 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/downstream_results/mlp_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/downstream_results/mlp_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..8e240d2558670c9a6a8ad0637985b769210ff97f --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/downstream_results/mlp_electricity_downstream_results.json @@ -0,0 +1,221 @@ +{ + "dataset": "electricity", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpxxknorc4", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpxxknorc4/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.4634443521499634, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.78291255235672, + "electricity_val_precision": 0.7822018265724182, + "electricity_val_recall": 0.78291255235672, + "electricity_val_f1-score": 0.7824164032936096, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8609635233879089, + "electricity_val_score": 0.78291255235672 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4719853401184082, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7798693180084229, + "electricity_test_precision": 0.7790170311927795, + "electricity_test_recall": 0.7798693180084229, + "electricity_test_f1-score": 0.7792123556137085, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.854252815246582, + "electricity_test_score": 0.7798693180084229 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.4530666768550873, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7826871275901794, + "electricity_val_precision": 0.7834422588348389, + "electricity_val_recall": 0.7826871275901794, + "electricity_val_f1-score": 0.7792355418205261, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8632467985153198, + "electricity_val_score": 0.7826871275901794 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.46097177267074585, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7796478867530823, + "electricity_test_precision": 0.7805474996566772, + "electricity_test_recall": 0.7796478867530823, + "electricity_test_f1-score": 0.7759845852851868, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8586447238922119, + "electricity_test_score": 0.7796478867530823 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.4652678668498993, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7795311212539673, + "electricity_val_precision": 0.7875102162361145, + "electricity_val_recall": 0.7795311212539673, + "electricity_val_f1-score": 0.7722758650779724, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8631045818328857, + "electricity_val_score": 0.7795311212539673 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4715755581855774, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.775550901889801, + "electricity_test_precision": 0.7834439277648926, + "electricity_test_recall": 0.775550901889801, + "electricity_test_f1-score": 0.7680745720863342, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8615056276321411, + "electricity_test_score": 0.775550901889801 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.45486560463905334, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7880973815917969, + "electricity_val_precision": 0.7871474623680115, + "electricity_val_recall": 0.7880973815917969, + "electricity_val_f1-score": 0.7867923378944397, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8619388937950134, + "electricity_val_score": 0.7880973815917969 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4607187509536743, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7816410064697266, + "electricity_test_precision": 0.7806061506271362, + "electricity_test_recall": 0.7816410064697266, + "electricity_test_f1-score": 0.780341386795044, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8581328392028809, + "electricity_test_score": 0.7816410064697266 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.4587739408016205, + "lr": 0.0009999999310821295, + "electricity_val_accuracy": 0.7813345193862915, + "electricity_val_precision": 0.7838097810745239, + "electricity_val_recall": 0.7813345193862915, + "electricity_val_f1-score": 0.7767515182495117, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8614188432693481, + "electricity_val_score": 0.7813345193862915 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.46351954340934753, + "lr": 0.0009999994654208422, + "electricity_test_accuracy": 0.7777654528617859, + "electricity_test_precision": 0.7799685001373291, + "electricity_test_recall": 0.7777654528617859, + "electricity_test_f1-score": 0.7731794118881226, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8589141964912415, + "electricity_test_score": 0.7777654528617859 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.46575419306755067, + "std": 0.005611068673516449, + "n": 5, + "ci95": 0.006965945040319385 + }, + "lr": { + "mean": 0.0009999994654208422, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.7788949131965637, + "std": 0.002319243111808993, + "n": 5, + "ci95": 0.0028792590132168147 + }, + "electricity_test_precision": { + "mean": 0.780716621875763, + "std": 0.001652823210437528, + "n": 5, + "ci95": 0.002051922069607568 + }, + "electricity_test_recall": { + "mean": 0.7788949131965637, + "std": 0.002319243111808993, + "n": 5, + "ci95": 0.0028792590132168147 + }, + "electricity_test_f1-score": { + "mean": 0.7753584623336792, + "std": 0.0049466498140553275, + "n": 5, + "ci95": 0.006141092320087518 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8582900404930115, + "std": 0.0026066871508275014, + "n": 5, + "ci95": 0.0032361107101887394 + }, + "electricity_test_score": { + "mean": 0.7788949131965637, + "std": 0.002319243111808993, + "n": 5, + "ci95": 0.0028792590132168147 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/downstream_results/resnet_electricity_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/downstream_results/resnet_electricity_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..40cd7157e73b0f14cb0e6bf3912df95f5ef85ecf --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/downstream_results/resnet_electricity_downstream_results.json @@ -0,0 +1,224 @@ +{ + "dataset": "electricity", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp4h46amba", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "electricity", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp4h46amba/split_data/split_config.json", + "sigma_aggregation": "mean", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "electricity_val_loss": 0.4116632044315338, + "lr": 0.0, + "electricity_val_accuracy": 0.8113164901733398, + "electricity_val_precision": 0.8107220530509949, + "electricity_val_recall": 0.8113164901733398, + "electricity_val_f1-score": 0.8101252913475037, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.889289379119873, + "electricity_val_score": 0.8113164901733398 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4123215079307556, + "lr": 0.0, + "electricity_test_accuracy": 0.8057801127433777, + "electricity_test_precision": 0.8050783276557922, + "electricity_test_recall": 0.8057801127433777, + "electricity_test_f1-score": 0.804643452167511, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8895878791809082, + "electricity_test_score": 0.8057801127433777 + } + } + }, + { + "seed": 1, + "val_metrics": { + "electricity_val_loss": 0.4013522267341614, + "lr": 0.0, + "electricity_val_accuracy": 0.8124436140060425, + "electricity_val_precision": 0.81708163022995, + "electricity_val_recall": 0.8124436140060425, + "electricity_val_f1-score": 0.8133341073989868, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.9013316631317139, + "electricity_val_score": 0.8124436140060425 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.41224122047424316, + "lr": 0.0, + "electricity_test_accuracy": 0.8106521964073181, + "electricity_test_precision": 0.8162451982498169, + "electricity_test_recall": 0.8106521964073181, + "electricity_test_f1-score": 0.8116121888160706, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8965729475021362, + "electricity_test_score": 0.8106521964073181 + } + } + }, + { + "seed": 2, + "val_metrics": { + "electricity_val_loss": 0.40094587206840515, + "lr": 0.0, + "electricity_val_accuracy": 0.8135707974433899, + "electricity_val_precision": 0.8143323063850403, + "electricity_val_recall": 0.8135707974433899, + "electricity_val_f1-score": 0.8138594627380371, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8971577882766724, + "electricity_val_score": 0.8135707974433899 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.4043063521385193, + "lr": 0.0, + "electricity_test_accuracy": 0.816963791847229, + "electricity_test_precision": 0.8182949423789978, + "electricity_test_recall": 0.816963791847229, + "electricity_test_f1-score": 0.8173933625221252, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8958669900894165, + "electricity_test_score": 0.816963791847229 + } + } + }, + { + "seed": 3, + "val_metrics": { + "electricity_val_loss": 0.40070128440856934, + "lr": 0.0, + "electricity_val_accuracy": 0.8074842095375061, + "electricity_val_precision": 0.8087881803512573, + "electricity_val_recall": 0.8074842095375061, + "electricity_val_f1-score": 0.8079221844673157, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.8958603739738464, + "electricity_val_score": 0.8074842095375061 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.40981781482696533, + "lr": 0.0, + "electricity_test_accuracy": 0.8098770976066589, + "electricity_test_precision": 0.8112821578979492, + "electricity_test_recall": 0.8098770976066589, + "electricity_test_f1-score": 0.8103334307670593, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.8916388154029846, + "electricity_test_score": 0.8098770976066589 + } + } + }, + { + "seed": 4, + "val_metrics": { + "electricity_val_loss": 0.3866618275642395, + "lr": 0.0, + "electricity_val_accuracy": 0.8178539276123047, + "electricity_val_precision": 0.8172124028205872, + "electricity_val_recall": 0.8178539276123047, + "electricity_val_f1-score": 0.8171406388282776, + "electricity_val_support": 4436.0, + "electricity_val_roc_auc": 0.902557909488678, + "electricity_val_score": 0.8178539276123047 + }, + "test_metrics": { + "0": { + "electricity_test_loss": 0.3917846977710724, + "lr": 0.0, + "electricity_test_accuracy": 0.8196213245391846, + "electricity_test_precision": 0.8190421462059021, + "electricity_test_recall": 0.8196213245391846, + "electricity_test_f1-score": 0.8191099166870117, + "electricity_test_support": 9031.0, + "electricity_test_roc_auc": 0.900456964969635, + "electricity_test_score": 0.8196213245391846 + } + } + } + ], + "summary": { + "0": { + "electricity_test_loss": { + "mean": 0.40609431862831114, + "std": 0.008636966859087955, + "n": 5, + "ci95": 0.010722491553068366 + }, + "lr": { + "mean": 0.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_accuracy": { + "mean": 0.8125789046287537, + "std": 0.005613787735776622, + "n": 5, + "ci95": 0.006969320660788081 + }, + "electricity_test_precision": { + "mean": 0.8139885544776917, + "std": 0.005828656184978704, + "n": 5, + "ci95": 0.007236072307422685 + }, + "electricity_test_recall": { + "mean": 0.8125789046287537, + "std": 0.005613787735776622, + "n": 5, + "ci95": 0.006969320660788081 + }, + "electricity_test_f1-score": { + "mean": 0.8126184701919555, + "std": 0.0058045971917500255, + "n": 5, + "ci95": 0.007206203910810929 + }, + "electricity_test_support": { + "mean": 9031.