Upload batch 10/14: sst2/mBERT/mean/seed_42 ... sst2/mBERT/multi_branch_average/seed_42
Browse files- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_42/all_results.json +15 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_42/eval_results.json +9 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_42/run_command.txt +1 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_42/train_results.json +9 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_42/trainer_state.json +419 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_43/all_results.json +15 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_43/eval_results.json +9 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_43/run_command.txt +1 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_43/train_results.json +9 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_43/trainer_state.json +419 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_44/all_results.json +15 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_44/eval_results.json +9 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_44/run_command.txt +1 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_44/train_results.json +9 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_44/trainer_state.json +419 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_42/all_results.json +15 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_42/eval_results.json +9 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_42/run_command.txt +1 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_42/train_results.json +9 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_42/trainer_state.json +419 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_43/all_results.json +15 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_43/eval_results.json +9 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_43/run_command.txt +1 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_43/train_results.json +9 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_43/trainer_state.json +419 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_44/all_results.json +15 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_44/eval_results.json +9 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_44/run_command.txt +1 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_44/train_results.json +9 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_44/trainer_state.json +419 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/multi_branch_average/seed_42/all_results.json +15 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/multi_branch_average/seed_42/eval_results.json +9 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/multi_branch_average/seed_42/run_command.txt +1 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/multi_branch_average/seed_42/train_results.json +9 -0
- mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/multi_branch_average/seed_42/trainer_state.json +419 -0
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_42/all_results.json
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.8727064220183486,
|
| 4 |
+
"eval_loss": 0.4059946835041046,
|
| 5 |
+
"eval_runtime": 2.4122,
|
| 6 |
+
"eval_samples": 872,
|
| 7 |
+
"eval_samples_per_second": 361.502,
|
| 8 |
+
"eval_steps_per_second": 2.487,
|
| 9 |
+
"total_flos": 6631676508966912.0,
|
| 10 |
+
"train_loss": 0.2541694050258054,
|
| 11 |
+
"train_runtime": 251.1926,
|
| 12 |
+
"train_samples": 67349,
|
| 13 |
+
"train_samples_per_second": 804.351,
|
| 14 |
+
"train_steps_per_second": 10.056
|
| 15 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_42/eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.8727064220183486,
|
| 4 |
+
"eval_loss": 0.4059946835041046,
|
| 5 |
+
"eval_runtime": 2.4122,
|
| 6 |
+
"eval_samples": 872,
|
| 7 |
+
"eval_samples_per_second": 361.502,
|
| 8 |
+
"eval_steps_per_second": 2.487
|
| 9 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_42/run_command.txt
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
/usr/local/bin/python /workspace/run_glue_MHA_gated.py --model_name_or_path bert-base-multilingual-cased --dataset_name tmnam20/ViGLUE --task_name sst2 --max_seq_length 256 --do_train --do_eval --evaluation_strategy epoch --save_strategy no --logging_strategy steps --logging_steps 50 --report_to none --num_train_epochs 3 --learning_rate 2e-05 --per_device_train_batch_size 80 --per_device_eval_batch_size 168 --gradient_accumulation_steps 1 --dataloader_num_workers 2 --seed 42 --output_dir result_ablation_mbert_paper/sst2/mBERT/mean/seed_42 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save --use_mha true --pooling_strategy mean --n_heads 8 --head_dim 32 --classifier_hidden_dim 512 --head_dropout 0.05
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_42/train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 6631676508966912.0,
|
| 4 |
+
"train_loss": 0.2541694050258054,
|
| 5 |
+
"train_runtime": 251.1926,
|
| 6 |
+
"train_samples": 67349,
|
| 7 |
+
"train_samples_per_second": 804.351,
|
| 8 |
+
"train_steps_per_second": 10.056
|
| 9 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_42/trainer_state.json
ADDED
|
@@ -0,0 +1,419 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 3.0,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 2526,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.05938242280285035,
|
| 13 |
+
"grad_norm": 7.129025459289551,
|
| 14 |
+
"learning_rate": 1.9604117181314332e-05,
|
| 15 |
+
"loss": 0.5463,
|
| 16 |
+
"step": 50
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 0.1187648456057007,
|
| 20 |
+
"grad_norm": 3.9759106636047363,
|
| 21 |
+
"learning_rate": 1.9208234362628665e-05,
|
| 22 |
+
"loss": 0.458,
|
| 23 |
+
"step": 100
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.17814726840855108,
|
| 27 |
+
"grad_norm": 6.411795616149902,
|
| 28 |
+
"learning_rate": 1.8812351543942996e-05,
|
| 29 |
+
"loss": 0.4158,
|
| 30 |
+
"step": 150
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.2375296912114014,
|
| 34 |
+
"grad_norm": 4.937953472137451,
|
| 35 |
+
"learning_rate": 1.8416468725257326e-05,
|
| 36 |
+
"loss": 0.4147,
|
| 37 |
+
"step": 200
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.29691211401425177,
|
| 41 |
+
"grad_norm": 7.532645225524902,
|
| 42 |
+
"learning_rate": 1.8020585906571656e-05,
|
| 43 |
+
"loss": 0.3916,
|
| 44 |
+
"step": 250
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.35629453681710216,
|
| 48 |
+
"grad_norm": 4.6888957023620605,
|
| 49 |
+
"learning_rate": 1.762470308788599e-05,
|
| 50 |
+
"loss": 0.3775,
|
| 51 |
+
"step": 300
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.4156769596199525,
|
| 55 |
+
"grad_norm": 4.964507102966309,
|
| 56 |
+
"learning_rate": 1.7228820269200316e-05,
|
| 57 |
+
"loss": 0.3422,
|
| 58 |
+
"step": 350
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.4750593824228028,
|
| 62 |
+
"grad_norm": 5.227758407592773,
|
| 63 |
+
"learning_rate": 1.6832937450514647e-05,
|
| 64 |
+
"loss": 0.3468,
|
| 65 |
+
"step": 400
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.5344418052256532,
|
| 69 |
+
"grad_norm": 5.890567302703857,
|
| 70 |
+
"learning_rate": 1.643705463182898e-05,
|
| 71 |
+
"loss": 0.3413,
|
| 72 |
+
"step": 450
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 0.5938242280285035,
|
| 76 |
+
"grad_norm": 4.4760589599609375,
|
| 77 |
+
"learning_rate": 1.604117181314331e-05,
|
| 78 |
+
"loss": 0.3299,
|
| 79 |
+
"step": 500
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.6532066508313539,
|
| 83 |
+
"grad_norm": 4.769516468048096,
|
| 84 |
+
"learning_rate": 1.564528899445764e-05,
|
| 85 |
+
"loss": 0.3316,
|
| 86 |
+
"step": 550
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 0.7125890736342043,
|
| 90 |
+
"grad_norm": 4.499186992645264,
|
| 91 |
+
"learning_rate": 1.5249406175771972e-05,
|
| 92 |
+
"loss": 0.323,
|
| 93 |
+
"step": 600
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.7719714964370546,
|
| 97 |
+
"grad_norm": 3.3831887245178223,
|
| 98 |
+
"learning_rate": 1.4853523357086304e-05,
|
| 99 |
+
"loss": 0.3104,
|
| 100 |
+
"step": 650
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.831353919239905,
|
| 104 |
+
"grad_norm": 4.017539978027344,
|
| 105 |
+
"learning_rate": 1.4457640538400634e-05,
|
| 106 |
+
"loss": 0.3133,
|
| 107 |
+
"step": 700
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.8907363420427553,
|
| 111 |
+
"grad_norm": 6.041660785675049,
|
| 112 |
+
"learning_rate": 1.4061757719714966e-05,
|
| 113 |
+
"loss": 0.3049,
|
| 114 |
+
"step": 750
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.9501187648456056,
|
| 118 |
+
"grad_norm": 3.813662528991699,
|
| 119 |
+
"learning_rate": 1.3665874901029297e-05,
|
| 120 |
+
"loss": 0.2837,
|
| 121 |
+
"step": 800
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 1.0,
|
| 125 |
+
"eval_accuracy": 0.8646788990825688,
|
| 126 |
+
"eval_loss": 0.33560505509376526,
|
| 127 |
+
"eval_runtime": 2.4354,
|
| 128 |
+
"eval_samples_per_second": 358.048,
|
| 129 |
+
"eval_steps_per_second": 2.464,
|
| 130 |
+
"step": 842
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 1.009501187648456,
|
| 134 |
+
"grad_norm": 3.738520383834839,
|
| 135 |
+
"learning_rate": 1.3269992082343628e-05,
|
| 136 |
+
"loss": 0.2972,
|
| 137 |
+
"step": 850
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.0688836104513064,
|
| 141 |
+
"grad_norm": 4.79104471206665,
|
| 142 |
+
"learning_rate": 1.2874109263657959e-05,
|
| 143 |
+
"loss": 0.2565,
|
| 144 |
+
"step": 900
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 1.1282660332541568,
|
| 148 |
+
"grad_norm": 3.6198294162750244,
|
| 149 |
+
"learning_rate": 1.247822644497229e-05,
|
| 150 |
+
"loss": 0.2418,
|
| 151 |
+
"step": 950
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"epoch": 1.187648456057007,
|
| 155 |
+
"grad_norm": 6.629349708557129,
|
| 156 |
+
"learning_rate": 1.208234362628662e-05,
|
| 157 |
+
"loss": 0.24,
|
| 158 |
+
"step": 1000
|
| 159 |
+
},
|
| 160 |
+
{
|
| 161 |
+
"epoch": 1.2470308788598574,
|
| 162 |
+
"grad_norm": 4.985299587249756,
|
| 163 |
+
"learning_rate": 1.168646080760095e-05,
|
| 164 |
+
"loss": 0.2566,
|
| 165 |
+
"step": 1050
|
| 166 |
+
},
|
| 167 |
+
{
|
| 168 |
+
"epoch": 1.3064133016627077,
|
| 169 |
+
"grad_norm": 4.4612908363342285,
|
| 170 |
+
"learning_rate": 1.1290577988915281e-05,
|
| 171 |
+
"loss": 0.2339,
|
| 172 |
+
"step": 1100
|
| 173 |
+
},
|
| 174 |
+
{
|
| 175 |
+
"epoch": 1.365795724465558,
|
| 176 |
+
"grad_norm": 5.318294525146484,
|
| 177 |
+
"learning_rate": 1.0894695170229613e-05,
|
| 178 |
+
"loss": 0.2247,
|
| 179 |
+
"step": 1150
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"epoch": 1.4251781472684084,
|
| 183 |
+
"grad_norm": 5.443617343902588,
|
| 184 |
+
"learning_rate": 1.0498812351543943e-05,
|
| 185 |
+
"loss": 0.2218,
|
| 186 |
+
"step": 1200
|
| 187 |
+
},
|
| 188 |
+
{
|
| 189 |
+
"epoch": 1.484560570071259,
|
| 190 |
+
"grad_norm": 7.907466888427734,
|
| 191 |
+
"learning_rate": 1.0102929532858275e-05,
|
| 192 |
+
"loss": 0.2394,
|
| 193 |
+