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "electricity_test_roc_auc": { + "mean": 0.8948247194290161, + "std": 0.004284836624723421, + "n": 5, + "ci95": 0.005319474447969162 + }, + "electricity_test_score": { + "mean": 0.8125789046287537, + "std": 0.005613787735776622, + "n": 5, + "ci95": 0.006969320660788081 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..0b977cb4f199d6b0113955ead77e24c7c55b3ec8 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ac00fa10d1e6825016144600050cbb02bcc8d9fe7fbb52091ece97295c880e55 +size 92799104 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..c1eaf755a91ef22883d7f279280d2c7349e47aa3 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d3970c990f954164d03f8632b501c8435de8dea4837c6c997e763686b710281e +size 8331758 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..0989237c656d1ed22f63d6cbe980929e2d742b8d --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/electricity/deterministic_sw/sw_f6.25e-05_sx1.25e-05_sy2.5e-06/seed_42/wjepa_deterministic/electricity/20260915_233553/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: electricity +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.0005 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.6 +mask_max_trgt_share: 0.9 +mask_min_ctx_share: 0.15 +mask_min_trgt_share: 0.15 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 64 +model_dim_hidden: 64 +model_dropout_prob: 0.001 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 4 +model_num_layers: 6 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 128 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.001 +recon_y_weight: 0.25 +res_dir: ./ +safe_exit: false +sigma_aggregation: mean +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 1.25e-05 +sw_s_y_weight: 2.5e-06 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 40 +vtjepa_kl_x_end_weight: 1.0e-06 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 15 +vtjepa_kl_y_end_weight: 1.0e-06 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 40 +vtjepa_kl_z_end_weight: 1.0e-05 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 40 +wjepa_joint_end_weight: 6.25e-05 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/bank-xgboost.exit_code b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/bank-xgboost.exit_code new file mode 100644 index 0000000000000000000000000000000000000000..573541ac9702dd3969c9bc859d2b91ec1f7e6e56 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/bank-xgboost.exit_code @@ -0,0 +1 @@ +0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/bank-xgboost.log b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/bank-xgboost.log new file mode 100644 index 0000000000000000000000000000000000000000..baa5754894dc9d2113ea192edde1615ce1512d87 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/bank-xgboost.log @@ -0,0 +1,195 @@ +[Config] Loaded sigma_aggregation=mean from /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/vtjepa_config.yaml +[Config] Loaded training data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets from /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/vtjepa_config.yaml +Split statistics: + val: 4425 samples + test: 9008 samples + train: 31778 samples +Running downstream for dataset 'bank_marketing', model 'xgboost'... +Running command: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python -m src.run_benchmark --data_path=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets --data_set=bank_marketing --model_name=xgboost --using_embedding=True --emb_train=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/embedding.npy --emb_test=/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/embedding.npy --use_split_assignments=True --exp_train_total_epochs=1 --split_config_file=/tmp/tmpl8bjxmwc/split_data/split_config.json --sigma_aggregation=mean --n_estimators=100 --max_depth=3 --learning_rate=0.1 --subsample=1.0 --colsample_bytree=1.0 --reg_alpha=0.0 --reg_lambda=1.0 --n_jobs=-1 --random_state=42 --encoder_type=linear_flatten --torch_seed=0 --np_seed=0 --output_file=/tmp/tmpl8bjxmwc/output_0.json +General args: +╒════════════════════════════╤═════════════════════════════════════════════════════════════════════════════════╕ +│ T_max │ 10 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ batch_size │ 64 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_loader_nprocs │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_path │ /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ data_set │ bank_marketing │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ eta_min │ 0.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cache_cadence │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_cadence_type │ recurrent │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_device │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_n_runs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_patience │ -1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_train_total_epochs │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_warmup │ 1e-05 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ exp_weight_decay_scheduler │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ filter_uncertainty │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ final_weight_decay │ 0.01 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ full_dataset_cuda │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ gradient_clipping │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ ipe_scale │ 1.0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_from_checkpoint │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ load_path │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ log_tensorboard │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ lr │ 0.0001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mock │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_amp │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_dtype │ float32 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ model_name │ xgboost │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_distributed │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_gpus │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nodes │ 1 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ mp_nr │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ np_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ output_file │ /tmp/tmpl8bjxmwc/output_0.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ pin_memory │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ random │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ res_dir │ ./ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ split_config_file │ /tmp/tmpl8bjxmwc/split_data/split_config.json │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ start_lr │ 0.001 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ test │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ torch_seed │ 0 │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ uncertainty_test │ │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ use_split_assignments │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ using_embedding │ True │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ verbose │ False │ +├────────────────────────────┼─────────────────────────────────────────────────────────────────────────────────┤ +│ weight_decay │ 1e-05 │ +╘════════════════════════════╧═════════════════════════════════════════════════════════════════════════════════╛ +Loading model: xgboost + +===================================================================================================================================================================== +Layer (type:depth-idx) Input Shape Output Shape Param # Kernel Shape Mult-Adds +===================================================================================================================================================================== +XGBoostLightning [1, 1024] [1, 1024] -- -- -- +===================================================================================================================================================================== +Total params: 0 +Trainable params: 0 +Non-trainable params: 0 +Total mult-adds (Units.MEGABYTES): 0 +===================================================================================================================================================================== +Input size (MB): 0.00 +Forward/backward pass size (MB): 0.00 +Params size (MB): 0.00 +Estimated Total Size (MB): 0.00 +===================================================================================================================================================================== +GPU available: False, used: False +TPU available: False, using: 0 TPU cores +💡 Tip: For seamless cloud logging and experiment tracking, try installing [litlogger](https://pypi.org/project/litlogger/) to enable LitLogger, which logs metrics and artifacts automatically to the Lightning Experiments platform. +`Trainer(limit_train_batches=1)` was configured so 1 batch per epoch will be used. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/core/optimizer.py:183: `LightningModule.configure_optimizers` returned `None`, this fit will run with no optimizer +┏━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┳━━━━━━━┓ +┃ ┃ Name ┃ Type ┃ Params ┃ Mode ┃ FLOPs ┃ +┡━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━╇━━━━━━━┩ +│ 0 │ loss │ BCEWithLogitsLoss │ 0 │ train │ 0 │ +│ 1 │ head │ Identity │ 0 │ train │ 0 │ +│ 2 │ encoder │ Identity │ 0 │ train │ 0 │ +└───┴─────────┴───────────────────┴────────┴───────┴───────┘ +Trainable params: 0 +Non-trainable params: 0 +Total params: 0 +Total estimated model params size (MB): 0.000 +Modules in train mode: 3 +Modules in eval mode: 0 +Total FLOPs: 0 +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/torch/utils/data/dataloader.py:759: UserWarning: 'pin_memory' argument is set as true but no accelerator is found, then device pinned memory won't be used. + super().__init__(loader) +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +`Trainer.fit` stopped: `max_epochs=1` reached. +Epoch 0/0 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1/1 0:00:00 • 0:00:00 0.00it/s +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/torch/utils/data/dataloader.py:759: UserWarning: 'pin_memory' argument is set as true but no accelerator is found, then device pinned memory won't be used. + super().__init__(loader) +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Validate metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_val_accuracy │ 0.9021468758583069 │ +│ bank_marketing_val_f1-score │ 0.8954094052314758 │ +│ bank_marketing_val_loss │ 0.24640412628650665 │ +│ bank_marketing_val_precision │ 0.8922626376152039 │ +│ bank_marketing_val_recall │ 0.9021468758583069 │ +│ bank_marketing_val_roc_auc │ 0.9228941202163696 │ +│ bank_marketing_val_score │ 0.9021468758583069 │ +│ bank_marketing_val_support │ 4425.0 │ +└──────────────────────────────┴──────────────────────────────┘ +Validation ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 70/70 0:00:06 • 0:00:00 10.11it/s +Validation metrics: [{'bank_marketing_val_loss': 0.24640412628650665, 'bank_marketing_val_accuracy': 0.9021468758583069, 'bank_marketing_val_precision': 0.8922626376152039, 'bank_marketing_val_recall': 0.9021468758583069, 'bank_marketing_val_f1-score': 0.8954094052314758, 'bank_marketing_val_support': 4425.0, 'bank_marketing_val_roc_auc': 0.9228941202163696, 'bank_marketing_val_score': 0.9021468758583069}] +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/utilities/_pytree.py:21: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:434: The 'test_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=31` in the `DataLoader` to improve performance. +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/lib/python3.11/site-packages/torch/utils/data/dataloader.py:759: UserWarning: 'pin_memory' argument is set as true but no accelerator is found, then device pinned memory won't be used. + super().