"step": 1250
|
| 194 |
+
},
|
| 195 |
+
{
|
| 196 |
+
"epoch": 1.5439429928741093,
|
| 197 |
+
"grad_norm": 6.404354572296143,
|
| 198 |
+
"learning_rate": 9.707046714172605e-06,
|
| 199 |
+
"loss": 0.2342,
|
| 200 |
+
"step": 1300
|
| 201 |
+
},
|
| 202 |
+
{
|
| 203 |
+
"epoch": 1.6033254156769596,
|
| 204 |
+
"grad_norm": 8.37304973602295,
|
| 205 |
+
"learning_rate": 9.311163895486937e-06,
|
| 206 |
+
"loss": 0.2015,
|
| 207 |
+
"step": 1350
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 1.66270783847981,
|
| 211 |
+
"grad_norm": 4.459824085235596,
|
| 212 |
+
"learning_rate": 8.915281076801267e-06,
|
| 213 |
+
"loss": 0.2468,
|
| 214 |
+
"step": 1400
|
| 215 |
+
},
|
| 216 |
+
{
|
| 217 |
+
"epoch": 1.7220902612826603,
|
| 218 |
+
"grad_norm": 3.3669028282165527,
|
| 219 |
+
"learning_rate": 8.5193982581156e-06,
|
| 220 |
+
"loss": 0.2098,
|
| 221 |
+
"step": 1450
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 1.7814726840855108,
|
| 225 |
+
"grad_norm": 5.465628147125244,
|
| 226 |
+
"learning_rate": 8.12351543942993e-06,
|
| 227 |
+
"loss": 0.2162,
|
| 228 |
+
"step": 1500
|
| 229 |
+
},
|
| 230 |
+
{
|
| 231 |
+
"epoch": 1.8408551068883612,
|
| 232 |
+
"grad_norm": 5.963727951049805,
|
| 233 |
+
"learning_rate": 7.72763262074426e-06,
|
| 234 |
+
"loss": 0.2278,
|
| 235 |
+
"step": 1550
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 1.9002375296912115,
|
| 239 |
+
"grad_norm": 3.128110647201538,
|
| 240 |
+
"learning_rate": 7.331749802058591e-06,
|
| 241 |
+
"loss": 0.2221,
|
| 242 |
+
"step": 1600
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 1.9596199524940618,
|
| 246 |
+
"grad_norm": 7.751652240753174,
|
| 247 |
+
"learning_rate": 6.935866983372922e-06,
|
| 248 |
+
"loss": 0.2141,
|
| 249 |
+
"step": 1650
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 2.0,
|
| 253 |
+
"eval_accuracy": 0.8738532110091743,
|
| 254 |
+
"eval_loss": 0.3798689544200897,
|
| 255 |
+
"eval_runtime": 2.4151,
|
| 256 |
+
"eval_samples_per_second": 361.059,
|
| 257 |
+
"eval_steps_per_second": 2.484,
|
| 258 |
+
"step": 1684
|
| 259 |
+
},
|
| 260 |
+
{
|
| 261 |
+
"epoch": 2.019002375296912,
|
| 262 |
+
"grad_norm": 7.893195629119873,
|
| 263 |
+
"learning_rate": 6.539984164687253e-06,
|
| 264 |
+
"loss": 0.1988,
|
| 265 |
+
"step": 1700
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 2.0783847980997625,
|
| 269 |
+
"grad_norm": 6.198106288909912,
|
| 270 |
+
"learning_rate": 6.1441013460015846e-06,
|
| 271 |
+
"loss": 0.172,
|
| 272 |
+
"step": 1750
|
| 273 |
+
},
|
| 274 |
+
{
|
| 275 |
+
"epoch": 2.137767220902613,
|
| 276 |
+
"grad_norm": 4.629214763641357,
|
| 277 |
+
"learning_rate": 5.748218527315916e-06,
|
| 278 |
+
"loss": 0.1712,
|
| 279 |
+
"step": 1800
|
| 280 |
+
},
|
| 281 |
+
{
|
| 282 |
+
"epoch": 2.197149643705463,
|
| 283 |
+
"grad_norm": 5.395705699920654,
|
| 284 |
+
"learning_rate": 5.352335708630247e-06,
|
| 285 |
+
"loss": 0.1787,
|
| 286 |
+
"step": 1850
|
| 287 |
+
},
|
| 288 |
+
{
|
| 289 |
+
"epoch": 2.2565320665083135,
|
| 290 |
+
"grad_norm": 7.353418827056885,
|
| 291 |
+
"learning_rate": 4.956452889944577e-06,
|
| 292 |
+
"loss": 0.1811,
|
| 293 |
+
"step": 1900
|
| 294 |
+
},
|
| 295 |
+
{
|
| 296 |
+
"epoch": 2.315914489311164,
|
| 297 |
+
"grad_norm": 5.178650856018066,
|
| 298 |
+
"learning_rate": 4.560570071258908e-06,
|
| 299 |
+
"loss": 0.1749,
|
| 300 |
+
"step": 1950
|
| 301 |
+
},
|
| 302 |
+
{
|
| 303 |
+
"epoch": 2.375296912114014,
|
| 304 |
+
"grad_norm": 4.736900329589844,
|
| 305 |
+
"learning_rate": 4.164687252573238e-06,
|
| 306 |
+
"loss": 0.1817,
|
| 307 |
+
"step": 2000
|
| 308 |
+
},
|
| 309 |
+
{
|
| 310 |
+
"epoch": 2.4346793349168645,
|
| 311 |
+
"grad_norm": 4.11307954788208,
|
| 312 |
+
"learning_rate": 3.7688044338875695e-06,
|
| 313 |
+
"loss": 0.1762,
|
| 314 |
+
"step": 2050
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"epoch": 2.494061757719715,
|
| 318 |
+
"grad_norm": 5.891166687011719,
|
| 319 |
+
"learning_rate": 3.3729216152019006e-06,
|
| 320 |
+
"loss": 0.1514,
|
| 321 |
+
"step": 2100
|
| 322 |
+
},
|
| 323 |
+
{
|
| 324 |
+
"epoch": 2.553444180522565,
|
| 325 |
+
"grad_norm": 4.036669731140137,
|
| 326 |
+
"learning_rate": 2.977038796516231e-06,
|
| 327 |
+
"loss": 0.1732,
|
| 328 |
+
"step": 2150
|
| 329 |
+
},
|
| 330 |
+
{
|
| 331 |
+
"epoch": 2.6128266033254155,
|
| 332 |
+
"grad_norm": 5.7535295486450195,
|
| 333 |
+
"learning_rate": 2.581155977830562e-06,
|
| 334 |
+
"loss": 0.173,
|
| 335 |
+
"step": 2200
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 2.6722090261282663,
|
| 339 |
+
"grad_norm": 5.590721607208252,
|
| 340 |
+
"learning_rate": 2.1852731591448932e-06,
|
| 341 |
+
"loss": 0.1766,
|
| 342 |
+
"step": 2250
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 2.731591448931116,
|
| 346 |
+
"grad_norm": 7.806019306182861,
|
| 347 |
+
"learning_rate": 1.789390340459224e-06,
|
| 348 |
+
"loss": 0.1598,
|
| 349 |
+
"step": 2300
|
| 350 |
+
},
|
| 351 |
+
{
|
| 352 |
+
"epoch": 2.790973871733967,
|
| 353 |
+
"grad_norm": 6.060710430145264,
|
| 354 |
+
"learning_rate": 1.3935075217735553e-06,
|
| 355 |
+
"loss": 0.1769,
|
| 356 |
+
"step": 2350
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 2.850356294536817,
|
| 360 |
+
"grad_norm": 5.39944314956665,
|
| 361 |
+
"learning_rate": 9.976247030878861e-07,
|
| 362 |
+
"loss": 0.1681,
|
| 363 |
+
"step": 2400
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"epoch": 2.9097387173396676,
|
| 367 |
+
"grad_norm": 10.039387702941895,
|
| 368 |
+
"learning_rate": 6.01741884402217e-07,
|
| 369 |
+
"loss": 0.1575,
|
| 370 |
+
"step": 2450
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"epoch": 2.969121140142518,
|
| 374 |
+
"grad_norm": 4.6573591232299805,
|
| 375 |
+
"learning_rate": 2.0585906571654793e-07,
|
| 376 |
+
"loss": 0.1672,
|
| 377 |
+
"step": 2500
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 3.0,
|
| 381 |
+
"eval_accuracy": 0.8727064220183486,
|
| 382 |
+
"eval_loss": 0.4059946835041046,
|
| 383 |
+
"eval_runtime": 2.3491,
|
| 384 |
+
"eval_samples_per_second": 371.204,
|
| 385 |
+
"eval_steps_per_second": 2.554,
|
| 386 |
+
"step": 2526
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 3.0,
|
| 390 |
+
"step": 2526,
|
| 391 |
+
"total_flos": 6631676508966912.0,
|
| 392 |
+
"train_loss": 0.2541694050258054,
|
| 393 |
+
"train_runtime": 251.1926,
|
| 394 |
+
"train_samples_per_second": 804.351,
|
| 395 |
+
"train_steps_per_second": 10.056
|
| 396 |
+
}
|
| 397 |
+
],
|
| 398 |
+
"logging_steps": 50,
|
| 399 |
+
"max_steps": 2526,
|
| 400 |
+
"num_input_tokens_seen": 0,
|
| 401 |
+
"num_train_epochs": 3,
|
| 402 |
+
"save_steps": 500,
|
| 403 |
+
"stateful_callbacks": {
|
| 404 |
+
"TrainerControl": {
|
| 405 |
+
"args": {
|
| 406 |
+
"should_epoch_stop": false,
|
| 407 |
+
"should_evaluate": false,
|
| 408 |
+
"should_log": false,
|
| 409 |
+
"should_save": false,
|
| 410 |
+
"should_training_stop": false
|
| 411 |
+
},
|
| 412 |
+
"attributes": {}
|
| 413 |
+
}
|
| 414 |
+
},
|
| 415 |
+
"total_flos": 6631676508966912.0,
|
| 416 |
+
"train_batch_size": 80,
|
| 417 |
+
"trial_name": null,
|
| 418 |
+
"trial_params": null
|
| 419 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_43/all_results.json
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.8899082568807339,
|
| 4 |
+
"eval_loss": 0.36042511463165283,
|
| 5 |
+
"eval_runtime": 2.4201,
|
| 6 |
+
"eval_samples": 872,
|
| 7 |
+
"eval_samples_per_second": 360.313,
|
| 8 |
+
"eval_steps_per_second": 2.479,
|
| 9 |
+
"total_flos": 6637229322706944.0,
|
| 10 |
+
"train_loss": 0.25073873817401565,
|
| 11 |
+
"train_runtime": 265.9468,
|
| 12 |
+
"train_samples": 67349,
|
| 13 |
+
"train_samples_per_second": 759.727,
|
| 14 |
+
"train_steps_per_second": 9.498
|
| 15 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_43/eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.8899082568807339,
|
| 4 |
+
"eval_loss": 0.36042511463165283,
|
| 5 |
+
"eval_runtime": 2.4201,
|
| 6 |
+
"eval_samples": 872,
|
| 7 |
+
"eval_samples_per_second": 360.313,
|
| 8 |
+
"eval_steps_per_second": 2.479
|
| 9 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_43/run_command.txt
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
/usr/local/bin/python /workspace/run_glue_MHA_gated.py --model_name_or_path bert-base-multilingual-cased --dataset_name tmnam20/ViGLUE --task_name sst2 --max_seq_length 256 --do_train --do_eval --evaluation_strategy epoch --save_strategy no --logging_strategy steps --logging_steps 50 --report_to none --num_train_epochs 3 --learning_rate 2e-05 --per_device_train_batch_size 80 --per_device_eval_batch_size 168 --gradient_accumulation_steps 1 --dataloader_num_workers 2 --seed 43 --output_dir result_ablation_mbert_paper/sst2/mBERT/mean/seed_43 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save --use_mha true --pooling_strategy mean --n_heads 8 --head_dim 32 --classifier_hidden_dim 512 --head_dropout 0.05
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_43/train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 6637229322706944.0,
|
| 4 |
+
"train_loss": 0.25073873817401565,
|
| 5 |
+
"train_runtime": 265.9468,
|
| 6 |
+
"train_samples": 67349,
|
| 7 |
+
"train_samples_per_second": 759.727,
|
| 8 |
+
"train_steps_per_second": 9.498
|
| 9 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_43/trainer_state.json
ADDED
|
@@ -0,0 +1,419 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 3.0,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 2526,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.05938242280285035,
|
| 13 |
+
"grad_norm": 6.11844539642334,
|
| 14 |
+
"learning_rate": 1.9604117181314332e-05,
|
| 15 |
+
"loss": 0.5674,
|
| 16 |
+
"step": 50
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 0.1187648456057007,
|
| 20 |
+
"grad_norm": 5.086767196655273,
|
| 21 |
+
"learning_rate": 1.9208234362628665e-05,
|
| 22 |
+
"loss": 0.4584,
|
| 23 |
+
"step": 100
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.17814726840855108,
|
| 27 |
+
"grad_norm": 8.495840072631836,
|
| 28 |
+
"learning_rate": 1.8812351543942996e-05,
|
| 29 |
+
"loss": 0.4207,
|
| 30 |
+
"step": 150
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.2375296912114014,
|
| 34 |
+
"grad_norm": 3.9398913383483887,
|
| 35 |
+
"learning_rate": 1.8416468725257326e-05,
|
| 36 |
+
"loss": 0.414,
|
| 37 |
+
"step": 200
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.29691211401425177,
|
| 41 |
+
"grad_norm": 4.006356239318848,
|
| 42 |
+
"learning_rate": 1.8020585906571656e-05,
|
| 43 |
+
"loss": 0.397,
|
| 44 |
+
"step": 250
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.35629453681710216,
|
| 48 |
+
"grad_norm": 3.638170003890991,
|
| 49 |
+
"learning_rate": 1.762470308788599e-05,
|
| 50 |
+
"loss": 0.3694,
|
| 51 |
+
"step": 300
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.4156769596199525,
|
| 55 |
+
"grad_norm": 4.466193675994873,
|
| 56 |
+
"learning_rate": 1.7228820269200316e-05,
|
| 57 |
+
"loss": 0.3461,
|
| 58 |
+
"step": 350
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.4750593824228028,
|
| 62 |
+
"grad_norm": 3.9991912841796875,
|
| 63 |
+
"learning_rate": 1.6832937450514647e-05,
|
| 64 |
+
"loss": 0.35,
|
| 65 |
+
"step": 400
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.5344418052256532,
|
| 69 |
+
"grad_norm": 5.2119550704956055,
|
| 70 |
+
"learning_rate": 1.643705463182898e-05,
|
| 71 |
+
"loss": 0.3348,
|
| 72 |
+
"step": 450
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 0.5938242280285035,
|
| 76 |
+
"grad_norm": 4.55248498916626,
|
| 77 |
+
"learning_rate": 1.604117181314331e-05,
|
| 78 |
+
"loss": 0.3212,
|
| 79 |
+
"step": 500
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.6532066508313539,
|
| 83 |
+
"grad_norm": 6.220978260040283,
|
| 84 |
+
"learning_rate": 1.564528899445764e-05,
|
| 85 |
+
"loss": 0.312,
|
| 86 |
+
"step": 550
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 0.7125890736342043,
|
| 90 |
+
"grad_norm": 4.904836654663086,
|
| 91 |
+
"learning_rate": 1.5249406175771972e-05,
|
| 92 |
+
"loss": 0.3084,
|
| 93 |
+
"step": 600
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.7719714964370546,
|
| 97 |
+
"grad_norm": 5.378131866455078,
|
| 98 |
+
"learning_rate": 1.4853523357086304e-05,
|
| 99 |
+
"loss": 0.3044,
|
| 100 |
+
"step": 650
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.831353919239905,
|
| 104 |
+
"grad_norm": 4.440019130706787,
|
| 105 |
+
"learning_rate": 1.4457640538400634e-05,
|
| 106 |
+
"loss": 0.3056,
|
| 107 |
+
"step": 700
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.8907363420427553,
|
| 111 |
+
"grad_norm": 5.2848711013793945,
|
| 112 |
+
"learning_rate": 1.4061757719714966e-05,
|
| 113 |
+
"loss": 0.3075,
|
| 114 |
+
"step": 750
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.9501187648456056,
|
| 118 |
+
"grad_norm": 5.052803993225098,
|
| 119 |
+
"learning_rate": 1.3665874901029297e-05,
|
| 120 |
+
"loss": 0.2941,
|
| 121 |
+
"step": 800
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 1.0,
|
| 125 |
+
"eval_accuracy": 0.8715596330275229,
|
| 126 |
+
"eval_loss": 0.3323845863342285,
|
| 127 |
+
"eval_runtime": 2.3798,
|
| 128 |
+
"eval_samples_per_second": 366.422,
|
| 129 |
+
"eval_steps_per_second": 2.521,
|
| 130 |
+
"step": 842
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 1.009501187648456,
|
| 134 |
+
"grad_norm": 3.9674901962280273,
|
| 135 |
+
"learning_rate": 1.3269992082343628e-05,
|
| 136 |
+
"loss": 0.276,
|
| 137 |
+
"step": 850
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.0688836104513064,
|
| 141 |
+
"grad_norm": 4.825985908508301,
|
| 142 |
+
"learning_rate": 1.2874109263657959e-05,
|
| 143 |
+
"loss": 0.243,
|
| 144 |
+
"step": 900
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 1.1282660332541568,
|
| 148 |
+
"grad_norm": 6.010342121124268,
|
| 149 |
+
"learning_rate": 1.247822644497229e-05,
|
| 150 |
+
"loss": 0.2302,
|
| 151 |
+
"step": 950
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"epoch": 1.187648456057007,
|
| 155 |
+
"grad_norm": 5.639437198638916,
|
| 156 |
+
"learning_rate": 1.208234362628662e-05,
|
| 157 |
+
"loss": 0.2416,
|
| 158 |
+
"step": 1000
|
| 159 |
+
},
|
| 160 |
+
{
|
| 161 |
+
"epoch": 1.2470308788598574,
|
| 162 |
+
"grad_norm": 5.301666259765625,
|
| 163 |
+
"learning_rate": 1.168646080760095e-05,
|
| 164 |
+
"loss": 0.2385,
|
| 165 |
+
"step": 1050
|
| 166 |
+
},
|
| 167 |
+
{
|
| 168 |
+
"epoch": 1.3064133016627077,
|
| 169 |
+
"grad_norm": 6.264228343963623,
|
| 170 |
+
"learning_rate": 1.1290577988915281e-05,
|
| 171 |
+
"loss": 0.2415,
|
| 172 |
+
"step": 1100
|
| 173 |
+
},
|
| 174 |
+
{
|
| 175 |
+
"epoch": 1.365795724465558,
|
| 176 |
+
"grad_norm": 6.046392917633057,
|
| 177 |
+
"learning_rate": 1.0894695170229613e-05,
|
| 178 |
+
"loss": 0.2348,
|
| 179 |
+
"step": 1150
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"epoch": 1.4251781472684084,
|
| 183 |
+
"grad_norm": 6.563099384307861,
|
| 184 |
+
"learning_rate": 1.0498812351543943e-05,
|
| 185 |
+
"loss": 0.2413,
|
| 186 |
+
"step": 1200
|
| 187 |
+
},
|
| 188 |
+
{
|
| 189 |
+
"epoch": 1.484560570071259,
|
| 190 |
+
"grad_norm": 6.6240997314453125,
|
| 191 |
+
"learning_rate": 1.0102929532858275e-05,
|
| 192 |
+
"loss": 0.2182,
|
| 193 |
+
"step": 1250
|
| 194 |
+
},
|
| 195 |
+
{
|
| 196 |
+
"epoch": 1.5439429928741093,
|
| 197 |
+
"grad_norm": 4.606036186218262,
|
| 198 |
+
"learning_rate": 9.707046714172605e-06,
|
| 199 |
+
"loss": 0.2357,
|
| 200 |
+
"step": 1300
|
| 201 |
+
},
|
| 202 |
+
{
|
| 203 |
+
"epoch": 1.6033254156769596,
|
| 204 |
+
"grad_norm": 4.885876178741455,
|
| 205 |
+
"learning_rate": 9.311163895486937e-06,
|
| 206 |
+
"loss": 0.2312,
|
| 207 |
+
"step": 1350
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 1.66270783847981,
|
| 211 |
+
"grad_norm": 6.166421890258789,
|
| 212 |
+
"learning_rate": 8.915281076801267e-06,
|
| 213 |
+
"loss": 0.2427,
|
| 214 |
+
"step": 1400
|
| 215 |
+
},
|
| 216 |
+
{
|
| 217 |
+
"epoch": 1.7220902612826603,
|
| 218 |
+
"grad_norm": 4.643629550933838,
|
| 219 |
+
"learning_rate": 8.5193982581156e-06,
|
| 220 |
+
"loss": 0.2101,
|
| 221 |
+
"step": 1450
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 1.7814726840855108,
|
| 225 |
+
"grad_norm": 5.984914302825928,
|
| 226 |
+
"learning_rate": 8.12351543942993e-06,
|
| 227 |
+
"loss": 0.2226,
|
| 228 |
+
"step": 1500
|
| 229 |
+
},
|
| 230 |
+
{
|
| 231 |
+
"epoch": 1.8408551068883612,
|
| 232 |
+
"grad_norm": 7.77462100982666,
|
| 233 |
+
"learning_rate": 7.72763262074426e-06,
|
| 234 |
+
"loss": 0.207,
|
| 235 |
+
"step": 1550
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 1.9002375296912115,
|
| 239 |
+
"grad_norm": 6.456450939178467,
|
| 240 |
+
"learning_rate": 7.331749802058591e-06,
|
| 241 |
+
"loss": 0.2078,
|
| 242 |
+
"step": 1600
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 1.9596199524940618,
|
| 246 |
+
"grad_norm": 5.926214694976807,
|
| 247 |
+
"learning_rate": 6.935866983372922e-06,
|
| 248 |
+
"loss": 0.2075,
|
| 249 |
+
"step": 1650
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 2.0,
|
| 253 |
+
"eval_accuracy": 0.8899082568807339,
|
| 254 |
+
"eval_loss": 0.3344023525714874,
|
| 255 |
+
"eval_runtime": 2.3625,
|
| 256 |
+
"eval_samples_per_second": 369.1,
|
| 257 |
+
"eval_steps_per_second": 2.54,
|
| 258 |
+
"step": 1684
|
| 259 |
+
},
|
| 260 |
+
{
|
| 261 |
+
"epoch": 2.019002375296912,
|
| 262 |
+
"grad_norm": 4.733188629150391,
|
| 263 |
+
"learning_rate": 6.539984164687253e-06,
|
| 264 |
+
"loss": 0.191,
|
| 265 |
+
"step": 1700
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 2.0783847980997625,
|
| 269 |
+
"grad_norm": 5.335888862609863,
|
| 270 |
+
"learning_rate": 6.1441013460015846e-06,
|
| 271 |
+
"loss": 0.1585,
|
| 272 |
+
"step": 1750
|
| 273 |
+
},
|
| 274 |
+
{
|
| 275 |
+
"epoch": 2.137767220902613,
|
| 276 |
+
"grad_norm": 4.269768714904785,
|
| 277 |
+
"learning_rate": 5.748218527315916e-06,
|
| 278 |
+
"loss": 0.1682,
|
| 279 |
+
"step": 1800
|
| 280 |
+
},
|
| 281 |
+
{
|
| 282 |
+
"epoch": 2.197149643705463,
|
| 283 |
+
"grad_norm": 9.054043769836426,
|
| 284 |
+
"learning_rate": 5.352335708630247e-06,
|
| 285 |
+
"loss": 0.1757,
|
| 286 |
+
"step": 1850
|
| 287 |
+
},
|
| 288 |
+
{
|
| 289 |
+
"epoch": 2.2565320665083135,
|
| 290 |
+
"grad_norm": 5.239339828491211,
|
| 291 |
+
"learning_rate": 4.956452889944577e-06,
|
| 292 |
+
"loss": 0.1608,
|
| 293 |
+
"step": 1900
|
| 294 |
+
},
|
| 295 |
+
{
|
| 296 |
+
"epoch": 2.315914489311164,
|
| 297 |
+
"grad_norm": 8.163712501525879,
|
| 298 |
+
"learning_rate": 4.560570071258908e-06,
|
| 299 |
+
"loss": 0.1657,
|
| 300 |
+
"step": 1950
|
| 301 |
+
},
|
| 302 |
+
{
|
| 303 |
+
"epoch": 2.375296912114014,
|
| 304 |
+
"grad_norm": 4.7352471351623535,
|
| 305 |
+
"learning_rate": 4.164687252573238e-06,
|
| 306 |
+
"loss": 0.1589,
|
| 307 |
+
"step": 2000
|
| 308 |
+
},
|
| 309 |
+
{
|
| 310 |
+
"epoch": 2.4346793349168645,
|
| 311 |
+
"grad_norm": 4.757707595825195,
|
| 312 |
+
"learning_rate": 3.7688044338875695e-06,
|
| 313 |
+
"loss": 0.1615,
|
| 314 |
+
"step": 2050
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"epoch": 2.494061757719715,
|
| 318 |
+
"grad_norm": 4.031062602996826,
|
| 319 |
+
"learning_rate": 3.3729216152019006e-06,
|
| 320 |
+
"loss": 0.1655,
|
| 321 |
+
"step": 2100
|
| 322 |
+
},
|
| 323 |
+
{
|
| 324 |
+
"epoch": 2.553444180522565,
|
| 325 |
+
"grad_norm": 5.558089256286621,
|
| 326 |
+
"learning_rate": 2.977038796516231e-06,
|
| 327 |
+
"loss": 0.1818,
|
| 328 |
+
"step": 2150
|
| 329 |
+
},
|
| 330 |
+
{
|
| 331 |
+
"epoch": 2.6128266033254155,
|
| 332 |
+
"grad_norm": 5.906740665435791,
|
| 333 |
+
"learning_rate": 2.581155977830562e-06,
|
| 334 |
+
"loss": 0.1681,
|
| 335 |
+
"step": 2200
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 2.6722090261282663,
|
| 339 |
+
"grad_norm": 7.199082851409912,
|
| 340 |
+
"learning_rate": 2.1852731591448932e-06,
|
| 341 |
+
"loss": 0.149,
|
| 342 |
+
"step": 2250
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 2.731591448931116,
|
| 346 |
+
"grad_norm": 5.645853519439697,
|
| 347 |
+
"learning_rate": 1.789390340459224e-06,
|
| 348 |
+
"loss": 0.1806,
|
| 349 |
+
"step": 2300
|
| 350 |
+
},
|
| 351 |
+
{
|
| 352 |
+
"epoch": 2.790973871733967,
|
| 353 |
+
"grad_norm": 7.038839340209961,
|
| 354 |
+
"learning_rate": 1.3935075217735553e-06,
|
| 355 |
+
"loss": 0.1677,
|
| 356 |
+
"step": 2350
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 2.850356294536817,
|
| 360 |
+
"grad_norm": 6.345663547515869,
|
| 361 |
+
"learning_rate": 9.976247030878861e-07,
|
| 362 |
+
"loss": 0.1643,
|
| 363 |
+
"step": 2400
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"epoch": 2.9097387173396676,
|
| 367 |
+
"grad_norm": 6.993391036987305,
|
| 368 |
+
"learning_rate": 6.01741884402217e-07,
|
| 369 |
+
"loss": 0.161,
|
| 370 |
+
"step": 2450
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"epoch": 2.969121140142518,
|
| 374 |
+
"grad_norm": 3.2426981925964355,
|
| 375 |
+
"learning_rate": 2.0585906571654793e-07,
|
| 376 |
+
"loss": 0.1612,
|
| 377 |
+
"step": 2500
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 3.0,
|
| 381 |
+
"eval_accuracy": 0.8899082568807339,
|
| 382 |
+
"eval_loss": 0.36042511463165283,
|
| 383 |
+
"eval_runtime": 2.5482,
|
| 384 |
+
"eval_samples_per_second": 342.207,
|
| 385 |
+
"eval_steps_per_second": 2.355,
|
| 386 |
+
"step": 2526
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 3.0,
|
| 390 |
+
"step": 2526,
|
| 391 |
+
"total_flos": 6637229322706944.0,
|
| 392 |
+
"train_loss": 0.25073873817401565,
|
| 393 |
+
"train_runtime": 265.9468,
|
| 394 |
+
"train_samples_per_second": 759.727,
|
| 395 |
+
"train_steps_per_second": 9.498
|
| 396 |
+
}
|
| 397 |
+
],
|
| 398 |
+
"logging_steps": 50,
|
| 399 |
+
"max_steps": 2526,
|
| 400 |
+
"num_input_tokens_seen": 0,
|
| 401 |
+
"num_train_epochs": 3,
|
| 402 |
+
"save_steps": 500,
|
| 403 |
+
"stateful_callbacks": {
|
| 404 |
+
"TrainerControl": {
|
| 405 |
+
"args": {
|
| 406 |
+
"should_epoch_stop": false,
|
| 407 |
+
"should_evaluate": false,
|
| 408 |
+
"should_log": false,
|
| 409 |
+
"should_save": false,
|
| 410 |
+
"should_training_stop": false
|