__init__(loader) + +=== CLASS FREQUENCIES (TEST) === +True class frequencies: + Class 0.0: 7909 samples (87.80%) + Class 1.0: 1099 samples (12.20%) +Predicted class frequencies: + Class 0: 8223 samples (91.29%) + Class 1: 785 samples (8.71%) +============================================================ +┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ +┃ Test metric ┃ DataLoader 0 ┃ +┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ +│ bank_marketing_test_accuracy │ 0.9023090600967407 │ +│ bank_marketing_test_f1-score │ 0.8951188921928406 │ +│ bank_marketing_test_loss │ 0.23810569941997528 │ +│ bank_marketing_test_precision │ 0.8922731876373291 │ +│ bank_marketing_test_recall │ 0.9023090600967407 │ +│ bank_marketing_test_roc_auc │ 0.9287988543510437 │ +│ bank_marketing_test_score │ 0.9023090600967407 │ +│ bank_marketing_test_support │ 9008.0 │ +└───────────────────────────────┴───────────────────────────────┘ +Testing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 141/141 0:00:15 • 0:00:00 9.37it/s +Test metrics: [{'bank_marketing_test_loss': 0.23810569941997528, 'bank_marketing_test_accuracy': 0.9023090600967407, 'bank_marketing_test_precision': 0.8922731876373291, 'bank_marketing_test_recall': 0.9023090600967407, 'bank_marketing_test_f1-score': 0.8951188921928406, 'bank_marketing_test_support': 9008.0, 'bank_marketing_test_roc_auc': 0.9287988543510437, 'bank_marketing_test_score': 0.9023090600967407}] +Wrote results for xgboost on bank_marketing to /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/downstream_results/xgboost_bank_marketing_downstream_results.json + +Downstream pipeline complete. Results under /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/bank_marketing/deterministic_sw/sw_f0.016_sx0.004_sy0.0004/seed_42/wjepa_deterministic/bank_marketing/20260916_041845/downstream_results diff --git "a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/bank-xgboost.log\\n" "b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/bank-xgboost.log\\n" new file mode 100644 index 0000000000000000000000000000000000000000..6940b7ca0035ef1e61f562ed60318b71915387b7 --- /dev/null +++ "b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/bank-xgboost.log\\n" @@ -0,0 +1 @@ +/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python: Error while finding module specification for 'src.downstream_pipeline' (ModuleNotFoundError: No module named 'src') diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue0-gpu2.exit_code b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue0-gpu2.exit_code new file mode 100644 index 0000000000000000000000000000000000000000..573541ac9702dd3969c9bc859d2b91ec1f7e6e56 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue0-gpu2.exit_code @@ -0,0 +1 @@ +0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue0-gpu2.log b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue0-gpu2.log new file mode 100644 index 0000000000000000000000000000000000000000..b1730e8d5eb07b86ded6a54f0c2ecc4e7f48e640 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue0-gpu2.log @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f34a509a1c44b70f4b80a5df16ea00f94713754078867f7dd4f0a15f7baf8346 +size 20376403 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue1-gpu3.exit_code b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue1-gpu3.exit_code new file mode 100644 index 0000000000000000000000000000000000000000..d00491fd7e5bb6fa28c517a0bb32b8b506539d4d --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue1-gpu3.exit_code @@ -0,0 +1 @@ +1 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue1-gpu3.log b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue1-gpu3.log new file mode 100644 index 0000000000000000000000000000000000000000..3a4c08275a0c6c732d99d59ee619f996e83ddc0d --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue1-gpu3.log @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:86095f12ca13ee39f1c0b33713bcd1188bfcbdb3b9677f25fe1c49ede3380d2d +size 18583927 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue2-gpu2.exit_code b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue2-gpu2.exit_code new file mode 100644 index 0000000000000000000000000000000000000000..573541ac9702dd3969c9bc859d2b91ec1f7e6e56 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue2-gpu2.exit_code @@ -0,0 +1 @@ +0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue2-gpu2.log b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue2-gpu2.log new file mode 100644 index 0000000000000000000000000000000000000000..fafd3a8062eff5c9384c705339209092a1861f01 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue2-gpu2.log @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0f3d4b5e0c309c22acdc108dff325273d23b754e2dc876831d51fa05bdf8fcfb +size 19350620 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue3-gpu3.exit_code b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue3-gpu3.exit_code new file mode 100644 index 0000000000000000000000000000000000000000..573541ac9702dd3969c9bc859d2b91ec1f7e6e56 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue3-gpu3.exit_code @@ -0,0 +1 @@ +0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue3-gpu3.log b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue3-gpu3.log new file mode 100644 index 0000000000000000000000000000000000000000..5ac3ef4c621e8bf34bad5e2279d0a51fd0e36b5d --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/logs/queue3-gpu3.log @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1fb6d2899e48a53507688c43ea464ff39fd0f3ae4b7db7ee53be5ab2fecfcd0b +size 18168961 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..97dcedd9aebb522f0f91c115d964e874b9195e5f --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-16T02:13:33.407745+00:00", + "spec": { + "dataset": "sim", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 2.5e-05, + "sw_z_weight": 0.0, + "sw_s_y_weight": 5e-06, + "lambda_joint": 0.000125, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..85f3220e1a64ec27d2f86441abde89ff07337bfe --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "sim", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 2.5e-05, + "sw_z_weight": 0.0, + "sw_s_y_weight": 5e-06, + "lambda_joint": 0.000125, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "sim", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.000125", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "2.5e-05", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "5e-06", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..d5657a8aebbbed00f091cb28df5ed832904e7f33 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "sim", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.000125, + "s_x": 2.5e-05, + "z": 0.0, + "s_y": 5e-06 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 50 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "sim", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "64", + "--model_num_layers", + "16", + "--model_num_heads", + "2", + "--model_dropout_prob", + "0.002", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.15", + "--mask_max_ctx_share", + "0.5", + "--mask_min_trgt_share", + "0.15", + "--mask_max_trgt_share", + "0.8", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.0005", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "1e-06", + "--vtjepa_kl_x_annealing_epochs", + "50", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "50", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "50", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.001", + "--recon_y_weight", + "0.1", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "p90", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.000125", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "50", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "2.5e-05", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "5e-06", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/downstream_results/autoint_sim_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/downstream_results/autoint_sim_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..8a469a4a5b544aef5a28380534ba6419c5e064e8 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/downstream_results/autoint_sim_downstream_results.json @@ -0,0 +1,210 @@ +{ + "dataset": "sim", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmplvvfqiav", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "sim", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmplvvfqiav/split_data/split_config.json", + "sigma_aggregation": "p90", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "sim_val_loss": 0.3782089054584503, + "lr": 0.0005626666243188083, + "sim_val_accuracy": 0.8637837767601013, + "sim_val_precision": 0.8643013834953308, + "sim_val_recall": 0.8637837767601013, + "sim_val_f1-score": 0.8637518882751465, + "sim_val_support": 925.0, + "sim_val_score": 0.8637837767601013 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.3815087378025055, + "lr": 0.0005626666243188083, + "sim_test_accuracy": 0.8552497625350952, + "sim_test_precision": 0.8556452989578247, + "sim_test_recall": 0.8552497625350952, + "sim_test_f1-score": 0.8553071618080139, + "sim_test_support": 1962.0, + "sim_test_score": 0.8552497625350952 + } + } + }, + { + "seed": 1, + "val_metrics": { + "sim_val_loss": 0.3487226963043213, + "lr": 0.0006545084179379046, + "sim_val_accuracy": 0.8670270442962646, + "sim_val_precision": 0.8671115636825562, + "sim_val_recall": 0.8670270442962646, + "sim_val_f1-score": 0.867058277130127, + "sim_val_support": 925.0, + "sim_val_score": 0.8670270442962646 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.3803064227104187, + "lr": 0.0006545084179379046, + "sim_test_accuracy": 0.8516819477081299, + "sim_test_precision": 0.8529751896858215, + "sim_test_recall": 0.8516819477081299, + "sim_test_f1-score": 0.8517122268676758, + "sim_test_support": 1962.0, + "sim_test_score": 0.8516819477081299 + } + } + }, + { + "seed": 2, + "val_metrics": { + "sim_val_loss": 0.3879034221172333, + "lr": 0.0003454915131442249, + "sim_val_accuracy": 0.8486486673355103, + "sim_val_precision": 0.8506736159324646, + "sim_val_recall": 0.8486486673355103, + "sim_val_f1-score": 0.8488385081291199, + "sim_val_support": 925.0, + "sim_val_score": 0.8486486673355103 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.3959185779094696, + "lr": 0.0003454914840403944, + "sim_test_accuracy": 0.8450560569763184, + "sim_test_precision": 0.8454021215438843, + "sim_test_recall": 0.8450560569763184, + "sim_test_f1-score": 0.8451074957847595, + "sim_test_support": 1962.0, + "sim_test_score": 0.8450560569763184 + } + } + }, + { + "seed": 3, + "val_metrics": { + "sim_val_loss": 0.3645848333835602, + "lr": 0.000437333423178643, + "sim_val_accuracy": 0.8659459352493286, + "sim_val_precision": 0.8673906922340393, + "sim_val_recall": 0.8659459352493286, + "sim_val_f1-score": 0.8660265207290649, + "sim_val_support": 925.0, + "sim_val_score": 0.8659459352493286 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.38190674781799316, + "lr": 0.0004373333358671516, + "sim_test_accuracy": 0.857798159122467, + "sim_test_precision": 0.8591137528419495, + "sim_test_recall": 0.857798159122467, + "sim_test_f1-score": 0.8578699231147766, + "sim_test_support": 1962.0, + "sim_test_score": 0.857798159122467 + } + } + }, + { + "seed": 4, + "val_metrics": { + "sim_val_loss": 0.38530775904655457, + "lr": 0.0006243449170142412, + "sim_val_accuracy": 0.846486508846283, + "sim_val_precision": 0.8505094051361084, + "sim_val_recall": 0.846486508846283, + "sim_val_f1-score": 0.8458940386772156, + "sim_val_support": 925.0, + "sim_val_score": 0.846486508846283 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.40989184379577637, + "lr": 0.0006243448005989194, + "sim_test_accuracy": 0.8379204869270325, + "sim_test_precision": 0.8446603417396545, + "sim_test_recall": 0.8379204869270325, + "sim_test_f1-score": 0.8370776772499084, + "sim_test_support": 1962.0, + "sim_test_score": 0.8379204869270325 + } + } + } + ], + "summary": { + "0": { + "sim_test_loss": { + "mean": 0.38990646600723267, + "std": 0.012867002140011649, + "n": 5, + "ci95": 0.015973932053984567 + }, + "lr": { + "mean": 0.0005248689325526356, + "std": 0.00013035457387373782, + "n": 5, + "ci95": 0.00016183063337730312 + }, + "sim_test_accuracy": { + "mean": 0.8495412826538086, + "std": 0.00807008892686456, + "n": 5, + "ci95": 0.01001873247432529 + }, + "sim_test_precision": { + "mean": 0.8515593409538269, + "std": 0.00634973149898988, + "n": 5, + "ci95": 0.007882969041444341 + }, + "sim_test_recall": { + "mean": 0.8495412826538086, + "std": 0.00807008892686456, + "n": 5, + "ci95": 0.01001873247432529 + }, + "sim_test_f1-score": { + "mean": 0.8494148969650268, + "std": 0.00840006952473212, + "n": 5, + "ci95": 0.010428391817823686 + }, + "sim_test_support": { + "mean": 1962.