| 411 |
+
},
|
| 412 |
+
"attributes": {}
|
| 413 |
+
}
|
| 414 |
+
},
|
| 415 |
+
"total_flos": 6637229322706944.0,
|
| 416 |
+
"train_batch_size": 80,
|
| 417 |
+
"trial_name": null,
|
| 418 |
+
"trial_params": null
|
| 419 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_44/all_results.json
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.8795871559633027,
|
| 4 |
+
"eval_loss": 0.36055734753608704,
|
| 5 |
+
"eval_runtime": 2.1451,
|
| 6 |
+
"eval_samples": 872,
|
| 7 |
+
"eval_samples_per_second": 406.509,
|
| 8 |
+
"eval_steps_per_second": 2.797,
|
| 9 |
+
"total_flos": 6633975382115328.0,
|
| 10 |
+
"train_loss": 0.25276644231776635,
|
| 11 |
+
"train_runtime": 266.5658,
|
| 12 |
+
"train_samples": 67349,
|
| 13 |
+
"train_samples_per_second": 757.963,
|
| 14 |
+
"train_steps_per_second": 9.476
|
| 15 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_44/eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.8795871559633027,
|
| 4 |
+
"eval_loss": 0.36055734753608704,
|
| 5 |
+
"eval_runtime": 2.1451,
|
| 6 |
+
"eval_samples": 872,
|
| 7 |
+
"eval_samples_per_second": 406.509,
|
| 8 |
+
"eval_steps_per_second": 2.797
|
| 9 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_44/run_command.txt
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
/usr/local/bin/python /workspace/run_glue_MHA_gated.py --model_name_or_path bert-base-multilingual-cased --dataset_name tmnam20/ViGLUE --task_name sst2 --max_seq_length 256 --do_train --do_eval --evaluation_strategy epoch --save_strategy no --logging_strategy steps --logging_steps 50 --report_to none --num_train_epochs 3 --learning_rate 2e-05 --per_device_train_batch_size 80 --per_device_eval_batch_size 168 --gradient_accumulation_steps 1 --dataloader_num_workers 2 --seed 44 --output_dir result_ablation_mbert_paper/sst2/mBERT/mean/seed_44 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save --use_mha true --pooling_strategy mean --n_heads 8 --head_dim 32 --classifier_hidden_dim 512 --head_dropout 0.05
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_44/train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 6633975382115328.0,
|
| 4 |
+
"train_loss": 0.25276644231776635,
|
| 5 |
+
"train_runtime": 266.5658,
|
| 6 |
+
"train_samples": 67349,
|
| 7 |
+
"train_samples_per_second": 757.963,
|
| 8 |
+
"train_steps_per_second": 9.476
|
| 9 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mean/seed_44/trainer_state.json
ADDED
|
@@ -0,0 +1,419 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 3.0,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 2526,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.05938242280285035,
|
| 13 |
+
"grad_norm": 3.8256235122680664,
|
| 14 |
+
"learning_rate": 1.9604117181314332e-05,
|
| 15 |
+
"loss": 0.5631,
|
| 16 |
+
"step": 50
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 0.1187648456057007,
|
| 20 |
+
"grad_norm": 4.138633728027344,
|
| 21 |
+
"learning_rate": 1.9208234362628665e-05,
|
| 22 |
+
"loss": 0.4688,
|
| 23 |
+
"step": 100
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.17814726840855108,
|
| 27 |
+
"grad_norm": 3.968381881713867,
|
| 28 |
+
"learning_rate": 1.8812351543942996e-05,
|
| 29 |
+
"loss": 0.4234,
|
| 30 |
+
"step": 150
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.2375296912114014,
|
| 34 |
+
"grad_norm": 3.6280550956726074,
|
| 35 |
+
"learning_rate": 1.8416468725257326e-05,
|
| 36 |
+
"loss": 0.3936,
|
| 37 |
+
"step": 200
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.29691211401425177,
|
| 41 |
+
"grad_norm": 3.4781851768493652,
|
| 42 |
+
"learning_rate": 1.8020585906571656e-05,
|
| 43 |
+
"loss": 0.3819,
|
| 44 |
+
"step": 250
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.35629453681710216,
|
| 48 |
+
"grad_norm": 6.165865421295166,
|
| 49 |
+
"learning_rate": 1.762470308788599e-05,
|
| 50 |
+
"loss": 0.3784,
|
| 51 |
+
"step": 300
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.4156769596199525,
|
| 55 |
+
"grad_norm": 5.5023298263549805,
|
| 56 |
+
"learning_rate": 1.7228820269200316e-05,
|
| 57 |
+
"loss": 0.3694,
|
| 58 |
+
"step": 350
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.4750593824228028,
|
| 62 |
+
"grad_norm": 4.349158763885498,
|
| 63 |
+
"learning_rate": 1.6832937450514647e-05,
|
| 64 |
+
"loss": 0.3472,
|
| 65 |
+
"step": 400
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.5344418052256532,
|
| 69 |
+
"grad_norm": 5.11631441116333,
|
| 70 |
+
"learning_rate": 1.643705463182898e-05,
|
| 71 |
+
"loss": 0.3261,
|
| 72 |
+
"step": 450
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 0.5938242280285035,
|
| 76 |
+
"grad_norm": 5.431488513946533,
|
| 77 |
+
"learning_rate": 1.604117181314331e-05,
|
| 78 |
+
"loss": 0.3399,
|
| 79 |
+
"step": 500
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.6532066508313539,
|
| 83 |
+
"grad_norm": 4.465195655822754,
|
| 84 |
+
"learning_rate": 1.564528899445764e-05,
|
| 85 |
+
"loss": 0.3239,
|
| 86 |
+
"step": 550
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 0.7125890736342043,
|
| 90 |
+
"grad_norm": 6.211365699768066,
|
| 91 |
+
"learning_rate": 1.5249406175771972e-05,
|
| 92 |
+
"loss": 0.3159,
|
| 93 |
+
"step": 600
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.7719714964370546,
|
| 97 |
+
"grad_norm": 4.169003009796143,
|
| 98 |
+
"learning_rate": 1.4853523357086304e-05,
|
| 99 |
+
"loss": 0.3101,
|
| 100 |
+
"step": 650
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.831353919239905,
|
| 104 |
+
"grad_norm": 3.7007365226745605,
|
| 105 |
+
"learning_rate": 1.4457640538400634e-05,
|
| 106 |
+
"loss": 0.3052,
|
| 107 |
+
"step": 700
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.8907363420427553,
|
| 111 |
+
"grad_norm": 7.2990922927856445,
|
| 112 |
+
"learning_rate": 1.4061757719714966e-05,
|
| 113 |
+
"loss": 0.2977,
|
| 114 |
+
"step": 750
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.9501187648456056,
|
| 118 |
+
"grad_norm": 5.146700382232666,
|
| 119 |
+
"learning_rate": 1.3665874901029297e-05,
|
| 120 |
+
"loss": 0.3048,
|
| 121 |
+
"step": 800
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 1.0,
|
| 125 |
+
"eval_accuracy": 0.8727064220183486,
|
| 126 |
+
"eval_loss": 0.3231830894947052,
|
| 127 |
+
"eval_runtime": 2.4162,
|
| 128 |
+
"eval_samples_per_second": 360.901,
|
| 129 |
+
"eval_steps_per_second": 2.483,
|
| 130 |
+
"step": 842
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 1.009501187648456,
|
| 134 |
+
"grad_norm": 9.58683967590332,
|
| 135 |
+
"learning_rate": 1.3269992082343628e-05,
|
| 136 |
+
"loss": 0.2719,
|
| 137 |
+
"step": 850
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.0688836104513064,
|
| 141 |
+
"grad_norm": 7.075453281402588,
|
| 142 |
+
"learning_rate": 1.2874109263657959e-05,
|
| 143 |
+
"loss": 0.2427,
|
| 144 |
+
"step": 900
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 1.1282660332541568,
|
| 148 |
+
"grad_norm": 5.452020168304443,
|
| 149 |
+
"learning_rate": 1.247822644497229e-05,
|
| 150 |
+
"loss": 0.2262,
|
| 151 |
+
"step": 950
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"epoch": 1.187648456057007,
|
| 155 |
+
"grad_norm": 3.3358142375946045,
|
| 156 |
+
"learning_rate": 1.208234362628662e-05,
|
| 157 |
+
"loss": 0.2392,
|
| 158 |
+
"step": 1000
|
| 159 |
+
},
|
| 160 |
+
{
|
| 161 |
+
"epoch": 1.2470308788598574,
|
| 162 |
+
"grad_norm": 4.201412200927734,
|
| 163 |
+
"learning_rate": 1.168646080760095e-05,
|
| 164 |
+
"loss": 0.2376,
|
| 165 |
+
"step": 1050
|
| 166 |
+
},
|
| 167 |
+
{
|
| 168 |
+
"epoch": 1.3064133016627077,
|
| 169 |
+
"grad_norm": 4.33805513381958,
|
| 170 |
+
"learning_rate": 1.1290577988915281e-05,
|
| 171 |
+
"loss": 0.2308,
|
| 172 |
+
"step": 1100
|
| 173 |
+
},
|
| 174 |
+
{
|
| 175 |
+
"epoch": 1.365795724465558,
|
| 176 |
+
"grad_norm": 4.972110271453857,
|
| 177 |
+
"learning_rate": 1.0894695170229613e-05,
|
| 178 |
+
"loss": 0.2441,
|
| 179 |
+
"step": 1150
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"epoch": 1.4251781472684084,
|
| 183 |
+
"grad_norm": 4.145775318145752,
|
| 184 |
+
"learning_rate": 1.0498812351543943e-05,
|
| 185 |
+
"loss": 0.2385,
|
| 186 |
+
"step": 1200
|
| 187 |
+
},
|
| 188 |
+
{
|
| 189 |
+
"epoch": 1.484560570071259,
|
| 190 |
+
"grad_norm": 4.327720642089844,
|
| 191 |
+
"learning_rate": 1.0102929532858275e-05,
|
| 192 |
+
"loss": 0.2226,
|
| 193 |
+
"step": 1250
|
| 194 |
+
},
|
| 195 |
+
{
|
| 196 |
+
"epoch": 1.5439429928741093,
|
| 197 |
+
"grad_norm": 6.255491256713867,
|
| 198 |
+
"learning_rate": 9.707046714172605e-06,
|
| 199 |
+
"loss": 0.232,
|
| 200 |
+
"step": 1300
|
| 201 |
+
},
|
| 202 |
+
{
|
| 203 |
+
"epoch": 1.6033254156769596,
|
| 204 |
+
"grad_norm": 5.9114813804626465,
|
| 205 |
+
"learning_rate": 9.311163895486937e-06,
|
| 206 |
+
"loss": 0.2224,
|
| 207 |
+
"step": 1350
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 1.66270783847981,
|
| 211 |
+
"grad_norm": 5.189817428588867,
|
| 212 |
+
"learning_rate": 8.915281076801267e-06,
|
| 213 |
+
"loss": 0.2131,
|
| 214 |
+
"step": 1400
|
| 215 |
+
},
|
| 216 |
+
{
|
| 217 |
+
"epoch": 1.7220902612826603,
|
| 218 |
+
"grad_norm": 4.231602191925049,
|
| 219 |
+
"learning_rate": 8.5193982581156e-06,
|
| 220 |
+
"loss": 0.2271,
|
| 221 |
+
"step": 1450
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 1.7814726840855108,
|
| 225 |
+
"grad_norm": 4.0591912269592285,
|
| 226 |
+
"learning_rate": 8.12351543942993e-06,
|
| 227 |
+
"loss": 0.2253,
|
| 228 |
+
"step": 1500
|
| 229 |
+
},
|
| 230 |
+
{
|
| 231 |
+
"epoch": 1.8408551068883612,
|
| 232 |
+
"grad_norm": 4.680384635925293,
|
| 233 |
+
"learning_rate": 7.72763262074426e-06,
|
| 234 |
+
"loss": 0.2227,
|
| 235 |
+
"step": 1550
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 1.9002375296912115,
|
| 239 |
+
"grad_norm": 7.154633522033691,
|
| 240 |
+
"learning_rate": 7.331749802058591e-06,
|
| 241 |
+
"loss": 0.2172,
|
| 242 |
+
"step": 1600
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 1.9596199524940618,
|
| 246 |
+
"grad_norm": 4.035558700561523,
|
| 247 |
+
"learning_rate": 6.935866983372922e-06,
|
| 248 |
+
"loss": 0.2188,
|
| 249 |
+
"step": 1650
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 2.0,
|
| 253 |
+
"eval_accuracy": 0.8841743119266054,
|
| 254 |
+
"eval_loss": 0.3439459204673767,
|
| 255 |
+
"eval_runtime": 2.3714,
|
| 256 |
+
"eval_samples_per_second": 367.714,
|
| 257 |
+
"eval_steps_per_second": 2.53,
|
| 258 |
+
"step": 1684
|
| 259 |
+
},
|
| 260 |
+
{
|
| 261 |
+
"epoch": 2.019002375296912,
|
| 262 |
+
"grad_norm": 4.597777366638184,
|
| 263 |
+
"learning_rate": 6.539984164687253e-06,
|
| 264 |
+
"loss": 0.2,
|
| 265 |
+
"step": 1700
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 2.0783847980997625,
|
| 269 |
+
"grad_norm": 5.824084758758545,
|
| 270 |
+
"learning_rate": 6.1441013460015846e-06,
|
| 271 |
+
"loss": 0.1807,
|
| 272 |
+
"step": 1750
|
| 273 |
+
},
|
| 274 |
+
{
|
| 275 |
+
"epoch": 2.137767220902613,
|
| 276 |
+
"grad_norm": 6.236695289611816,
|
| 277 |
+
"learning_rate": 5.748218527315916e-06,
|
| 278 |
+
"loss": 0.1609,
|
| 279 |
+
"step": 1800
|
| 280 |
+
},