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "sim_test_score": { + "mean": 0.8495412826538086, + "std": 0.00807008892686456, + "n": 5, + "ci95": 0.01001873247432529 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/downstream_results/dcnv2_sim_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/downstream_results/dcnv2_sim_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..81123165df6eb2664724e8d4f6c73dd7721c93f8 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/downstream_results/dcnv2_sim_downstream_results.json @@ -0,0 +1,209 @@ +{ + "dataset": "sim", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp17sz61l4", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "sim", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp17sz61l4/split_data/split_config.json", + "sigma_aggregation": "p90", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "sim_val_loss": 0.39409396052360535, + "lr": 0.0004686047323048115, + "sim_val_accuracy": 0.8508108258247375, + "sim_val_precision": 0.8526347875595093, + "sim_val_recall": 0.8508108258247375, + "sim_val_f1-score": 0.8505774736404419, + "sim_val_support": 925.0, + "sim_val_score": 0.8508108258247375 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.3970380425453186, + "lr": 0.0004686047323048115, + "sim_test_accuracy": 0.8486238718032837, + "sim_test_precision": 0.849159836769104, + "sim_test_recall": 0.8486238718032837, + "sim_test_f1-score": 0.8483309745788574, + "sim_test_support": 1962.0, + "sim_test_score": 0.8486238718032837 + } + } + }, + { + "seed": 1, + "val_metrics": { + "sim_val_loss": 0.3911454677581787, + "lr": 0.0005000000237487257, + "sim_val_accuracy": 0.8551351428031921, + "sim_val_precision": 0.8585590720176697, + "sim_val_recall": 0.8551351428031921, + "sim_val_f1-score": 0.8554569482803345, + "sim_val_support": 925.0, + "sim_val_score": 0.8551351428031921 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.4121691882610321, + "lr": 0.0005000000819563866, + "sim_test_accuracy": 0.8455657362937927, + "sim_test_precision": 0.8497406840324402, + "sim_test_recall": 0.8455657362937927, + "sim_test_f1-score": 0.8454813957214355, + "sim_test_support": 1962.0, + "sim_test_score": 0.8455657362937927 + } + } + }, + { + "seed": 2, + "val_metrics": { + "sim_val_loss": 0.37148424983024597, + "lr": 0.0006243449170142412, + "sim_val_accuracy": 0.8572973012924194, + "sim_val_precision": 0.8585994243621826, + "sim_val_recall": 0.8572973012924194, + "sim_val_f1-score": 0.8569865822792053, + "sim_val_support": 925.0, + "sim_val_score": 0.8572973012924194 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.39458781480789185, + "lr": 0.0006243448005989194, + "sim_test_accuracy": 0.8501529097557068, + "sim_test_precision": 0.850886881351471, + "sim_test_recall": 0.8501529097557068, + "sim_test_f1-score": 0.849794864654541, + "sim_test_support": 1962.0, + "sim_test_score": 0.8501529097557068 + } + } + }, + { + "seed": 3, + "val_metrics": { + "sim_val_loss": 0.37747451663017273, + "lr": 0.0005000000237487257, + "sim_val_accuracy": 0.8540540337562561, + "sim_val_precision": 0.8552247285842896, + "sim_val_recall": 0.8540540337562561, + "sim_val_f1-score": 0.8541227579116821, + "sim_val_support": 925.0, + "sim_val_score": 0.8540540337562561 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.40048158168792725, + "lr": 0.0005000000819563866, + "sim_test_accuracy": 0.8455657362937927, + "sim_test_precision": 0.8478516340255737, + "sim_test_recall": 0.8455657362937927, + "sim_test_f1-score": 0.8454481363296509, + "sim_test_support": 1962.0, + "sim_test_score": 0.8455657362937927 + } + } + }, + { + "seed": 4, + "val_metrics": { + "sim_val_loss": 0.4187757670879364, + "lr": 0.0003159377083647996, + "sim_val_accuracy": 0.8594594597816467, + "sim_val_precision": 0.8633977770805359, + "sim_val_recall": 0.8594594597816467, + "sim_val_f1-score": 0.8598502278327942, + "sim_val_support": 925.0, + "sim_val_score": 0.8594594597816467 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.44202905893325806, + "lr": 0.0003159377083647996, + "sim_test_accuracy": 0.8486238718032837, + "sim_test_precision": 0.8506619930267334, + "sim_test_recall": 0.8486238718032837, + "sim_test_f1-score": 0.8488503694534302, + "sim_test_support": 1962.0, + "sim_test_score": 0.8486238718032837 + } + } + } + ], + "summary": { + "0": { + "sim_test_loss": { + "mean": 0.40926113724708557, + "std": 0.019519121929790453, + "n": 5, + "ci95": 0.024232305557044895 + }, + "lr": { + "mean": 0.0004817774810362607, + "std": 0.0001103017661743607, + "n": 5, + "ci95": 0.0001369357756477482 + }, + "sim_test_accuracy": { + "mean": 0.847706425189972, + "std": 0.0020514507797765465, + "n": 5, + "ci95": 0.002546804221501011 + }, + "sim_test_precision": { + "mean": 0.8496602058410645, + "std": 0.0012284225995611829, + "n": 5, + "ci95": 0.0015250435902198166 + }, + "sim_test_recall": { + "mean": 0.847706425189972, + "std": 0.0020514507797765465, + "n": 5, + "ci95": 0.002546804221501011 + }, + "sim_test_f1-score": { + "mean": 0.847581148147583, + "std": 0.002002024153199878, + "n": 5, + "ci95": 0.002485442797448846 + }, + "sim_test_support": { + "mean": 1962.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "sim_test_score": { + "mean": 0.847706425189972, + "std": 0.0020514507797765465, + "n": 5, + "ci95": 0.002546804221501011 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/downstream_results/fttransformer_sim_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/downstream_results/fttransformer_sim_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..fbc5ee4c42537601bb01fd8efef78cfcf1db07b1 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/downstream_results/fttransformer_sim_downstream_results.json @@ -0,0 +1,209 @@ +{ + "dataset": "sim", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpq49ttmv6", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "sim", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpq49ttmv6/split_data/split_config.json", + "sigma_aggregation": "p90", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "sim_val_loss": 0.3959009349346161, + "lr": 4.7586476284777746e-05, + "sim_val_accuracy": 0.8518918752670288, + "sim_val_precision": 0.8521439433097839, + "sim_val_recall": 0.8518918752670288, + "sim_val_f1-score": 0.8519467711448669, + "sim_val_support": 925.0, + "sim_val_score": 0.8518918752670288 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.40963634848594666, + "lr": 4.758647264679894e-05, + "sim_test_accuracy": 0.8542304039001465, + "sim_test_precision": 0.8546965718269348, + "sim_test_recall": 0.8542304039001465, + "sim_test_f1-score": 0.8542208671569824, + "sim_test_support": 1962.0, + "sim_test_score": 0.8542304039001465 + } + } + }, + { + "seed": 1, + "val_metrics": { + "sim_val_loss": 0.3818179666996002, + "lr": 0.0003454915131442249, + "sim_val_accuracy": 0.8583783507347107, + "sim_val_precision": 0.8593643307685852, + "sim_val_recall": 0.8583783507347107, + "sim_val_f1-score": 0.8585767149925232, + "sim_val_support": 925.0, + "sim_val_score": 0.8583783507347107 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.40157413482666016, + "lr": 0.0003454914840403944, + "sim_test_accuracy": 0.8557594418525696, + "sim_test_precision": 0.8570339679718018, + "sim_test_recall": 0.8557594418525696, + "sim_test_f1-score": 0.8557974696159363, + "sim_test_support": 1962.0, + "sim_test_score": 0.8557594418525696 + } + } + }, + { + "seed": 2, + "val_metrics": { + "sim_val_loss": 0.38185203075408936, + "lr": 0.00018128800729755312, + "sim_val_accuracy": 0.861621618270874, + "sim_val_precision": 0.863429069519043, + "sim_val_recall": 0.861621618270874, + "sim_val_f1-score": 0.8616964221000671, + "sim_val_support": 925.0, + "sim_val_score": 0.861621618270874 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.40728646516799927, + "lr": 0.00018128800729755312, + "sim_test_accuracy": 0.847604513168335, + "sim_test_precision": 0.8510030508041382, + "sim_test_recall": 0.847604513168335, + "sim_test_f1-score": 0.8475886583328247, + "sim_test_support": 1962.0, + "sim_test_score": 0.847604513168335 + } + } + }, + { + "seed": 3, + "val_metrics": { + "sim_val_loss": 0.37150830030441284, + "lr": 0.00023208661878015846, + "sim_val_accuracy": 0.861621618270874, + "sim_val_precision": 0.8645786643028259, + "sim_val_recall": 0.861621618270874, + "sim_val_f1-score": 0.8617915511131287, + "sim_val_support": 925.0, + "sim_val_score": 0.861621618270874 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.3833673298358917, + "lr": 0.00023208657512441278, + "sim_test_accuracy": 0.853210985660553, + "sim_test_precision": 0.8575186133384705, + "sim_test_recall": 0.853210985660553, + "sim_test_f1-score": 0.8534896969795227, + "sim_test_support": 1962.0, + "sim_test_score": 0.853210985660553 + } + } + }, + { + "seed": 4, + "val_metrics": { + "sim_val_loss": 0.37652772665023804, + "lr": 0.0003454915131442249, + "sim_val_accuracy": 0.8594594597816467, + "sim_val_precision": 0.8626759648323059, + "sim_val_recall": 0.8594594597816467, + "sim_val_f1-score": 0.8596131801605225, + "sim_val_support": 925.0, + "sim_val_score": 0.8594594597816467 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.37782225012779236, + "lr": 0.0003454914840403944, + "sim_test_accuracy": 0.8552497625350952, + "sim_test_precision": 0.8565954566001892, + "sim_test_recall": 0.8552497625350952, + "sim_test_f1-score": 0.8553430438041687, + "sim_test_support": 1962.0, + "sim_test_score": 0.8552497625350952 + } + } + } + ], + "summary": { + "0": { + "sim_test_loss": { + "mean": 0.39593730568885804, + "std": 0.014442996296586751, + "n": 5, + "ci95": 0.017930473546763446 + }, + "lr": { + "mean": 0.00023038880462991073, + "std": 0.00012482759802697304, + "n": 5, + "ci95": 0.0001549690866331946 + }, + "sim_test_accuracy": { + "mean": 0.8532110214233398, + "std": 0.003283404766915523, + "n": 5, + "ci95": 0.004076231905592123 + }, + "sim_test_precision": { + "mean": 0.8553695321083069, + "std": 0.002665090160636475, + "n": 5, + "ci95": 0.0033086159993217603 + }, + "sim_test_recall": { + "mean": 0.8532110214233398, + "std": 0.003283404766915523, + "n": 5, + "ci95": 0.004076231905592123 + }, + "sim_test_f1-score": { + "mean": 0.8532879471778869, + "std": 0.0033133842600781365, + "n": 5, + "ci95": 0.00411345039530569 + }, + "sim_test_support": { + "mean": 1962.