|
| 281 |
+
{
|
| 282 |
+
"epoch": 2.197149643705463,
|
| 283 |
+
"grad_norm": 6.860519886016846,
|
| 284 |
+
"learning_rate": 5.352335708630247e-06,
|
| 285 |
+
"loss": 0.1801,
|
| 286 |
+
"step": 1850
|
| 287 |
+
},
|
| 288 |
+
{
|
| 289 |
+
"epoch": 2.2565320665083135,
|
| 290 |
+
"grad_norm": 5.984694957733154,
|
| 291 |
+
"learning_rate": 4.956452889944577e-06,
|
| 292 |
+
"loss": 0.158,
|
| 293 |
+
"step": 1900
|
| 294 |
+
},
|
| 295 |
+
{
|
| 296 |
+
"epoch": 2.315914489311164,
|
| 297 |
+
"grad_norm": 7.1133270263671875,
|
| 298 |
+
"learning_rate": 4.560570071258908e-06,
|
| 299 |
+
"loss": 0.1649,
|
| 300 |
+
"step": 1950
|
| 301 |
+
},
|
| 302 |
+
{
|
| 303 |
+
"epoch": 2.375296912114014,
|
| 304 |
+
"grad_norm": 4.818303108215332,
|
| 305 |
+
"learning_rate": 4.164687252573238e-06,
|
| 306 |
+
"loss": 0.1602,
|
| 307 |
+
"step": 2000
|
| 308 |
+
},
|
| 309 |
+
{
|
| 310 |
+
"epoch": 2.4346793349168645,
|
| 311 |
+
"grad_norm": 4.123764514923096,
|
| 312 |
+
"learning_rate": 3.7688044338875695e-06,
|
| 313 |
+
"loss": 0.1683,
|
| 314 |
+
"step": 2050
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"epoch": 2.494061757719715,
|
| 318 |
+
"grad_norm": 4.992451190948486,
|
| 319 |
+
"learning_rate": 3.3729216152019006e-06,
|
| 320 |
+
"loss": 0.1782,
|
| 321 |
+
"step": 2100
|
| 322 |
+
},
|
| 323 |
+
{
|
| 324 |
+
"epoch": 2.553444180522565,
|
| 325 |
+
"grad_norm": 6.76478910446167,
|
| 326 |
+
"learning_rate": 2.977038796516231e-06,
|
| 327 |
+
"loss": 0.1629,
|
| 328 |
+
"step": 2150
|
| 329 |
+
},
|
| 330 |
+
{
|
| 331 |
+
"epoch": 2.6128266033254155,
|
| 332 |
+
"grad_norm": 5.525815963745117,
|
| 333 |
+
"learning_rate": 2.581155977830562e-06,
|
| 334 |
+
"loss": 0.1701,
|
| 335 |
+
"step": 2200
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 2.6722090261282663,
|
| 339 |
+
"grad_norm": 6.1062541007995605,
|
| 340 |
+
"learning_rate": 2.1852731591448932e-06,
|
| 341 |
+
"loss": 0.1643,
|
| 342 |
+
"step": 2250
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 2.731591448931116,
|
| 346 |
+
"grad_norm": 4.965120315551758,
|
| 347 |
+
"learning_rate": 1.789390340459224e-06,
|
| 348 |
+
"loss": 0.1834,
|
| 349 |
+
"step": 2300
|
| 350 |
+
},
|
| 351 |
+
{
|
| 352 |
+
"epoch": 2.790973871733967,
|
| 353 |
+
"grad_norm": 4.1455302238464355,
|
| 354 |
+
"learning_rate": 1.3935075217735553e-06,
|
| 355 |
+
"loss": 0.1675,
|
| 356 |
+
"step": 2350
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 2.850356294536817,
|
| 360 |
+
"grad_norm": 4.758089065551758,
|
| 361 |
+
"learning_rate": 9.976247030878861e-07,
|
| 362 |
+
"loss": 0.1632,
|
| 363 |
+
"step": 2400
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"epoch": 2.9097387173396676,
|
| 367 |
+
"grad_norm": 5.925333499908447,
|
| 368 |
+
"learning_rate": 6.01741884402217e-07,
|
| 369 |
+
"loss": 0.1716,
|
| 370 |
+
"step": 2450
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"epoch": 2.969121140142518,
|
| 374 |
+
"grad_norm": 6.4998860359191895,
|
| 375 |
+
"learning_rate": 2.0585906571654793e-07,
|
| 376 |
+
"loss": 0.1662,
|
| 377 |
+
"step": 2500
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 3.0,
|
| 381 |
+
"eval_accuracy": 0.8795871559633027,
|
| 382 |
+
"eval_loss": 0.36055734753608704,
|
| 383 |
+
"eval_runtime": 2.131,
|
| 384 |
+
"eval_samples_per_second": 409.201,
|
| 385 |
+
"eval_steps_per_second": 2.816,
|
| 386 |
+
"step": 2526
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 3.0,
|
| 390 |
+
"step": 2526,
|
| 391 |
+
"total_flos": 6633975382115328.0,
|
| 392 |
+
"train_loss": 0.25276644231776635,
|
| 393 |
+
"train_runtime": 266.5658,
|
| 394 |
+
"train_samples_per_second": 757.963,
|
| 395 |
+
"train_steps_per_second": 9.476
|
| 396 |
+
}
|
| 397 |
+
],
|
| 398 |
+
"logging_steps": 50,
|
| 399 |
+
"max_steps": 2526,
|
| 400 |
+
"num_input_tokens_seen": 0,
|
| 401 |
+
"num_train_epochs": 3,
|
| 402 |
+
"save_steps": 500,
|
| 403 |
+
"stateful_callbacks": {
|
| 404 |
+
"TrainerControl": {
|
| 405 |
+
"args": {
|
| 406 |
+
"should_epoch_stop": false,
|
| 407 |
+
"should_evaluate": false,
|
| 408 |
+
"should_log": false,
|
| 409 |
+
"should_save": false,
|
| 410 |
+
"should_training_stop": false
|
| 411 |
+
},
|
| 412 |
+
"attributes": {}
|
| 413 |
+
}
|
| 414 |
+
},
|
| 415 |
+
"total_flos": 6633975382115328.0,
|
| 416 |
+
"train_batch_size": 80,
|
| 417 |
+
"trial_name": null,
|
| 418 |
+
"trial_params": null
|
| 419 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_42/all_results.json
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.8738532110091743,
|
| 4 |
+
"eval_loss": 0.3983316421508789,
|
| 5 |
+
"eval_runtime": 2.4368,
|
| 6 |
+
"eval_samples": 872,
|
| 7 |
+
"eval_samples_per_second": 357.841,
|
| 8 |
+
"eval_steps_per_second": 2.462,
|
| 9 |
+
"total_flos": 6692349849708888.0,
|
| 10 |
+
"train_loss": 0.255206602108639,
|
| 11 |
+
"train_runtime": 261.9382,
|
| 12 |
+
"train_samples": 67349,
|
| 13 |
+
"train_samples_per_second": 771.354,
|
| 14 |
+
"train_steps_per_second": 9.643
|
| 15 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_42/eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.8738532110091743,
|
| 4 |
+
"eval_loss": 0.3983316421508789,
|
| 5 |
+
"eval_runtime": 2.4368,
|
| 6 |
+
"eval_samples": 872,
|
| 7 |
+
"eval_samples_per_second": 357.841,
|
| 8 |
+
"eval_steps_per_second": 2.462
|
| 9 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_42/run_command.txt
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
/usr/local/bin/python /workspace/run_glue_MHA_gated.py --model_name_or_path bert-base-multilingual-cased --dataset_name tmnam20/ViGLUE --task_name sst2 --max_seq_length 256 --do_train --do_eval --evaluation_strategy epoch --save_strategy no --logging_strategy steps --logging_steps 50 --report_to none --num_train_epochs 3 --learning_rate 2e-05 --per_device_train_batch_size 80 --per_device_eval_batch_size 168 --gradient_accumulation_steps 1 --dataloader_num_workers 2 --seed 42 --output_dir result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_42 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save --use_mha true --pooling_strategy mha_attention --n_heads 8 --head_dim 32 --classifier_hidden_dim 512 --head_dropout 0.05
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_42/train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 6692349849708888.0,
|
| 4 |
+
"train_loss": 0.255206602108639,
|
| 5 |
+
"train_runtime": 261.9382,
|
| 6 |
+
"train_samples": 67349,
|
| 7 |
+
"train_samples_per_second": 771.354,
|
| 8 |
+
"train_steps_per_second": 9.643
|
| 9 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_42/trainer_state.json
ADDED
|
@@ -0,0 +1,419 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 3.0,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 2526,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.05938242280285035,
|
| 13 |
+
"grad_norm": 7.136999130249023,
|
| 14 |
+
"learning_rate": 1.9604117181314332e-05,
|
| 15 |
+
"loss": 0.545,
|
| 16 |
+
"step": 50
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 0.1187648456057007,
|
| 20 |
+
"grad_norm": 4.816132545471191,
|
| 21 |
+
"learning_rate": 1.9208234362628665e-05,
|
| 22 |
+
"loss": 0.4584,
|
| 23 |
+
"step": 100
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.17814726840855108,
|
| 27 |
+
"grad_norm": 7.979624271392822,
|
| 28 |
+
"learning_rate": 1.8812351543942996e-05,
|
| 29 |
+
"loss": 0.425,
|
| 30 |
+
"step": 150
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.2375296912114014,
|
| 34 |
+
"grad_norm": 8.028812408447266,
|
| 35 |
+
"learning_rate": 1.8416468725257326e-05,
|
| 36 |
+
"loss": 0.4196,
|
| 37 |
+
"step": 200
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.29691211401425177,
|
| 41 |
+
"grad_norm": 6.038510799407959,
|
| 42 |
+
"learning_rate": 1.8020585906571656e-05,
|
| 43 |
+
"loss": 0.3864,
|
| 44 |
+
"step": 250
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.35629453681710216,
|
| 48 |
+
"grad_norm": 4.557120323181152,
|
| 49 |
+
"learning_rate": 1.762470308788599e-05,
|
| 50 |
+
"loss": 0.3764,
|
| 51 |
+
"step": 300
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.4156769596199525,
|
| 55 |
+
"grad_norm": 5.02539587020874,
|
| 56 |
+
"learning_rate": 1.7228820269200316e-05,
|
| 57 |
+
"loss": 0.3496,
|
| 58 |
+
"step": 350
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.4750593824228028,
|
| 62 |
+
"grad_norm": 4.409160137176514,
|
| 63 |
+
"learning_rate": 1.6832937450514647e-05,
|
| 64 |
+
"loss": 0.3429,
|
| 65 |
+
"step": 400
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.5344418052256532,
|
| 69 |
+
"grad_norm": 5.63561487197876,
|
| 70 |
+
"learning_rate": 1.643705463182898e-05,
|
| 71 |
+
"loss": 0.3538,
|
| 72 |
+
"step": 450
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 0.5938242280285035,
|
| 76 |
+
"grad_norm": 4.248507976531982,
|
| 77 |
+
"learning_rate": 1.604117181314331e-05,
|
| 78 |
+
"loss": 0.3238,
|
| 79 |
+
"step": 500
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.6532066508313539,
|
| 83 |
+
"grad_norm": 3.6183316707611084,
|
| 84 |
+
"learning_rate": 1.564528899445764e-05,
|
| 85 |
+
"loss": 0.3329,
|
| 86 |
+
"step": 550
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 0.7125890736342043,
|
| 90 |
+
"grad_norm": 5.196035861968994,
|
| 91 |
+
"learning_rate": 1.5249406175771972e-05,
|
| 92 |
+
"loss": 0.3322,
|
| 93 |
+
"step": 600
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.7719714964370546,
|
| 97 |
+
"grad_norm": 4.75089693069458,
|
| 98 |
+
"learning_rate": 1.4853523357086304e-05,
|
| 99 |
+
"loss": 0.313,
|
| 100 |
+
"step": 650
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.831353919239905,
|
| 104 |
+
"grad_norm": 5.5071187019348145,
|
| 105 |
+
"learning_rate": 1.4457640538400634e-05,
|
| 106 |
+
"loss": 0.303,
|
| 107 |
+
"step": 700
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.8907363420427553,
|
| 111 |
+
"grad_norm": 5.43211030960083,
|
| 112 |
+
"learning_rate": 1.4061757719714966e-05,
|
| 113 |
+
"loss": 0.2997,
|
| 114 |
+
"step": 750
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.9501187648456056,
|
| 118 |
+
"grad_norm": 4.37415885925293,
|
| 119 |
+
"learning_rate": 1.3665874901029297e-05,
|
| 120 |
+
"loss": 0.2946,
|
| 121 |
+
"step": 800
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 1.0,
|
| 125 |
+
"eval_accuracy": 0.8646788990825688,
|
| 126 |
+
"eval_loss": 0.36605337262153625,
|
| 127 |
+
"eval_runtime": 2.868,
|
| 128 |
+
"eval_samples_per_second": 304.041,
|
| 129 |
+
"eval_steps_per_second": 2.092,
|
| 130 |
+
"step": 842
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 1.009501187648456,
|
| 134 |
+
"grad_norm": 3.4254796504974365,
|
| 135 |
+
"learning_rate": 1.3269992082343628e-05,
|
| 136 |
+
"loss": 0.298,
|
| 137 |
+
"step": 850
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.0688836104513064,
|
| 141 |
+
"grad_norm": 4.131380558013916,
|
| 142 |
+
"learning_rate": 1.2874109263657959e-05,
|
| 143 |
+
"loss": 0.2575,
|
| 144 |
+
"step": 900
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 1.1282660332541568,
|
| 148 |
+
"grad_norm": 3.091102123260498,