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "sim_test_score": { + "mean": 0.8532110214233398, + "std": 0.003283404766915523, + "n": 5, + "ci95": 0.004076231905592123 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/downstream_results/mlp_sim_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/downstream_results/mlp_sim_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..e5a61022fc859af0665eef55b83023583c89e05e --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/downstream_results/mlp_sim_downstream_results.json @@ -0,0 +1,205 @@ +{ + "dataset": "sim", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpwa7rrbwr", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "sim", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpwa7rrbwr/split_data/split_config.json", + "sigma_aggregation": "p90", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "sim_val_loss": 0.4164837896823883, + "lr": 0.0010000000474974513, + "sim_val_accuracy": 0.8335134983062744, + "sim_val_precision": 0.8485895991325378, + "sim_val_recall": 0.8335134983062744, + "sim_val_f1-score": 0.8329811096191406, + "sim_val_support": 925.0, + "sim_val_score": 0.8335134983062744 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.4374564290046692, + "lr": 0.0010000001639127731, + "sim_test_accuracy": 0.8272171020507812, + "sim_test_precision": 0.8430460095405579, + "sim_test_recall": 0.8272171020507812, + "sim_test_f1-score": 0.8264910578727722, + "sim_test_support": 1962.0, + "sim_test_score": 0.8272171020507812 + } + } + }, + { + "seed": 1, + "val_metrics": { + "sim_val_loss": 0.3836711347103119, + "lr": 0.0010000000474974513, + "sim_val_accuracy": 0.8486486673355103, + "sim_val_precision": 0.8551791906356812, + "sim_val_recall": 0.8486486673355103, + "sim_val_f1-score": 0.8474687337875366, + "sim_val_support": 925.0, + "sim_val_score": 0.8486486673355103 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.415079802274704, + "lr": 0.0010000001639127731, + "sim_test_accuracy": 0.8307849168777466, + "sim_test_precision": 0.8409544825553894, + "sim_test_recall": 0.8307849168777466, + "sim_test_f1-score": 0.8291705846786499, + "sim_test_support": 1962.0, + "sim_test_score": 0.8307849168777466 + } + } + }, + { + "seed": 2, + "val_metrics": { + "sim_val_loss": 0.42745241522789, + "lr": 0.0010000000474974513, + "sim_val_accuracy": 0.8270270228385925, + "sim_val_precision": 0.8492650985717773, + "sim_val_recall": 0.8270270228385925, + "sim_val_f1-score": 0.8267048001289368, + "sim_val_support": 925.0, + "sim_val_score": 0.8270270228385925 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.43095648288726807, + "lr": 0.0010000001639127731, + "sim_test_accuracy": 0.819571852684021, + "sim_test_precision": 0.8391041159629822, + "sim_test_recall": 0.819571852684021, + "sim_test_f1-score": 0.8181103467941284, + "sim_test_support": 1962.0, + "sim_test_score": 0.819571852684021 + } + } + }, + { + "seed": 3, + "val_metrics": { + "sim_val_loss": 0.35156771540641785, + "lr": 0.0010000000474974513, + "sim_val_accuracy": 0.8627027273178101, + "sim_val_precision": 0.8636327981948853, + "sim_val_recall": 0.8627027273178101, + "sim_val_f1-score": 0.8627711534500122, + "sim_val_support": 925.0, + "sim_val_score": 0.8627027273178101 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.3588317930698395, + "lr": 0.0010000001639127731, + "sim_test_accuracy": 0.856269121170044, + "sim_test_precision": 0.8563961386680603, + "sim_test_recall": 0.856269121170044, + "sim_test_f1-score": 0.8562073111534119, + "sim_test_support": 1962.0, + "sim_test_score": 0.856269121170044 + } + } + }, + { + "seed": 4, + "val_metrics": { + "sim_val_loss": 0.37364333868026733, + "lr": 0.0010000000474974513, + "sim_val_accuracy": 0.8583783507347107, + "sim_val_precision": 0.8640233278274536, + "sim_val_recall": 0.8583783507347107, + "sim_val_f1-score": 0.8584708571434021, + "sim_val_support": 925.0, + "sim_val_score": 0.8583783507347107 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.3680511713027954, + "lr": 0.0010000001639127731, + "sim_test_accuracy": 0.8567788004875183, + "sim_test_precision": 0.8608500957489014, + "sim_test_recall": 0.8567788004875183, + "sim_test_f1-score": 0.8569513559341431, + "sim_test_support": 1962.0, + "sim_test_score": 0.8567788004875183 + } + } + } + ], + "summary": { + "0": { + "sim_test_loss": { + "mean": 0.40207513570785525, + "std": 0.03634111417909979, + "n": 5, + "ci95": 0.045116219174150984 + }, + "lr": { + "mean": 0.0010000001639127731, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "sim_test_accuracy": { + "mean": 0.8381243586540222, + "std": 0.017278973760363567, + "n": 5, + "ci95": 0.02145124014181441 + }, + "sim_test_precision": { + "mean": 0.8480701684951782, + "std": 0.009860445952583445, + "n": 5, + "ci95": 0.01224139795382147 + }, + "sim_test_recall": { + "mean": 0.8381243586540222, + "std": 0.017278973760363567, + "n": 5, + "ci95": 0.02145124014181441 + }, + "sim_test_f1-score": { + "mean": 0.8373861312866211, + "std": 0.01799158501666832, + "n": 5, + "ci95": 0.022335922033155608 + }, + "sim_test_support": { + "mean": 1962.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "sim_test_score": { + "mean": 0.8381243586540222, + "std": 0.017278973760363567, + "n": 5, + "ci95": 0.02145124014181441 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/downstream_results/resnet_sim_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/downstream_results/resnet_sim_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..45a9ab64386f009e00154a016429197b82647721 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/downstream_results/resnet_sim_downstream_results.json @@ -0,0 +1,208 @@ +{ + "dataset": "sim", + "model": "resnet", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmps2944q7b", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "sim", + "model_name": "resnet", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmps2944q7b/split_data/split_config.json", + "sigma_aggregation": "p90", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_out": 128, + "d_block": 128, + "d_hidden": 256, + "dropout1": 0.1, + "dropout2": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "sim_val_loss": 0.4334232807159424, + "lr": 9.549150126986206e-05, + "sim_val_accuracy": 0.8508108258247375, + "sim_val_precision": 0.8509776592254639, + "sim_val_recall": 0.8508108258247375, + "sim_val_f1-score": 0.8508463501930237, + "sim_val_support": 925.0, + "sim_val_score": 0.8508108258247375 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.43478503823280334, + "lr": 9.549151582177728e-05, + "sim_test_accuracy": 0.8379204869270325, + "sim_test_precision": 0.8381434679031372, + "sim_test_recall": 0.8379204869270325, + "sim_test_f1-score": 0.8378437161445618, + "sim_test_support": 1962.0, + "sim_test_score": 0.8379204869270325 + } + } + }, + { + "seed": 1, + "val_metrics": { + "sim_val_loss": 0.41572412848472595, + "lr": 9.549150126986206e-05, + "sim_val_accuracy": 0.838918924331665, + "sim_val_precision": 0.8439152240753174, + "sim_val_recall": 0.838918924331665, + "sim_val_f1-score": 0.8383347392082214, + "sim_val_support": 925.0, + "sim_val_score": 0.838918924331665 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.44139349460601807, + "lr": 9.549151582177728e-05, + "sim_test_accuracy": 0.8419979810714722, + "sim_test_precision": 0.8466127514839172, + "sim_test_recall": 0.8419979810714722, + "sim_test_f1-score": 0.8414115309715271, + "sim_test_support": 1962.0, + "sim_test_score": 0.8419979810714722 + } + } + }, + { + "seed": 2, + "val_metrics": { + "sim_val_loss": 0.3817296028137207, + "lr": 0.0006545084179379046, + "sim_val_accuracy": 0.8345946073532104, + "sim_val_precision": 0.8351615071296692, + "sim_val_recall": 0.8345946073532104, + "sim_val_f1-score": 0.8346790075302124, + "sim_val_support": 925.0, + "sim_val_score": 0.8345946073532104 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.3901840150356293, + "lr": 0.0006545084179379046, + "sim_test_accuracy": 0.8419979810714722, + "sim_test_precision": 0.842574417591095, + "sim_test_recall": 0.8419979810714722, + "sim_test_f1-score": 0.8418765068054199, + "sim_test_support": 1962.0, + "sim_test_score": 0.8419979810714722 + } + } + }, + { + "seed": 3, + "val_metrics": { + "sim_val_loss": 0.4370410740375519, + "lr": 9.549150126986206e-05, + "sim_val_accuracy": 0.8432432413101196, + "sim_val_precision": 0.8439783453941345, + "sim_val_recall": 0.8432432413101196, + "sim_val_f1-score": 0.8430055975914001, + "sim_val_support": 925.0, + "sim_val_score": 0.8432432413101196 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.44849780201911926, + "lr": 9.549151582177728e-05, + "sim_test_accuracy": 0.8363914489746094, + "sim_test_precision": 0.8383257389068604, + "sim_test_recall": 0.8363914489746094, + "sim_test_f1-score": 0.8359894752502441, + "sim_test_support": 1962.0, + "sim_test_score": 0.8363914489746094 + } + } + }, + { + "seed": 4, + "val_metrics": { + "sim_val_loss": 0.4372842013835907, + "lr": 9.549150126986206e-05, + "sim_val_accuracy": 0.8237837553024292, + "sim_val_precision": 0.8270774483680725, + "sim_val_recall": 0.8237837553024292, + "sim_val_f1-score": 0.8239543437957764, + "sim_val_support": 925.0, + "sim_val_score": 0.8237837553024292 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.4394420087337494, + "lr": 9.549151582177728e-05, + "sim_test_accuracy": 0.8353720903396606, + "sim_test_precision": 0.8381948471069336, + "sim_test_recall": 0.8353720903396606, + "sim_test_f1-score": 0.8357051014900208, + "sim_test_support": 1962.0, + "sim_test_score": 0.8353720903396606 + } + } + } + ], + "summary": { + "0": { + "sim_test_loss": { + "mean": 0.4308604717254639, + "std": 0.023268213937821202, + "n": 5, + "ci95": 0.028886671846002823 + }, + "lr": { + "mean": 0.00020729489624500276, + "std": 0.00024999995874060133, + "n": 5, + "ci95": 0.00031036618405487384 + }, + "sim_test_accuracy": { + "mean": 0.8387359976768494, + "std": 0.0031128327254698433, + "n": 5, + "ci95": 0.00386447269620411 + }, + "sim_test_precision": { + "mean": 0.8407702445983887, + "std": 0.0037715435506130365, + "n": 5, + "ci95": 0.0046822390919476296 + }, + "sim_test_recall": { + "mean": 0.8387359976768494, + "std": 0.0031128327254698433, + "n": 5, + "ci95": 0.00386447269620411 + }, + "sim_test_f1-score": { + "mean": 0.8385652661323547, + "std": 0.0029326358947274578, + "n": 5, + "ci95": 0.0036407646483386873 + }, + "sim_test_support": { + "mean": 1962.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "sim_test_score": { + "mean": 0.8387359976768494, + "std": 0.0031128327254698433, + "n": 5, + "ci95": 0.00386447269620411 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/embedding.npy b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/embedding.npy new file mode 100644 index 0000000000000000000000000000000000000000..b8cf0d7aef202673d04a03f48a32e3c266f80d4a --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/embedding.npy @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:be7eadb7723032019a035da16174f3bca6a6df0e8a8e1f5d5546e8872480a0dc +size 81920128 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/vtjepa_best_model.pth b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/vtjepa_best_model.pth new file mode 100644 index 0000000000000000000000000000000000000000..1a3be6cb57fe3d3518ff8ddf0947363fc888746a --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/vtjepa_best_model.