|
| 149 |
+
"learning_rate": 1.247822644497229e-05,
|
| 150 |
+
"loss": 0.244,
|
| 151 |
+
"step": 950
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"epoch": 1.187648456057007,
|
| 155 |
+
"grad_norm": 6.612815856933594,
|
| 156 |
+
"learning_rate": 1.208234362628662e-05,
|
| 157 |
+
"loss": 0.2439,
|
| 158 |
+
"step": 1000
|
| 159 |
+
},
|
| 160 |
+
{
|
| 161 |
+
"epoch": 1.2470308788598574,
|
| 162 |
+
"grad_norm": 4.433372497558594,
|
| 163 |
+
"learning_rate": 1.168646080760095e-05,
|
| 164 |
+
"loss": 0.2546,
|
| 165 |
+
"step": 1050
|
| 166 |
+
},
|
| 167 |
+
{
|
| 168 |
+
"epoch": 1.3064133016627077,
|
| 169 |
+
"grad_norm": 4.32538366317749,
|
| 170 |
+
"learning_rate": 1.1290577988915281e-05,
|
| 171 |
+
"loss": 0.2334,
|
| 172 |
+
"step": 1100
|
| 173 |
+
},
|
| 174 |
+
{
|
| 175 |
+
"epoch": 1.365795724465558,
|
| 176 |
+
"grad_norm": 3.685180902481079,
|
| 177 |
+
"learning_rate": 1.0894695170229613e-05,
|
| 178 |
+
"loss": 0.2268,
|
| 179 |
+
"step": 1150
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"epoch": 1.4251781472684084,
|
| 183 |
+
"grad_norm": 6.372713088989258,
|
| 184 |
+
"learning_rate": 1.0498812351543943e-05,
|
| 185 |
+
"loss": 0.222,
|
| 186 |
+
"step": 1200
|
| 187 |
+
},
|
| 188 |
+
{
|
| 189 |
+
"epoch": 1.484560570071259,
|
| 190 |
+
"grad_norm": 6.241898059844971,
|
| 191 |
+
"learning_rate": 1.0102929532858275e-05,
|
| 192 |
+
"loss": 0.2409,
|
| 193 |
+
"step": 1250
|
| 194 |
+
},
|
| 195 |
+
{
|
| 196 |
+
"epoch": 1.5439429928741093,
|
| 197 |
+
"grad_norm": 6.684188365936279,
|
| 198 |
+
"learning_rate": 9.707046714172605e-06,
|
| 199 |
+
"loss": 0.2386,
|
| 200 |
+
"step": 1300
|
| 201 |
+
},
|
| 202 |
+
{
|
| 203 |
+
"epoch": 1.6033254156769596,
|
| 204 |
+
"grad_norm": 10.182158470153809,
|
| 205 |
+
"learning_rate": 9.311163895486937e-06,
|
| 206 |
+
"loss": 0.2087,
|
| 207 |
+
"step": 1350
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 1.66270783847981,
|
| 211 |
+
"grad_norm": 4.2762041091918945,
|
| 212 |
+
"learning_rate": 8.915281076801267e-06,
|
| 213 |
+
"loss": 0.247,
|
| 214 |
+
"step": 1400
|
| 215 |
+
},
|
| 216 |
+
{
|
| 217 |
+
"epoch": 1.7220902612826603,
|
| 218 |
+
"grad_norm": 4.479830265045166,
|
| 219 |
+
"learning_rate": 8.5193982581156e-06,
|
| 220 |
+
"loss": 0.221,
|
| 221 |
+
"step": 1450
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 1.7814726840855108,
|
| 225 |
+
"grad_norm": 4.608367919921875,
|
| 226 |
+
"learning_rate": 8.12351543942993e-06,
|
| 227 |
+
"loss": 0.2219,
|
| 228 |
+
"step": 1500
|
| 229 |
+
},
|
| 230 |
+
{
|
| 231 |
+
"epoch": 1.8408551068883612,
|
| 232 |
+
"grad_norm": 5.412030220031738,
|
| 233 |
+
"learning_rate": 7.72763262074426e-06,
|
| 234 |
+
"loss": 0.2292,
|
| 235 |
+
"step": 1550
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 1.9002375296912115,
|
| 239 |
+
"grad_norm": 3.252452850341797,
|
| 240 |
+
"learning_rate": 7.331749802058591e-06,
|
| 241 |
+
"loss": 0.2268,
|
| 242 |
+
"step": 1600
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 1.9596199524940618,
|
| 246 |
+
"grad_norm": 6.23402738571167,
|
| 247 |
+
"learning_rate": 6.935866983372922e-06,
|
| 248 |
+
"loss": 0.2111,
|
| 249 |
+
"step": 1650
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 2.0,
|
| 253 |
+
"eval_accuracy": 0.8738532110091743,
|
| 254 |
+
"eval_loss": 0.3735749125480652,
|
| 255 |
+
"eval_runtime": 2.4558,
|
| 256 |
+
"eval_samples_per_second": 355.077,
|
| 257 |
+
"eval_steps_per_second": 2.443,
|
| 258 |
+
"step": 1684
|
| 259 |
+
},
|
| 260 |
+
{
|
| 261 |
+
"epoch": 2.019002375296912,
|
| 262 |
+
"grad_norm": 4.806388854980469,
|
| 263 |
+
"learning_rate": 6.539984164687253e-06,
|
| 264 |
+
"loss": 0.1942,
|
| 265 |
+
"step": 1700
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 2.0783847980997625,
|
| 269 |
+
"grad_norm": 6.469438076019287,
|
| 270 |
+
"learning_rate": 6.1441013460015846e-06,
|
| 271 |
+
"loss": 0.178,
|
| 272 |
+
"step": 1750
|
| 273 |
+
},
|
| 274 |
+
{
|
| 275 |
+
"epoch": 2.137767220902613,
|
| 276 |
+
"grad_norm": 6.294427871704102,
|
| 277 |
+
"learning_rate": 5.748218527315916e-06,
|
| 278 |
+
"loss": 0.1704,
|
| 279 |
+
"step": 1800
|
| 280 |
+
},
|
| 281 |
+
{
|
| 282 |
+
"epoch": 2.197149643705463,
|
| 283 |
+
"grad_norm": 4.624752998352051,
|
| 284 |
+
"learning_rate": 5.352335708630247e-06,
|
| 285 |
+
"loss": 0.1835,
|
| 286 |
+
"step": 1850
|
| 287 |
+
},
|
| 288 |
+
{
|
| 289 |
+
"epoch": 2.2565320665083135,
|
| 290 |
+
"grad_norm": 8.190455436706543,
|
| 291 |
+
"learning_rate": 4.956452889944577e-06,
|
| 292 |
+
"loss": 0.178,
|
| 293 |
+
"step": 1900
|
| 294 |
+
},
|
| 295 |
+
{
|
| 296 |
+
"epoch": 2.315914489311164,
|
| 297 |
+
"grad_norm": 5.218827247619629,
|
| 298 |
+
"learning_rate": 4.560570071258908e-06,
|
| 299 |
+
"loss": 0.1662,
|
| 300 |
+
"step": 1950
|
| 301 |
+
},
|
| 302 |
+
{
|
| 303 |
+
"epoch": 2.375296912114014,
|
| 304 |
+
"grad_norm": 6.203914642333984,
|
| 305 |
+
"learning_rate": 4.164687252573238e-06,
|
| 306 |
+
"loss": 0.1812,
|
| 307 |
+
"step": 2000
|
| 308 |
+
},
|
| 309 |
+
{
|
| 310 |
+
"epoch": 2.4346793349168645,
|
| 311 |
+
"grad_norm": 5.3609418869018555,
|
| 312 |
+
"learning_rate": 3.7688044338875695e-06,
|
| 313 |
+
"loss": 0.1764,
|
| 314 |
+
"step": 2050
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"epoch": 2.494061757719715,
|
| 318 |
+
"grad_norm": 5.475010395050049,
|
| 319 |
+
"learning_rate": 3.3729216152019006e-06,
|
| 320 |
+
"loss": 0.151,
|
| 321 |
+
"step": 2100
|
| 322 |
+
},
|
| 323 |
+
{
|
| 324 |
+
"epoch": 2.553444180522565,
|
| 325 |
+
"grad_norm": 3.98654842376709,
|
| 326 |
+
"learning_rate": 2.977038796516231e-06,
|
| 327 |
+
"loss": 0.1845,
|
| 328 |
+
"step": 2150
|
| 329 |
+
},
|
| 330 |
+
{
|
| 331 |
+
"epoch": 2.6128266033254155,
|
| 332 |
+
"grad_norm": 4.515543460845947,
|
| 333 |
+
"learning_rate": 2.581155977830562e-06,
|
| 334 |
+
"loss": 0.178,
|
| 335 |
+
"step": 2200
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 2.6722090261282663,
|
| 339 |
+
"grad_norm": 6.1017351150512695,
|
| 340 |
+
"learning_rate": 2.1852731591448932e-06,
|
| 341 |
+
"loss": 0.173,
|
| 342 |
+
"step": 2250
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 2.731591448931116,
|
| 346 |
+
"grad_norm": 4.544620990753174,
|
| 347 |
+
"learning_rate": 1.789390340459224e-06,
|
| 348 |
+
"loss": 0.1606,
|
| 349 |
+
"step": 2300
|
| 350 |
+
},
|
| 351 |
+
{
|
| 352 |
+
"epoch": 2.790973871733967,
|
| 353 |
+
"grad_norm": 5.618455410003662,
|
| 354 |
+
"learning_rate": 1.3935075217735553e-06,
|
| 355 |
+
"loss": 0.1799,
|
| 356 |
+
"step": 2350
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 2.850356294536817,
|
| 360 |
+
"grad_norm": 7.705366134643555,
|
| 361 |
+
"learning_rate": 9.976247030878861e-07,
|
| 362 |
+
"loss": 0.1556,
|
| 363 |
+
"step": 2400
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"epoch": 2.9097387173396676,
|
| 367 |
+
"grad_norm": 5.122476577758789,
|
| 368 |
+
"learning_rate": 6.01741884402217e-07,
|
| 369 |
+
"loss": 0.16,
|
| 370 |
+
"step": 2450
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"epoch": 2.969121140142518,
|
| 374 |
+
"grad_norm": 3.9589831829071045,
|
| 375 |
+
"learning_rate": 2.0585906571654793e-07,
|
| 376 |
+
"loss": 0.1521,
|
| 377 |
+
"step": 2500
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 3.0,
|
| 381 |
+
"eval_accuracy": 0.8738532110091743,
|
| 382 |
+
"eval_loss": 0.3983316421508789,
|
| 383 |
+
"eval_runtime": 2.3939,
|
| 384 |
+
"eval_samples_per_second": 364.264,
|
| 385 |
+
"eval_steps_per_second": 2.506,
|
| 386 |
+
"step": 2526
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 3.0,
|
| 390 |
+
"step": 2526,
|
| 391 |
+
"total_flos": 6692349849708888.0,
|
| 392 |
+
"train_loss": 0.255206602108639,
|
| 393 |
+
"train_runtime": 261.9382,
|
| 394 |
+
"train_samples_per_second": 771.354,
|
| 395 |
+
"train_steps_per_second": 9.643
|
| 396 |
+
}
|
| 397 |
+
],
|
| 398 |
+
"logging_steps": 50,
|
| 399 |
+
"max_steps": 2526,
|
| 400 |
+
"num_input_tokens_seen": 0,
|
| 401 |
+
"num_train_epochs": 3,
|
| 402 |
+
"save_steps": 500,
|
| 403 |
+
"stateful_callbacks": {
|
| 404 |
+
"TrainerControl": {
|
| 405 |
+
"args": {
|
| 406 |
+
"should_epoch_stop": false,
|
| 407 |
+
"should_evaluate": false,
|
| 408 |
+
"should_log": false,
|
| 409 |
+
"should_save": false,
|
| 410 |
+
"should_training_stop": false
|
| 411 |
+
},
|
| 412 |
+
"attributes": {}
|
| 413 |
+
}
|
| 414 |
+
},
|
| 415 |
+
"total_flos": 6692349849708888.0,
|
| 416 |
+
"train_batch_size": 80,
|
| 417 |
+
"trial_name": null,
|
| 418 |
+
"trial_params": null
|
| 419 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_43/all_results.json
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.8772935779816514,
|
| 4 |
+
"eval_loss": 0.3741566240787506,
|
| 5 |
+
"eval_runtime": 2.3158,
|
| 6 |
+
"eval_samples": 872,
|
| 7 |
+
"eval_samples_per_second": 376.55,
|
| 8 |
+
"eval_steps_per_second": 2.591,
|
| 9 |
+
"total_flos": 6697953466252656.0,
|
| 10 |
+
"train_loss": 0.25444065910997793,
|
| 11 |
+
"train_runtime": 275.6879,
|
| 12 |
+
"train_samples": 67349,
|
| 13 |
+
"train_samples_per_second": 732.883,
|
| 14 |
+
"train_steps_per_second": 9.163
|
| 15 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_43/eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.8772935779816514,
|
| 4 |
+
"eval_loss": 0.3741566240787506,
|
| 5 |
+
"eval_runtime": 2.3158,
|
| 6 |
+
"eval_samples": 872,
|
| 7 |
+
"eval_samples_per_second": 376.55,
|
| 8 |
+
"eval_steps_per_second": 2.591
|
| 9 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_43/run_command.txt
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
/usr/local/bin/python /workspace/run_glue_MHA_gated.py --model_name_or_path bert-base-multilingual-cased --dataset_name tmnam20/ViGLUE --task_name sst2 --max_seq_length 256 --do_train --do_eval --evaluation_strategy epoch --save_strategy no --logging_strategy steps --logging_steps 50 --report_to none --num_train_epochs 3 --learning_rate 2e-05 --per_device_train_batch_size 80 --per_device_eval_batch_size 168 --gradient_accumulation_steps 1 --dataloader_num_workers 2 --seed 43 --output_dir result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_43 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save --use_mha true --pooling_strategy mha_attention --n_heads 8 --head_dim 32 --classifier_hidden_dim 512 --head_dropout 0.05
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_43/train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 6697953466252656.0,
|
| 4 |
+
"train_loss": 0.25444065910997793,
|
| 5 |
+
"train_runtime": 275.6879,
|
| 6 |
+
"train_samples": 67349,
|
| 7 |
+
"train_samples_per_second": 732.883,
|
| 8 |
+
"train_steps_per_second": 9.163
|
| 9 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_43/trainer_state.json
ADDED
|
@@ -0,0 +1,419 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 3.0,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 2526,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.05938242280285035,
|
| 13 |
+
"grad_norm": 5.68363094329834,
|
| 14 |
+
"learning_rate": 1.9604117181314332e-05,