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cd5791d3ad0009164e3f93380024de27c57845629f4de5d5afc5998b148bcdb7 +size 15815126 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/vtjepa_config.yaml b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/vtjepa_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..87caad641e9c91af9d91ecdde50d89a7d099de31 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.000125_sx2.5e-05_sy5e-06/seed_42/wjepa_deterministic/sim/20260916_015515/vtjepa_config.yaml @@ -0,0 +1,130 @@ +activation_checkpointing: false +batch_size: 512 +data_loader_nprocs: 0 +data_path: /pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets +data_set: sim +deterministic_mode: warn +downstream_evaluation_epochs: 50 +downstream_probe_patience: 20 +exp_cadence_type: improvement +exp_device: null +exp_final_lr: 0.0 +exp_final_weight_decay: 0.0 +exp_gradient_clipping: 1.0 +exp_ipe_scale: 1.0 +exp_lr: 0.0005 +exp_n_runs: 1 +exp_patience: 15 +exp_scheduler: true +exp_start_lr: 0.0 +exp_train_total_epochs: 500 +exp_val_cache_cadence: 1 +exp_warmup: 10 +exp_weight_decay: 0.0 +exp_weight_decay_scheduler: true +full_dataset_cuda: false +init_type: trunc_normal +is_batchlearning: true +iteration: 0 +jepa_objective: w_jepa_joint +latent_architecture: deterministic_no_z +load_from_checkpoint: false +load_path: null +log_sigma_x_init: -2.0 +log_sigma_y_init: -2.0 +log_tensorboard: false +logvar_init_bias: -4.0 +logvar_init_std: 0.005 +mask_allow_overlap: false +mask_max_ctx_share: 0.5 +mask_max_trgt_share: 0.8 +mask_min_ctx_share: 0.15 +mask_min_trgt_share: 0.15 +mask_num_encs: 1 +mask_num_preds: 4 +mask_seed: 0 +mock: false +model_act_func: relu +model_amp: false +model_dim_feedforward: 64 +model_dim_hidden: 64 +model_dropout_prob: 0.002 +model_dtype: float32 +model_feature_index_embedding: true +model_feature_type_embedding: true +model_layer_norm_eps: 1.0e-05 +model_num_heads: 2 +model_num_layers: 16 +model_selection: unfiltered +mp_distributed: false +mp_gpus: 1 +mp_nodes: 1 +mp_nr: 0 +mu_init_std: 0.03 +n_cls_tokens: 1 +np_seed: 42 +pin_memory: false +pooled_context_representation_size_for_auxiliary_encoder: 1 +pooled_context_representation_size_for_target_posterior: 1 +pre_evaluation: true +pred_activation: relu +pred_dim_feedforward: 256 +pred_embed_dim: 16 +pred_layer_norm_eps: 1.0e-05 +pred_num_heads: 4 +pred_num_layers: 16 +pred_p_dropout: 0.002 +pred_type: transformer +probe_cadence: 2 +probe_encoder_type: target_posterior +probe_model: linear_probe +probe_representation_type: mean +random: false +recon_x_weight: 0.001 +recon_y_weight: 0.1 +res_dir: ./ +safe_exit: false +sigma_aggregation: p90 +subset_size: -1 +sw_full_num_projections: 2048 +sw_include_marginals: true +sw_marginal_num_projections: 1024 +sw_p: 2 +sw_s_x_weight: 2.5e-05 +sw_s_y_weight: 5.0e-06 +sw_seed: 0 +sw_z_weight: 0.0 +test: false +torch_seed: 42 +using_embedding: false +verbose: false +vtjepa_kl_x_annealing_epochs: 50 +vtjepa_kl_x_end_weight: 1.0e-06 +vtjepa_kl_x_start_epoch: 1 +vtjepa_kl_x_start_weight: 0.0 +vtjepa_kl_y_annealing_epochs: 50 +vtjepa_kl_y_end_weight: 1.0e-05 +vtjepa_kl_y_start_epoch: 1 +vtjepa_kl_y_start_weight: 0.0 +vtjepa_kl_z_annealing_epochs: 50 +vtjepa_kl_z_end_weight: 1.0e-06 +vtjepa_kl_z_start_epoch: 1 +vtjepa_kl_z_start_weight: 0.0 +wjepa_joint_annealing_epochs: 50 +wjepa_joint_end_weight: 0.000125 +wjepa_joint_start_epoch: 1 +wjepa_joint_start_weight: 0.0 +wjepa_marginal_end_scale: 1.0 +wjepa_marginal_start_scale: 0.0 +wjepa_s_x_annealing_epochs: 30 +wjepa_s_x_end_scale: 1.0 +wjepa_s_x_start_epoch: 0 +wjepa_s_x_start_scale: 0.0 +wjepa_s_y_annealing_epochs: 30 +wjepa_s_y_end_scale: 1.0 +wjepa_s_y_start_epoch: 0 +wjepa_s_y_start_scale: 0.0 +wjepa_z_annealing_epochs: 0 +wjepa_z_end_scale: 0.0 +wjepa_z_start_epoch: 0 +wjepa_z_start_scale: 0.0 diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/comparison_complete.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/comparison_complete.json new file mode 100644 index 0000000000000000000000000000000000000000..d79d0b4d5c4709998c1a92cd855024144baf62e6 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/comparison_complete.json @@ -0,0 +1,31 @@ +{ + "completed_at_utc": "2026-09-16T02:33:00.743249+00:00", + "spec": { + "dataset": "sim", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 2.5e-05, + "sw_z_weight": 0.0, + "sw_s_y_weight": 1e-05, + "lambda_joint": 0.00025, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/comparison_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/comparison_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..7d88dcd54794011ee5fd30bca1464eda06dafa69 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/comparison_manifest.json @@ -0,0 +1,73 @@ +{ + "spec": { + "dataset": "sim", + "objective": "w_jepa_joint", + "torch_seed": 42, + "np_seed": 42, + "mask_seed": 0, + "sw_seed": 0, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_p": 2, + "sw_include_marginals": true, + "sw_s_x_weight": 2.5e-05, + "sw_z_weight": 0.0, + "sw_s_y_weight": 1e-05, + "lambda_joint": 0.00025, + "sw_full_start_epoch": null, + "sw_full_annealing_epochs": null, + "sw_s_x_start_epoch": null, + "sw_s_x_annealing_epochs": null, + "sw_z_start_epoch": null, + "sw_z_annealing_epochs": null, + "sw_s_y_start_epoch": null, + "sw_s_y_annealing_epochs": null, + "epochs": 500, + "smoke": false, + "downstream_models": "mlp,dcnv2,resnet,autoint,fttransformer", + "latent_architecture": "deterministic_no_z" + }, + "train_command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/runners/hessian_train_vtjepa.py", + "--dataset", + "sim", + "--results-root", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42", + "--epochs", + "500", + "--objective", + "w_jepa_joint", + "--latent-architecture", + "deterministic_no_z", + "--sw-full-end-weight", + "0.00025", + "--torch-seed", + "42", + "--np-seed", + "42", + "--mask-seed", + "0", + "--sw-seed", + "0", + "--sw-full-num-projections", + "2048", + "--sw-marginal-num-projections", + "1024", + "--sw-p", + "2", + "--sw-s-x-end-weight", + "2.5e-05", + "--sw-z-end-weight", + "0.0", + "--sw-s-y-end-weight", + "1e-05", + "--model-selection", + "unfiltered", + "--deterministic-mode", + "warn", + "--sw-include-marginals", + "--batch-size", + "512" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/paper_alignment_manifest.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/paper_alignment_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..0902e74e7de9ba41c9b64d50e7169bd90fc98243 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/paper_alignment_manifest.json @@ -0,0 +1,239 @@ +{ + "code_variant": "paper-aligned-v5-flattened-block-pooled-sw-deterministic-no-z", + "upstream_commit": "eba7a9b1e01bb6613f4968b0f460d1bd1ae409ef", + "source_sha256": { + "src/encoder.py": "c29e003ed6a59e239cfe5cbc6823817fed99c54bfa2f4ffa94bc9ea8eb6f777c", + "src/mask.py": "e7f57ec6d07cbc6a2d9bc3ef8b48dbaeae8279ecb676b0425226675cbf42c437", + "src/run_vtjepa.py": "2f44128407503aee95afb939be1faeed557de3124783e7d116a50f53eaf20dc2", + "src/train_vtjepa.py": "60011dfabb086ae93bb0bb56db2057ca9c1009f2d7fb1a13fcbddb212fbbb1f5", + "src/vtjepa.py": "3d11b290bc62128f969661cb5600129da45a5b6f1b9afb598350f026584f92d5", + "src/d_jepa.py": "7e86669cb99f31b2460849ce67589e6e2b28a1aee49e443614439cc4fc4d2c0b", + "src/jepa_model.py": "69f08f390513d6f90566dcb6ea79743143bdaa46020493582f47b4d5cb278927", + "src/vtjepa_configs.py": "1be90275f2e0b4e00d8f7a61bc6ad8a94c1d21457b8a8611dc87c15fa871ae76", + "src/downstream_pipeline.py": "13e4032713b8fddde763d5310cef40aaf98aeb2dde43d5b2a234cb2f4a60edc5", + "src/sliced_wasserstein.py": "2577a99ea0e5c7c2239cd1670ca9d1833f0fd87b9c11754264afd07a33b13053" + }, + "dataset": "sim", + "objective": "w_jepa_joint", + "latent_architecture": "deterministic_no_z", + "model_selection": "unfiltered", + "seeds": { + "torch": 42, + "numpy": 42, + "mask": 0, + "sliced_wasserstein": 0 + }, + "smoke": false, + "sw_include_marginals": true, + "sw_full_num_projections": 2048, + "sw_marginal_num_projections": 1024, + "sw_component_weights": { + "full": 0.00025, + "s_x": 2.5e-05, + "z": 0.0, + "s_y": 1e-05 + }, + "sw_annealing_schedule": { + "full": { + "start_epoch": 1, + "annealing_epochs": 50 + }, + "s_x": { + "start_epoch": 0, + "annealing_epochs": 30 + }, + "z": { + "start_epoch": 0, + "annealing_epochs": 0 + }, + "s_y": { + "start_epoch": 0, + "annealing_epochs": 30 + } + }, + "command": [ + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/.venv/bin/python", + "-W", + "ignore", + "-m", + "src.run_vtjepa", + "--data_set", + "sim", + "--data_path", + "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "--batch_size", + "512", + "--model_dim_hidden", + "64", + "--model_dim_feedforward", + "64", + "--model_num_layers", + "16", + "--model_num_heads", + "2", + "--model_dropout_prob", + "0.002", + "--model_feature_index_embedding", + "true", + "--model_feature_type_embedding", + "true", + "--exp_train_total_epochs", + "500", + "--init_type", + "trunc_normal", + "--pre_evaluation", + "true", + "--pin_memory", + "false", + "--mask_min_ctx_share", + "0.15", + "--mask_max_ctx_share", + "0.5", + "--mask_min_trgt_share", + "0.15", + "--mask_max_trgt_share", + "0.8", + "--mask_num_preds", + "4", + "--pred_num_layers", + "16", + "--pred_embed_dim", + "16", + "--pred_num_heads", + "4", + "--pred_p_dropout", + "0.002", + "--pred_dim_feedforward", + "256", + "--pred_type", + "transformer", + "--probe_model", + "linear_probe", + "--probe_cadence", + "2", + "--probe_representation_type", + "mean", + "--probe_encoder_type", + "target_posterior", + "--downstream_evaluation_epochs", + "50", + "--downstream_probe_patience", + "20", + "--exp_final_lr", + "0.0", + "--exp_final_weight_decay", + "0.0", + "--exp_start_lr", + "0.0", + "--exp_patience", + "15", + "--exp_weight_decay", + "0.0", + "--exp_lr", + "0.0005", + "--exp_warmup", + "10", + "--log_tensorboard", + "false", + "--n_cls_tokens", + "1", + "--vtjepa_kl_x_start_weight", + "0", + "--vtjepa_kl_x_end_weight", + "1e-06", + "--vtjepa_kl_x_annealing_epochs", + "50", + "--vtjepa_kl_z_start_weight", + "0", + "--vtjepa_kl_z_end_weight", + "1e-06", + "--vtjepa_kl_z_annealing_epochs", + "50", + "--vtjepa_kl_y_start_weight", + "0", + "--vtjepa_kl_y_end_weight", + "1e-05", + "--vtjepa_kl_y_annealing_epochs", + "50", + "--vtjepa_kl_x_start_epoch", + "1", + "--vtjepa_kl_z_start_epoch", + "1", + "--vtjepa_kl_y_start_epoch", + "1", + "--recon_x_weight", + "0.001", + "--recon_y_weight", + "0.1", + "--mu_init_std", + "0.03", + "--logvar_init_std", + "0.005", + "--logvar_init_bias", + "-4.0", + "--log_sigma_x_init", + "-2.0", + "--log_sigma_y_init", + "-2.0", + "--pooled_context_representation_size_for_target_posterior", + "1", + "--pooled_context_representation_size_for_auxiliary_encoder", + "1", + "--sigma_aggregation", + "p90", + "--model_selection", + "unfiltered", + "--jepa_objective", + "w_jepa_joint", + "--latent_architecture", + "deterministic_no_z", + "--wjepa_joint_start_weight", + "0.0", + "--wjepa_joint_end_weight", + "0.00025", + "--wjepa_joint_start_epoch", + "1", + "--wjepa_joint_annealing_epochs", + "50", + "--wjepa_s_x_start_epoch", + "0", + "--wjepa_s_x_annealing_epochs", + "30", + "--wjepa_z_start_epoch", + "0", + "--wjepa_z_annealing_epochs", + "0", + "--wjepa_s_y_start_epoch", + "0", + "--wjepa_s_y_annealing_epochs", + "30", + "--sw_full_num_projections", + "2048", + "--sw_marginal_num_projections", + "1024", + "--sw_p", + "2", + "--sw_include_marginals", + "true", + "--sw_s_x_weight", + "2.5e-05", + "--sw_z_weight", + "0.0", + "--sw_s_y_weight", + "1e-05", + "--torch_seed", + "42", + "--np_seed", + "42", + "--mask_seed", + "0", + "--sw_seed", + "0", + "--deterministic_mode", + "warn", + "--random", + "false", + "--safe_exit", + "false" + ] +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260916_021347/downstream_results/autoint_sim_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260916_021347/downstream_results/autoint_sim_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..c05db1a25a9c24e0a61b8df9c3a7813d671f9932 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260916_021347/downstream_results/autoint_sim_downstream_results.json @@ -0,0 +1,210 @@ +{ + "dataset": "sim", + "model": "autoint", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmph23mlm5q", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "sim", + "model_name": "autoint", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260916_021347/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260916_021347/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmph23mlm5q/split_data/split_config.json", + "sigma_aggregation": "p90", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "activation": "relu", + "initialization": "kaiming", + "n_layers": 3, + "d_token": 128, + "n_heads": 2, + "attention_dropout": 0.1, + "residual_dropout": 0.1, + "prenormalization": false + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "sim_val_loss": 0.36400794982910156, + "lr": 0.0005626666243188083, + "sim_val_accuracy": 0.8583783507347107, + "sim_val_precision": 0.8583271503448486, + "sim_val_recall": 0.8583783507347107, + "sim_val_f1-score": 0.8582327365875244, + "sim_val_support": 925.0, + "sim_val_score": 0.8583783507347107 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.3825133144855499, + "lr": 0.0005626666243188083, + "sim_test_accuracy": 0.8572884798049927, + "sim_test_precision": 0.8580183386802673, + "sim_test_recall": 0.8572884798049927, + "sim_test_f1-score": 0.8571587800979614, + "sim_test_support": 1962.0, + "sim_test_score": 0.8572884798049927 + } + } + }, + { + "seed": 1, + "val_metrics": { + "sim_val_loss": 0.3858925998210907, + "lr": 0.000531395198777318, + "sim_val_accuracy": 0.8594594597816467, + "sim_val_precision": 0.8662502765655518, + "sim_val_recall": 0.8594594597816467, + "sim_val_f1-score": 0.8596591949462891, + "sim_val_support": 925.0, + "sim_val_score": 0.8594594597816467 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.3820371925830841, + "lr": 0.0005313953151926398, + "sim_test_accuracy": 0.8455657362937927, + "sim_test_precision": 0.8508884310722351, + "sim_test_recall": 0.8455657362937927, + "sim_test_f1-score": 0.8461118340492249, + "sim_test_support": 1962.0, + "sim_test_score": 0.8455657362937927 + } + } + }, + { + "seed": 2, + "val_metrics": { + "sim_val_loss": 0.3799585998058319, + "lr": 0.0005936906673014164, + "sim_val_accuracy": 0.8572973012924194, + "sim_val_precision": 0.8597105145454407, + "sim_val_recall": 0.8572973012924194, + "sim_val_f1-score": 0.8573763966560364, + "sim_val_support": 925.0, + "sim_val_score": 0.8572973012924194 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.4110434651374817, + "lr": 0.0005936907255090773, + "sim_test_accuracy": 0.8506625890731812, + "sim_test_precision": 0.8547523617744446, + "sim_test_recall": 0.8506625890731812, + "sim_test_f1-score": 0.8503130674362183, + "sim_test_support": 1962.0, + "sim_test_score": 0.8506625890731812 + } + } + }, + { + "seed": 3, + "val_metrics": { + "sim_val_loss": 0.3794053792953491, + "lr": 0.0003756550431717187, + "sim_val_accuracy": 0.8681080937385559, + "sim_val_precision": 0.8708376288414001, + "sim_val_recall": 0.8681080937385559, + "sim_val_f1-score": 0.8683998584747314, + "sim_val_support": 925.0, + "sim_val_score": 0.8681080937385559 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.3876788318157196, + "lr": 0.0003756550431717187, + "sim_test_accuracy": 0.8537206649780273, + "sim_test_precision": 0.8558109402656555, + "sim_test_recall": 0.8537206649780273, + "sim_test_f1-score": 0.8537439107894897, + "sim_test_support": 1962.0, + "sim_test_score": 0.8537206649780273 + } + } + }, + { + "seed": 4, + "val_metrics": { + "sim_val_loss": 0.37984699010849, + "lr": 0.000531395198777318, + "sim_val_accuracy": 0.8562161922454834, + "sim_val_precision": 0.8612796664237976, + "sim_val_recall": 0.8562161922454834, + "sim_val_f1-score": 0.8567438125610352, + "sim_val_support": 925.0, + "sim_val_score": 0.8562161922454834 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.3842598497867584, + "lr": 0.0005313953151926398, + "sim_test_accuracy": 0.847604513168335, + "sim_test_precision": 0.8534238338470459, + "sim_test_recall": 0.847604513168335, + "sim_test_f1-score": 0.8479011058807373, + "sim_test_support": 1962.0, + "sim_test_score": 0.847604513168335 + } + } + } + ], + "summary": { + "0": { + "sim_test_loss": { + "mean": 0.38950653076171876, + "std": 0.012241361992321545, + "n": 5, + "ci95": 0.015197221744877747 + }, + "lr": { + "mean": 0.0005189606046769768, + "std": 8.417323672353716e-05, + "n": 5, + "ci95": 0.00010449812237184595 + }, + "sim_test_accuracy": { + "mean": 0.8509683966636657, + "std": 0.004693516840075176, + "n": 5, + "ci95": 0.005826836607452786 + }, + "sim_test_precision": { + "mean": 0.8545787811279297, + "std": 0.0026613666269679277, + "n": 5, + "ci95": 0.0033039933628152243 + }, + "sim_test_recall": { + "mean": 0.8509683966636657, + "std": 0.004693516840075176, + "n": 5, + "ci95": 0.005826836607452786 + }, + "sim_test_f1-score": { + "mean": 0.8510457396507263, + "std": 0.0044558485614241685, + "n": 5, + "ci95": 0.0055317797718941 + }, + "sim_test_support": { + "mean": 1962.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "sim_test_score": { + "mean": 0.8509683966636657, + "std": 0.004693516840075176, + "n": 5, + "ci95": 0.005826836607452786 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260916_021347/downstream_results/dcnv2_sim_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260916_021347/downstream_results/dcnv2_sim_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..aac8307ef272e3e0d76aee0c8748aa442cc56c41 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260916_021347/downstream_results/dcnv2_sim_downstream_results.json @@ -0,0 +1,209 @@ +{ + "dataset": "sim", + "model": "dcnv2", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpektm136l", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "sim", + "model_name": "dcnv2", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260916_021347/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260916_021347/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpektm136l/split_data/split_config.json", + "sigma_aggregation": "p90", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "hidden_dim": 128, + "n_hidden_layers": 4, + "n_cross_layers": 2, + "hidden_dropout": 0.1, + "cross_dropout": 0.1, + "d_embedding": 128, + "stacked": true + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "sim_val_loss": 0.40635114908218384, + "lr": 0.0003454915131442249, + "sim_val_accuracy": 0.8475675582885742, + "sim_val_precision": 0.847938060760498, + "sim_val_recall": 0.8475675582885742, + "sim_val_f1-score": 0.8476515412330627, + "sim_val_support": 925.0, + "sim_val_score": 0.8475675582885742 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.4280955195426941, + "lr": 0.0003454914840403944, + "sim_test_accuracy": 0.8527013063430786, + "sim_test_precision": 0.852744460105896, + "sim_test_recall": 0.8527013063430786, + "sim_test_f1-score": 0.8526800870895386, + "sim_test_support": 1962.0, + "sim_test_score": 0.8527013063430786 + } + } + }, + { + "seed": 1, + "val_metrics": { + "sim_val_loss": 0.4002242684364319, + "lr": 0.000437333423178643, + "sim_val_accuracy": 0.8583783507347107, + "sim_val_precision": 0.8592825531959534, + "sim_val_recall": 0.8583783507347107, + "sim_val_f1-score": 0.858418881893158, + "sim_val_support": 925.0, + "sim_val_score": 0.8583783507347107 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.4435223937034607, + "lr": 0.0004373333358671516, + "sim_test_accuracy": 0.8527013063430786, + "sim_test_precision": 0.8542618155479431, + "sim_test_recall": 0.8527013063430786, + "sim_test_f1-score": 0.8526677489280701, + "sim_test_support": 1962.0, + "sim_test_score": 0.8527013063430786 + } + } + }, + { + "seed": 2, + "val_metrics": { + "sim_val_loss": 0.44371917843818665, + "lr": 0.000406309321988374, + "sim_val_accuracy": 0.8291891813278198, + "sim_val_precision": 0.8344312310218811, + "sim_val_recall": 0.8291891813278198, + "sim_val_f1-score": 0.828852117061615, + "sim_val_support": 925.0, + "sim_val_score": 0.8291891813278198 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.47529271245002747, + "lr": 0.0004063093801960349, + "sim_test_accuracy": 0.8363914489746094, + "sim_test_precision": 0.844687819480896, + "sim_test_recall": 0.8363914489746094, + "sim_test_f1-score": 0.8358990550041199, + "sim_test_support": 1962.0, + "sim_test_score": 0.8363914489746094 + } + } + }, + { + "seed": 3, + "val_metrics": { + "sim_val_loss": 0.38766875863075256, + "lr": 0.0005000000237487257, + "sim_val_accuracy": 0.8572973012924194, + "sim_val_precision": 0.8587242960929871, + "sim_val_recall": 0.8572973012924194, + "sim_val_f1-score": 0.8573572635650635, + "sim_val_support": 925.0, + "sim_val_score": 0.8572973012924194 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.4333261549472809, + "lr": 0.0005000000819563866, + "sim_test_accuracy": 0.8491335511207581, + "sim_test_precision": 0.852593719959259, + "sim_test_recall": 0.8491335511207581, + "sim_test_f1-score": 0.8487865924835205, + "sim_test_support": 1962.0, + "sim_test_score": 0.8491335511207581 + } + } + }, + { + "seed": 4, + "val_metrics": { + "sim_val_loss": 0.40547969937324524, + "lr": 0.0005936906673014164, + "sim_val_accuracy": 0.8432432413101196, + "sim_val_precision": 0.8467190861701965, + "sim_val_recall": 0.8432432413101196, + "sim_val_f1-score": 0.8424417972564697, + "sim_val_support": 925.0, + "sim_val_score": 0.8432432413101196 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.43578553199768066, + "lr": 0.0005936907255090773, + "sim_test_accuracy": 0.8511722683906555, + "sim_test_precision": 0.8552222847938538, + "sim_test_recall": 0.8511722683906555, + "sim_test_f1-score": 0.850665271282196, + "sim_test_support": 1962.0, + "sim_test_score": 0.8511722683906555 + } + } + } + ], + "summary": { + "0": { + "sim_test_loss": { + "mean": 0.44320446252822876, + "std": 0.018779367083963768, + "n": 5, + "ci95": 0.023313925850936397 + }, + "lr": { + "mean": 0.00045656500151380897, + "std": 9.476666227605696e-05, + "n": 5, + "ci95": 0.0001176494888015357 + }, + "sim_test_accuracy": { + "mean": 0.848419976234436, + "std": 0.00688166666350363, + "n": 5, + "ci95": 0.008543347899130617 + }, + "sim_test_precision": { + "mean": 0.8519020199775695, + "std": 0.004178086781796735, + "n": 5, + "ci95": 0.005186948260506908 + }, + "sim_test_recall": { + "mean": 0.848419976234436, + "std": 0.00688166666350363, + "n": 5, + "ci95": 0.008543347899130617 + }, + "sim_test_f1-score": { + "mean": 0.8481397509574891, + "std": 0.007031157659068591, + "n": 5, + "ci95": 0.00872893572913583 + }, + "sim_test_support": { + "mean": 1962.