|
| 15 |
+
"loss": 0.5604,
|
| 16 |
+
"step": 50
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 0.1187648456057007,
|
| 20 |
+
"grad_norm": 5.383595943450928,
|
| 21 |
+
"learning_rate": 1.9208234362628665e-05,
|
| 22 |
+
"loss": 0.4559,
|
| 23 |
+
"step": 100
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.17814726840855108,
|
| 27 |
+
"grad_norm": 8.439021110534668,
|
| 28 |
+
"learning_rate": 1.8812351543942996e-05,
|
| 29 |
+
"loss": 0.4164,
|
| 30 |
+
"step": 150
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.2375296912114014,
|
| 34 |
+
"grad_norm": 5.211700439453125,
|
| 35 |
+
"learning_rate": 1.8416468725257326e-05,
|
| 36 |
+
"loss": 0.4252,
|
| 37 |
+
"step": 200
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.29691211401425177,
|
| 41 |
+
"grad_norm": 3.3225319385528564,
|
| 42 |
+
"learning_rate": 1.8020585906571656e-05,
|
| 43 |
+
"loss": 0.4033,
|
| 44 |
+
"step": 250
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.35629453681710216,
|
| 48 |
+
"grad_norm": 3.4626541137695312,
|
| 49 |
+
"learning_rate": 1.762470308788599e-05,
|
| 50 |
+
"loss": 0.3653,
|
| 51 |
+
"step": 300
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.4156769596199525,
|
| 55 |
+
"grad_norm": 4.0393290519714355,
|
| 56 |
+
"learning_rate": 1.7228820269200316e-05,
|
| 57 |
+
"loss": 0.352,
|
| 58 |
+
"step": 350
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.4750593824228028,
|
| 62 |
+
"grad_norm": 4.3066301345825195,
|
| 63 |
+
"learning_rate": 1.6832937450514647e-05,
|
| 64 |
+
"loss": 0.3545,
|
| 65 |
+
"step": 400
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.5344418052256532,
|
| 69 |
+
"grad_norm": 7.805842399597168,
|
| 70 |
+
"learning_rate": 1.643705463182898e-05,
|
| 71 |
+
"loss": 0.3399,
|
| 72 |
+
"step": 450
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 0.5938242280285035,
|
| 76 |
+
"grad_norm": 6.688758373260498,
|
| 77 |
+
"learning_rate": 1.604117181314331e-05,
|
| 78 |
+
"loss": 0.3302,
|
| 79 |
+
"step": 500
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.6532066508313539,
|
| 83 |
+
"grad_norm": 6.068962097167969,
|
| 84 |
+
"learning_rate": 1.564528899445764e-05,
|
| 85 |
+
"loss": 0.336,
|
| 86 |
+
"step": 550
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 0.7125890736342043,
|
| 90 |
+
"grad_norm": 8.058384895324707,
|
| 91 |
+
"learning_rate": 1.5249406175771972e-05,
|
| 92 |
+
"loss": 0.3052,
|
| 93 |
+
"step": 600
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.7719714964370546,
|
| 97 |
+
"grad_norm": 5.7268171310424805,
|
| 98 |
+
"learning_rate": 1.4853523357086304e-05,
|
| 99 |
+
"loss": 0.321,
|
| 100 |
+
"step": 650
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.831353919239905,
|
| 104 |
+
"grad_norm": 4.237183094024658,
|
| 105 |
+
"learning_rate": 1.4457640538400634e-05,
|
| 106 |
+
"loss": 0.3108,
|
| 107 |
+
"step": 700
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.8907363420427553,
|
| 111 |
+
"grad_norm": 5.330963611602783,
|
| 112 |
+
"learning_rate": 1.4061757719714966e-05,
|
| 113 |
+
"loss": 0.3159,
|
| 114 |
+
"step": 750
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.9501187648456056,
|
| 118 |
+
"grad_norm": 5.5379438400268555,
|
| 119 |
+
"learning_rate": 1.3665874901029297e-05,
|
| 120 |
+
"loss": 0.2955,
|
| 121 |
+
"step": 800
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 1.0,
|
| 125 |
+
"eval_accuracy": 0.8704128440366973,
|
| 126 |
+
"eval_loss": 0.33496010303497314,
|
| 127 |
+
"eval_runtime": 2.435,
|
| 128 |
+
"eval_samples_per_second": 358.116,
|
| 129 |
+
"eval_steps_per_second": 2.464,
|
| 130 |
+
"step": 842
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 1.009501187648456,
|
| 134 |
+
"grad_norm": 4.633709907531738,
|
| 135 |
+
"learning_rate": 1.3269992082343628e-05,
|
| 136 |
+
"loss": 0.2851,
|
| 137 |
+
"step": 850
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.0688836104513064,
|
| 141 |
+
"grad_norm": 4.599423408508301,
|
| 142 |
+
"learning_rate": 1.2874109263657959e-05,
|
| 143 |
+
"loss": 0.2422,
|
| 144 |
+
"step": 900
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 1.1282660332541568,
|
| 148 |
+
"grad_norm": 7.752073287963867,
|
| 149 |
+
"learning_rate": 1.247822644497229e-05,
|
| 150 |
+
"loss": 0.241,
|
| 151 |
+
"step": 950
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"epoch": 1.187648456057007,
|
| 155 |
+
"grad_norm": 4.42056131362915,
|
| 156 |
+
"learning_rate": 1.208234362628662e-05,
|
| 157 |
+
"loss": 0.2503,
|
| 158 |
+
"step": 1000
|
| 159 |
+
},
|
| 160 |
+
{
|
| 161 |
+
"epoch": 1.2470308788598574,
|
| 162 |
+
"grad_norm": 3.891842842102051,
|
| 163 |
+
"learning_rate": 1.168646080760095e-05,
|
| 164 |
+
"loss": 0.2386,
|
| 165 |
+
"step": 1050
|
| 166 |
+
},
|
| 167 |
+
{
|
| 168 |
+
"epoch": 1.3064133016627077,
|
| 169 |
+
"grad_norm": 5.100167274475098,
|
| 170 |
+
"learning_rate": 1.1290577988915281e-05,
|
| 171 |
+
"loss": 0.2385,
|
| 172 |
+
"step": 1100
|
| 173 |
+
},
|
| 174 |
+
{
|
| 175 |
+
"epoch": 1.365795724465558,
|
| 176 |
+
"grad_norm": 5.943244934082031,
|
| 177 |
+
"learning_rate": 1.0894695170229613e-05,
|
| 178 |
+
"loss": 0.2274,
|
| 179 |
+
"step": 1150
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"epoch": 1.4251781472684084,
|
| 183 |
+
"grad_norm": 5.428085803985596,
|
| 184 |
+
"learning_rate": 1.0498812351543943e-05,
|
| 185 |
+
"loss": 0.2455,
|
| 186 |
+
"step": 1200
|
| 187 |
+
},
|
| 188 |
+
{
|
| 189 |
+
"epoch": 1.484560570071259,
|
| 190 |
+
"grad_norm": 6.86477518081665,
|
| 191 |
+
"learning_rate": 1.0102929532858275e-05,
|
| 192 |
+
"loss": 0.2239,
|
| 193 |
+
"step": 1250
|
| 194 |
+
},
|
| 195 |
+
{
|
| 196 |
+
"epoch": 1.5439429928741093,
|
| 197 |
+
"grad_norm": 3.724605083465576,
|
| 198 |
+
"learning_rate": 9.707046714172605e-06,
|
| 199 |
+
"loss": 0.2273,
|
| 200 |
+
"step": 1300
|
| 201 |
+
},
|
| 202 |
+
{
|
| 203 |
+
"epoch": 1.6033254156769596,
|
| 204 |
+
"grad_norm": 5.287256240844727,
|
| 205 |
+
"learning_rate": 9.311163895486937e-06,
|
| 206 |
+
"loss": 0.2298,
|
| 207 |
+
"step": 1350
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 1.66270783847981,
|
| 211 |
+
"grad_norm": 5.92859411239624,
|
| 212 |
+
"learning_rate": 8.915281076801267e-06,
|
| 213 |
+
"loss": 0.2399,
|
| 214 |
+
"step": 1400
|
| 215 |
+
},
|
| 216 |
+
{
|
| 217 |
+
"epoch": 1.7220902612826603,
|
| 218 |
+
"grad_norm": 8.584405899047852,
|
| 219 |
+
"learning_rate": 8.5193982581156e-06,
|
| 220 |
+
"loss": 0.2122,
|
| 221 |
+
"step": 1450
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 1.7814726840855108,
|
| 225 |
+
"grad_norm": 6.123090744018555,
|
| 226 |
+
"learning_rate": 8.12351543942993e-06,
|
| 227 |
+
"loss": 0.2276,
|
| 228 |
+
"step": 1500
|
| 229 |
+
},
|
| 230 |
+
{
|
| 231 |
+
"epoch": 1.8408551068883612,
|
| 232 |
+
"grad_norm": 9.330291748046875,
|
| 233 |
+
"learning_rate": 7.72763262074426e-06,
|
| 234 |
+
"loss": 0.2194,
|
| 235 |
+
"step": 1550
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 1.9002375296912115,
|
| 239 |
+
"grad_norm": 7.708232402801514,
|
| 240 |
+
"learning_rate": 7.331749802058591e-06,
|
| 241 |
+
"loss": 0.2031,
|
| 242 |
+
"step": 1600
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 1.9596199524940618,
|
| 246 |
+
"grad_norm": 5.512419700622559,
|
| 247 |
+
"learning_rate": 6.935866983372922e-06,
|
| 248 |
+
"loss": 0.2188,
|
| 249 |
+
"step": 1650
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 2.0,
|
| 253 |
+
"eval_accuracy": 0.8704128440366973,
|
| 254 |
+
"eval_loss": 0.3823597729206085,
|
| 255 |
+
"eval_runtime": 2.403,
|
| 256 |
+
"eval_samples_per_second": 362.882,
|
| 257 |
+
"eval_steps_per_second": 2.497,
|
| 258 |
+
"step": 1684
|
| 259 |
+
},
|
| 260 |
+
{
|
| 261 |
+
"epoch": 2.019002375296912,
|
| 262 |
+
"grad_norm": 5.040703296661377,
|
| 263 |
+
"learning_rate": 6.539984164687253e-06,
|
| 264 |
+
"loss": 0.1973,
|
| 265 |
+
"step": 1700
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 2.0783847980997625,
|
| 269 |
+
"grad_norm": 11.075889587402344,
|
| 270 |
+
"learning_rate": 6.1441013460015846e-06,
|
| 271 |
+
"loss": 0.1776,
|
| 272 |
+
"step": 1750
|
| 273 |
+
},
|
| 274 |
+
{
|
| 275 |
+
"epoch": 2.137767220902613,
|
| 276 |
+
"grad_norm": 3.3569436073303223,
|
| 277 |
+
"learning_rate": 5.748218527315916e-06,
|
| 278 |
+
"loss": 0.1744,
|
| 279 |
+
"step": 1800
|
| 280 |
+
},
|
| 281 |
+
{
|
| 282 |
+
"epoch": 2.197149643705463,
|
| 283 |
+
"grad_norm": 7.500093460083008,
|
| 284 |
+
"learning_rate": 5.352335708630247e-06,
|
| 285 |
+
"loss": 0.1679,
|
| 286 |
+
"step": 1850
|
| 287 |
+
},
|
| 288 |
+
{
|
| 289 |
+
"epoch": 2.2565320665083135,
|
| 290 |
+
"grad_norm": 5.835613250732422,
|
| 291 |
+
"learning_rate": 4.956452889944577e-06,
|
| 292 |
+
"loss": 0.1765,
|
| 293 |
+
"step": 1900
|
| 294 |
+
},
|
| 295 |
+
{
|
| 296 |
+
"epoch": 2.315914489311164,
|
| 297 |
+
"grad_norm": 7.2991814613342285,
|
| 298 |
+
"learning_rate": 4.560570071258908e-06,
|
| 299 |
+
"loss": 0.1737,
|
| 300 |
+
"step": 1950
|
| 301 |
+
},
|
| 302 |
+
{
|
| 303 |
+
"epoch": 2.375296912114014,
|
| 304 |
+
"grad_norm": 7.424198150634766,
|
| 305 |
+
"learning_rate": 4.164687252573238e-06,
|
| 306 |
+
"loss": 0.1702,
|
| 307 |
+
"step": 2000
|
| 308 |
+
},
|
| 309 |
+
{
|
| 310 |
+
"epoch": 2.4346793349168645,
|
| 311 |
+
"grad_norm": 8.05273723602295,
|
| 312 |
+
"learning_rate": 3.7688044338875695e-06,
|
| 313 |
+
"loss": 0.1551,
|
| 314 |
+
"step": 2050
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"epoch": 2.494061757719715,
|
| 318 |
+
"grad_norm": 3.840614080429077,
|
| 319 |
+
"learning_rate": 3.3729216152019006e-06,
|
| 320 |
+
"loss": 0.1738,
|
| 321 |
+
"step": 2100
|
| 322 |
+
},
|
| 323 |
+
{
|
| 324 |
+
"epoch": 2.553444180522565,
|
| 325 |
+
"grad_norm": 4.635272979736328,
|
| 326 |
+
"learning_rate": 2.977038796516231e-06,
|
| 327 |
+
"loss": 0.1767,
|
| 328 |
+
"step": 2150
|
| 329 |
+
},
|
| 330 |
+
{
|
| 331 |
+
"epoch": 2.6128266033254155,
|
| 332 |
+
"grad_norm": 5.418165683746338,
|
| 333 |
+
"learning_rate": 2.581155977830562e-06,
|
| 334 |
+
"loss": 0.1631,
|
| 335 |
+
"step": 2200
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 2.6722090261282663,
|
| 339 |
+
"grad_norm": 6.6639933586120605,
|
| 340 |
+
"learning_rate": 2.1852731591448932e-06,
|
| 341 |
+
"loss": 0.1522,
|
| 342 |
+
"step": 2250
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 2.731591448931116,
|
| 346 |
+
"grad_norm": 6.732873916625977,
|
| 347 |
+
"learning_rate": 1.789390340459224e-06,
|
| 348 |
+
"loss": 0.18,
|
| 349 |
+
"step": 2300
|
| 350 |
+
},
|
| 351 |
+
{
|
| 352 |
+
"epoch": 2.790973871733967,
|
| 353 |
+
"grad_norm": 6.801496505737305,
|
| 354 |
+
"learning_rate": 1.3935075217735553e-06,
|
| 355 |
+
"loss": 0.1713,
|
| 356 |
+
"step": 2350
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 2.850356294536817,
|
| 360 |
+
"grad_norm": 5.231013298034668,
|
| 361 |
+
"learning_rate": 9.976247030878861e-07,
|
| 362 |
+
"loss": 0.1636,
|
| 363 |
+
"step": 2400
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"epoch": 2.9097387173396676,
|
| 367 |
+