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "sim_test_score": { + "mean": 0.848419976234436, + "std": 0.00688166666350363, + "n": 5, + "ci95": 0.008543347899130617 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260916_021347/downstream_results/fttransformer_sim_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260916_021347/downstream_results/fttransformer_sim_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..e6bc6eb5fe3fce0e3faf6badcbbee879f6605745 --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260916_021347/downstream_results/fttransformer_sim_downstream_results.json @@ -0,0 +1,209 @@ +{ + "dataset": "sim", + "model": "fttransformer", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmp0rzxon_s", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "sim", + "model_name": "fttransformer", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260916_021347/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260916_021347/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmp0rzxon_s/split_data/split_config.json", + "sigma_aggregation": "p90", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_blocks": 3, + "d_block": 128, + "attention_n_heads": 4, + "attention_dropout": 0.1, + "ffn_d_hidden_multiplier": 2.0, + "ffn_dropout": 0.1, + "residual_dropout": 0.0 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "sim_val_loss": 0.39820966124534607, + "lr": 9.549150126986206e-05, + "sim_val_accuracy": 0.8508108258247375, + "sim_val_precision": 0.8518505096435547, + "sim_val_recall": 0.8508108258247375, + "sim_val_f1-score": 0.8510094285011292, + "sim_val_support": 925.0, + "sim_val_score": 0.8508108258247375 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.42575085163116455, + "lr": 9.549151582177728e-05, + "sim_test_accuracy": 0.8521916270256042, + "sim_test_precision": 0.8531011939048767, + "sim_test_recall": 0.8521916270256042, + "sim_test_f1-score": 0.8522056937217712, + "sim_test_support": 1962.0, + "sim_test_score": 0.8521916270256042 + } + } + }, + { + "seed": 1, + "val_metrics": { + "sim_val_loss": 0.35576528310775757, + "lr": 0.00023208661878015846, + "sim_val_accuracy": 0.8540540337562561, + "sim_val_precision": 0.855444610118866, + "sim_val_recall": 0.8540540337562561, + "sim_val_f1-score": 0.8541958928108215, + "sim_val_support": 925.0, + "sim_val_score": 0.8540540337562561 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.37184810638427734, + "lr": 0.00023208657512441278, + "sim_test_accuracy": 0.856269121170044, + "sim_test_precision": 0.8584330677986145, + "sim_test_recall": 0.856269121170044, + "sim_test_f1-score": 0.8564442992210388, + "sim_test_support": 1962.0, + "sim_test_score": 0.856269121170044 + } + } + }, + { + "seed": 2, + "val_metrics": { + "sim_val_loss": 0.37365588545799255, + "lr": 0.00018128800729755312, + "sim_val_accuracy": 0.8497297167778015, + "sim_val_precision": 0.8522219657897949, + "sim_val_recall": 0.8497297167778015, + "sim_val_f1-score": 0.8497582077980042, + "sim_val_support": 925.0, + "sim_val_score": 0.8497297167778015 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.402897447347641, + "lr": 0.00018128800729755312, + "sim_test_accuracy": 0.8491335511207581, + "sim_test_precision": 0.8536770939826965, + "sim_test_recall": 0.8491335511207581, + "sim_test_f1-score": 0.8490366339683533, + "sim_test_support": 1962.0, + "sim_test_score": 0.8491335511207581 + } + } + }, + { + "seed": 3, + "val_metrics": { + "sim_val_loss": 0.37134307622909546, + "lr": 2.4471739379805513e-05, + "sim_val_accuracy": 0.8551351428031921, + "sim_val_precision": 0.8565653562545776, + "sim_val_recall": 0.8551351428031921, + "sim_val_f1-score": 0.8554383516311646, + "sim_val_support": 925.0, + "sim_val_score": 0.8551351428031921 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.3938498795032501, + "lr": 2.447174301778432e-05, + "sim_test_accuracy": 0.856269121170044, + "sim_test_precision": 0.8573340177536011, + "sim_test_recall": 0.856269121170044, + "sim_test_f1-score": 0.8563140630722046, + "sim_test_support": 1962.0, + "sim_test_score": 0.856269121170044 + } + } + }, + { + "seed": 4, + "val_metrics": { + "sim_val_loss": 0.37412169575691223, + "lr": 0.0003454915131442249, + "sim_val_accuracy": 0.8529729843139648, + "sim_val_precision": 0.85638827085495, + "sim_val_recall": 0.8529729843139648, + "sim_val_f1-score": 0.853297770023346, + "sim_val_support": 925.0, + "sim_val_score": 0.8529729843139648 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.3788248598575592, + "lr": 0.0003454914840403944, + "sim_test_accuracy": 0.8572884798049927, + "sim_test_precision": 0.8595444560050964, + "sim_test_recall": 0.8572884798049927, + "sim_test_f1-score": 0.8575030565261841, + "sim_test_support": 1962.0, + "sim_test_score": 0.8572884798049927 + } + } + } + ], + "summary": { + "0": { + "sim_test_loss": { + "mean": 0.39463422894477845, + "std": 0.02125019836439681, + "n": 5, + "ci95": 0.026381376260986716 + }, + "lr": { + "mean": 0.0001757658650603844, + "std": 0.000123837781679779, + "n": 5, + "ci95": 0.00015374026434001773 + }, + "sim_test_accuracy": { + "mean": 0.8542303800582886, + "std": 0.0034568440754915004, + "n": 5, + "ci95": 0.00429155072659919 + }, + "sim_test_precision": { + "mean": 0.856417965888977, + "std": 0.0028804549281427998, + "n": 5, + "ci95": 0.0035759838077307126 + }, + "sim_test_recall": { + "mean": 0.8542303800582886, + "std": 0.0034568440754915004, + "n": 5, + "ci95": 0.00429155072659919 + }, + "sim_test_f1-score": { + "mean": 0.8543007493019104, + "std": 0.00357082547462595, + "n": 5, + "ci95": 0.004433054637563034 + }, + "sim_test_support": { + "mean": 1962.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "sim_test_score": { + "mean": 0.8542303800582886, + "std": 0.0034568440754915004, + "n": 5, + "ci95": 0.00429155072659919 + } + } + } +} \ No newline at end of file diff --git a/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260916_021347/downstream_results/mlp_sim_downstream_results.json b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260916_021347/downstream_results/mlp_sim_downstream_results.json new file mode 100644 index 0000000000000000000000000000000000000000..e1dd43e33414e11249efbc53e631b34b34cc622d --- /dev/null +++ b/JEPA/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260916_021347/downstream_results/mlp_sim_downstream_results.json @@ -0,0 +1,205 @@ +{ + "dataset": "sim", + "model": "mlp", + "config": { + "script_path": "src/run_benchmark.py", + "folder_name": "/tmp/tmpcat_8w0c", + "data_path": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/baselines/Var-T-JEPA/datasets", + "data_set": "sim", + "model_name": "mlp", + "using_embedding": true, + "emb_train": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260916_021347/embedding.npy", + "emb_test": "/pfss/mlde/workspaces/mlde_wsp_MGPATH/truong/JEPA/reproduction/artifacts/djepa_six_dataset_sw_tune_v1/sim/deterministic_sw/sw_f0.00025_sx2.5e-05_sy1e-05/seed_42/wjepa_deterministic/sim/20260916_021347/embedding.npy", + "use_split_assignments": true, + "exp_train_total_epochs": 50, + "split_config_file": "/tmp/tmpcat_8w0c/split_data/split_config.json", + "sigma_aggregation": "p90", + "lr": 0.001, + "weight_decay": 0.01, + "batch_size": 128, + "exp_patience": 16, + "encoder_type": "linear_flatten", + "n_hidden": 2, + "hidden_dim": 128, + "dropout": 0.1 + }, + "runs": [ + { + "seed": 0, + "val_metrics": { + "sim_val_loss": 0.45564520359039307, + "lr": 0.0010000000474974513, + "sim_val_accuracy": 0.8183783888816833, + "sim_val_precision": 0.8303206562995911, + "sim_val_recall": 0.8183783888816833, + "sim_val_f1-score": 0.815730094909668, + "sim_val_support": 925.0, + "sim_val_score": 0.8183783888816833 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.48241764307022095, + "lr": 0.0010000001639127731, + "sim_test_accuracy": 0.8165137767791748, + "sim_test_precision": 0.8337898254394531, + "sim_test_recall": 0.8165137767791748, + "sim_test_f1-score": 0.8118561506271362, + "sim_test_support": 1962.0, + "sim_test_score": 0.8165137767791748 + } + } + }, + { + "seed": 1, + "val_metrics": { + "sim_val_loss": 0.3617926836013794, + "lr": 0.0010000000474974513, + "sim_val_accuracy": 0.8583783507347107, + "sim_val_precision": 0.8585049510002136, + "sim_val_recall": 0.8583783507347107, + "sim_val_f1-score": 0.8581913113594055, + "sim_val_support": 925.0, + "sim_val_score": 0.8583783507347107 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.36500853300094604, + "lr": 0.0010000001639127731, + "sim_test_accuracy": 0.8593271970748901, + "sim_test_precision": 0.8609239459037781, + "sim_test_recall": 0.8593271970748901, + "sim_test_f1-score": 0.859132707118988, + "sim_test_support": 1962.0, + "sim_test_score": 0.8593271970748901 + } + } + }, + { + "seed": 2, + "val_metrics": { + "sim_val_loss": 0.4245508909225464, + "lr": 0.0010000000474974513, + "sim_val_accuracy": 0.8421621918678284, + "sim_val_precision": 0.8508545160293579, + "sim_val_recall": 0.8421621918678284, + "sim_val_f1-score": 0.8413665890693665, + "sim_val_support": 925.0, + "sim_val_score": 0.8421621918678284 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.4545876383781433, + "lr": 0.0010000001639127731, + "sim_test_accuracy": 0.8297655582427979, + "sim_test_precision": 0.8420844674110413, + "sim_test_recall": 0.8297655582427979, + "sim_test_f1-score": 0.8291539549827576, + "sim_test_support": 1962.0, + "sim_test_score": 0.8297655582427979 + } + } + }, + { + "seed": 3, + "val_metrics": { + "sim_val_loss": 0.36281323432922363, + "lr": 0.0010000000474974513, + "sim_val_accuracy": 0.8670270442962646, + "sim_val_precision": 0.8716503977775574, + "sim_val_recall": 0.8670270442962646, + "sim_val_f1-score": 0.8670254349708557, + "sim_val_support": 925.0, + "sim_val_score": 0.8670270442962646 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.37134167551994324, + "lr": 0.0010000001639127731, + "sim_test_accuracy": 0.8521916270256042, + "sim_test_precision": 0.8558180332183838, + "sim_test_recall": 0.8521916270256042, + "sim_test_f1-score": 0.8521935343742371, + "sim_test_support": 1962.0, + "sim_test_score": 0.8521916270256042 + } + } + }, + { + "seed": 4, + "val_metrics": { + "sim_val_loss": 0.39330244064331055, + "lr": 0.0010000000474974513, + "sim_val_accuracy": 0.8475675582885742, + "sim_val_precision": 0.857559859752655, + "sim_val_recall": 0.8475675582885742, + "sim_val_f1-score": 0.8476885557174683, + "sim_val_support": 925.0, + "sim_val_score": 0.8475675582885742 + }, + "test_metrics": { + "0": { + "sim_test_loss": 0.4053797423839569, + "lr": 0.0010000001639127731, + "sim_test_accuracy": 0.8399592041969299, + "sim_test_precision": 0.8495704531669617, + "sim_test_recall": 0.8399592041969299, + "sim_test_f1-score": 0.8391807675361633, + "sim_test_support": 1962.0, + "sim_test_score": 0.8399592041969299 + } + } + } + ], + "summary": { + "0": { + "sim_test_loss": { + "mean": 0.4157470464706421, + "std": 0.05149581231294036, + "n": 5, + "ci95": 0.06393024560038711 + }, + "lr": { + "mean": 0.0010000001639127731, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "sim_test_accuracy": { + "mean": 0.8395514726638794, + "std": 0.017157495271778365, + "n": 5, + "ci95": 0.02130042885713711 + }, + "sim_test_precision": { + "mean": 0.8484373450279236, + "std": 0.01080026131353026, + "n": 5, + "ci95": 0.013408145775551591 + }, + "sim_test_recall": { + "mean": 0.8395514726638794, + "std": 0.017157495271778365, + "n": 5, + "ci95": 0.02130042885713711 + }, + "sim_test_f1-score": { + "mean": 0.8383034229278564, + "std": 0.018779876099806664, + "n": 5, + "ci95": 0.02331455777625982 + }, + "sim_test_support": { + "mean": 1962.0, + "std": 0.0, + "n": 5, + "ci95": 0.0 + }, + "sim_test_score": { + "mean": 0.8395514726638794, + "std": 0.017157495271778365, + "n": 5, + "ci95": 0.02130042885713711 + } + } + } +} \ No newline at end of file