"grad_norm": 4.750856399536133,
|
| 368 |
+
"learning_rate": 6.01741884402217e-07,
|
| 369 |
+
"loss": 0.1611,
|
| 370 |
+
"step": 2450
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"epoch": 2.969121140142518,
|
| 374 |
+
"grad_norm": 2.7668511867523193,
|
| 375 |
+
"learning_rate": 2.0585906571654793e-07,
|
| 376 |
+
"loss": 0.1707,
|
| 377 |
+
"step": 2500
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 3.0,
|
| 381 |
+
"eval_accuracy": 0.8772935779816514,
|
| 382 |
+
"eval_loss": 0.3741566240787506,
|
| 383 |
+
"eval_runtime": 2.3974,
|
| 384 |
+
"eval_samples_per_second": 363.735,
|
| 385 |
+
"eval_steps_per_second": 2.503,
|
| 386 |
+
"step": 2526
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 3.0,
|
| 390 |
+
"step": 2526,
|
| 391 |
+
"total_flos": 6697953466252656.0,
|
| 392 |
+
"train_loss": 0.25444065910997793,
|
| 393 |
+
"train_runtime": 275.6879,
|
| 394 |
+
"train_samples_per_second": 732.883,
|
| 395 |
+
"train_steps_per_second": 9.163
|
| 396 |
+
}
|
| 397 |
+
],
|
| 398 |
+
"logging_steps": 50,
|
| 399 |
+
"max_steps": 2526,
|
| 400 |
+
"num_input_tokens_seen": 0,
|
| 401 |
+
"num_train_epochs": 3,
|
| 402 |
+
"save_steps": 500,
|
| 403 |
+
"stateful_callbacks": {
|
| 404 |
+
"TrainerControl": {
|
| 405 |
+
"args": {
|
| 406 |
+
"should_epoch_stop": false,
|
| 407 |
+
"should_evaluate": false,
|
| 408 |
+
"should_log": false,
|
| 409 |
+
"should_save": false,
|
| 410 |
+
"should_training_stop": false
|
| 411 |
+
},
|
| 412 |
+
"attributes": {}
|
| 413 |
+
}
|
| 414 |
+
},
|
| 415 |
+
"total_flos": 6697953466252656.0,
|
| 416 |
+
"train_batch_size": 80,
|
| 417 |
+
"trial_name": null,
|
| 418 |
+
"trial_params": null
|
| 419 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_44/all_results.json
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.8772935779816514,
|
| 4 |
+
"eval_loss": 0.37144818902015686,
|
| 5 |
+
"eval_runtime": 2.1881,
|
| 6 |
+
"eval_samples": 872,
|
| 7 |
+
"eval_samples_per_second": 398.518,
|
| 8 |
+
"eval_steps_per_second": 2.742,
|
| 9 |
+
"total_flos": 6694669755293622.0,
|
| 10 |
+
"train_loss": 0.2535441892069668,
|
| 11 |
+
"train_runtime": 275.9521,
|
| 12 |
+
"train_samples": 67349,
|
| 13 |
+
"train_samples_per_second": 732.181,
|
| 14 |
+
"train_steps_per_second": 9.154
|
| 15 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_44/eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.8772935779816514,
|
| 4 |
+
"eval_loss": 0.37144818902015686,
|
| 5 |
+
"eval_runtime": 2.1881,
|
| 6 |
+
"eval_samples": 872,
|
| 7 |
+
"eval_samples_per_second": 398.518,
|
| 8 |
+
"eval_steps_per_second": 2.742
|
| 9 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_44/run_command.txt
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
/usr/local/bin/python /workspace/run_glue_MHA_gated.py --model_name_or_path bert-base-multilingual-cased --dataset_name tmnam20/ViGLUE --task_name sst2 --max_seq_length 256 --do_train --do_eval --evaluation_strategy epoch --save_strategy no --logging_strategy steps --logging_steps 50 --report_to none --num_train_epochs 3 --learning_rate 2e-05 --per_device_train_batch_size 80 --per_device_eval_batch_size 168 --gradient_accumulation_steps 1 --dataloader_num_workers 2 --seed 44 --output_dir result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_44 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save --use_mha true --pooling_strategy mha_attention --n_heads 8 --head_dim 32 --classifier_hidden_dim 512 --head_dropout 0.05
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_44/train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 6694669755293622.0,
|
| 4 |
+
"train_loss": 0.2535441892069668,
|
| 5 |
+
"train_runtime": 275.9521,
|
| 6 |
+
"train_samples": 67349,
|
| 7 |
+
"train_samples_per_second": 732.181,
|
| 8 |
+
"train_steps_per_second": 9.154
|
| 9 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/mha_attention/seed_44/trainer_state.json
ADDED
|
@@ -0,0 +1,419 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 3.0,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 2526,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.05938242280285035,
|
| 13 |
+
"grad_norm": 3.813469886779785,
|
| 14 |
+
"learning_rate": 1.9604117181314332e-05,
|
| 15 |
+
"loss": 0.5785,
|
| 16 |
+
"step": 50
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 0.1187648456057007,
|
| 20 |
+
"grad_norm": 3.149343967437744,
|
| 21 |
+
"learning_rate": 1.9208234362628665e-05,
|
| 22 |
+
"loss": 0.4709,
|
| 23 |
+
"step": 100
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.17814726840855108,
|
| 27 |
+
"grad_norm": 6.103271007537842,
|
| 28 |
+
"learning_rate": 1.8812351543942996e-05,
|
| 29 |
+
"loss": 0.4368,
|
| 30 |
+
"step": 150
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.2375296912114014,
|
| 34 |
+
"grad_norm": 4.508041858673096,
|
| 35 |
+
"learning_rate": 1.8416468725257326e-05,
|
| 36 |
+
"loss": 0.405,
|
| 37 |
+
"step": 200
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.29691211401425177,
|
| 41 |
+
"grad_norm": 5.787165641784668,
|
| 42 |
+
"learning_rate": 1.8020585906571656e-05,
|
| 43 |
+
"loss": 0.3782,
|
| 44 |
+
"step": 250
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.35629453681710216,
|
| 48 |
+
"grad_norm": 5.334907054901123,
|
| 49 |
+
"learning_rate": 1.762470308788599e-05,
|
| 50 |
+
"loss": 0.3946,
|
| 51 |
+
"step": 300
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.4156769596199525,
|
| 55 |
+
"grad_norm": 3.885115623474121,
|
| 56 |
+
"learning_rate": 1.7228820269200316e-05,
|
| 57 |
+
"loss": 0.373,
|
| 58 |
+
"step": 350
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.4750593824228028,
|
| 62 |
+
"grad_norm": 4.70413875579834,
|
| 63 |
+
"learning_rate": 1.6832937450514647e-05,
|
| 64 |
+
"loss": 0.3428,
|
| 65 |
+
"step": 400
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.5344418052256532,
|
| 69 |
+
"grad_norm": 4.990809917449951,
|
| 70 |
+
"learning_rate": 1.643705463182898e-05,
|
| 71 |
+
"loss": 0.3244,
|
| 72 |
+
"step": 450
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 0.5938242280285035,
|
| 76 |
+
"grad_norm": 7.042197227478027,
|
| 77 |
+
"learning_rate": 1.604117181314331e-05,
|
| 78 |
+
"loss": 0.3423,
|
| 79 |
+
"step": 500
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.6532066508313539,
|
| 83 |
+
"grad_norm": 4.276707649230957,
|
| 84 |
+
"learning_rate": 1.564528899445764e-05,
|
| 85 |
+
"loss": 0.3391,
|
| 86 |
+
"step": 550
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 0.7125890736342043,
|
| 90 |
+
"grad_norm": 4.52197265625,
|
| 91 |
+
"learning_rate": 1.5249406175771972e-05,
|
| 92 |
+
"loss": 0.32,
|
| 93 |
+
"step": 600
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.7719714964370546,
|
| 97 |
+
"grad_norm": 5.242362022399902,
|
| 98 |
+
"learning_rate": 1.4853523357086304e-05,
|
| 99 |
+
"loss": 0.3104,
|
| 100 |
+
"step": 650
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.831353919239905,
|
| 104 |
+
"grad_norm": 4.8427252769470215,
|
| 105 |
+
"learning_rate": 1.4457640538400634e-05,
|
| 106 |
+
"loss": 0.2987,
|
| 107 |
+
"step": 700
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.8907363420427553,
|
| 111 |
+
"grad_norm": 8.30374526977539,
|
| 112 |
+
"learning_rate": 1.4061757719714966e-05,
|
| 113 |
+
"loss": 0.2991,
|
| 114 |
+
"step": 750
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.9501187648456056,
|
| 118 |
+
"grad_norm": 4.5999979972839355,
|
| 119 |
+
"learning_rate": 1.3665874901029297e-05,
|
| 120 |
+
"loss": 0.3045,
|
| 121 |
+
"step": 800
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 1.0,
|
| 125 |
+
"eval_accuracy": 0.8761467889908257,
|
| 126 |
+
"eval_loss": 0.33226579427719116,
|
| 127 |
+
"eval_runtime": 2.3765,
|
| 128 |
+
"eval_samples_per_second": 366.923,
|
| 129 |
+
"eval_steps_per_second": 2.525,
|
| 130 |
+
"step": 842
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 1.009501187648456,
|
| 134 |
+
"grad_norm": 9.606904983520508,
|
| 135 |
+
"learning_rate": 1.3269992082343628e-05,
|
| 136 |
+
"loss": 0.2782,
|
| 137 |
+
"step": 850
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.0688836104513064,
|
| 141 |
+
"grad_norm": 7.778744220733643,
|
| 142 |
+
"learning_rate": 1.2874109263657959e-05,
|
| 143 |
+
"loss": 0.2403,
|
| 144 |
+
"step": 900
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 1.1282660332541568,
|
| 148 |
+
"grad_norm": 4.415440559387207,
|
| 149 |
+
"learning_rate": 1.247822644497229e-05,
|
| 150 |
+
"loss": 0.2173,
|
| 151 |
+
"step": 950
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"epoch": 1.187648456057007,
|
| 155 |
+
"grad_norm": 6.505444049835205,
|
| 156 |
+
"learning_rate": 1.208234362628662e-05,
|
| 157 |
+
"loss": 0.232,
|
| 158 |
+
"step": 1000
|
| 159 |
+
},
|
| 160 |
+
{
|
| 161 |
+
"epoch": 1.2470308788598574,
|
| 162 |
+
"grad_norm": 6.76679801940918,
|
| 163 |
+
"learning_rate": 1.168646080760095e-05,
|
| 164 |
+
"loss": 0.2335,
|
| 165 |
+
"step": 1050
|
| 166 |
+
},
|
| 167 |
+
{
|
| 168 |
+
"epoch": 1.3064133016627077,
|
| 169 |
+
"grad_norm": 5.687557220458984,
|
| 170 |
+
"learning_rate": 1.1290577988915281e-05,
|
| 171 |
+
"loss": 0.2346,
|
| 172 |
+
"step": 1100
|
| 173 |
+
},
|
| 174 |
+
{
|
| 175 |
+
"epoch": 1.365795724465558,
|
| 176 |
+
"grad_norm": 5.521519660949707,
|
| 177 |
+
"learning_rate": 1.0894695170229613e-05,
|
| 178 |
+
"loss": 0.2417,
|
| 179 |
+
"step": 1150
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"epoch": 1.4251781472684084,
|
| 183 |
+
"grad_norm": 5.35023832321167,
|
| 184 |
+
"learning_rate": 1.0498812351543943e-05,
|
| 185 |
+
"loss": 0.2298,
|
| 186 |
+
"step": 1200
|
| 187 |
+
},
|
| 188 |
+
{
|
| 189 |
+
"epoch": 1.484560570071259,
|
| 190 |
+
"grad_norm": 5.0389533042907715,
|
| 191 |
+
"learning_rate": 1.0102929532858275e-05,
|
| 192 |
+
"loss": 0.2219,
|
| 193 |
+
"step": 1250
|
| 194 |
+
},
|
| 195 |
+
{
|
| 196 |
+
"epoch": 1.5439429928741093,
|
| 197 |
+
"grad_norm": 5.653523921966553,
|
| 198 |
+
"learning_rate": 9.707046714172605e-06,
|
| 199 |
+
"loss": 0.2389,
|
| 200 |
+
"step": 1300
|
| 201 |
+
},
|
| 202 |
+
{
|
| 203 |
+
"epoch": 1.6033254156769596,
|
| 204 |
+
"grad_norm": 4.59339714050293,
|
| 205 |
+
"learning_rate": 9.311163895486937e-06,
|
| 206 |
+
"loss": 0.2334,
|
| 207 |
+
"step": 1350
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 1.66270783847981,
|
| 211 |
+
"grad_norm": 4.2259063720703125,
|
| 212 |
+
"learning_rate": 8.915281076801267e-06,
|
| 213 |
+
"loss": 0.2064,
|
| 214 |
+
"step": 1400
|
| 215 |
+
},
|
| 216 |
+
{
|
| 217 |
+
"epoch": 1.7220902612826603,
|
| 218 |
+
"grad_norm": 6.655894756317139,
|
| 219 |
+
"learning_rate": 8.5193982581156e-06,
|
| 220 |
+
"loss": 0.2284,
|
| 221 |
+
"step": 1450
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 1.7814726840855108,
|
| 225 |
+
"grad_norm": 5.455095291137695,
|
| 226 |
+
"learning_rate": 8.12351543942993e-06,
|
| 227 |
+
"loss": 0.2221,
|
| 228 |
+
"step": 1500
|
| 229 |
+
},
|
| 230 |
+
{
|
| 231 |
+
"epoch": 1.8408551068883612,
|
| 232 |
+
"grad_norm": 6.1030778884887695,
|
| 233 |
+
"learning_rate": 7.72763262074426e-06,
|
| 234 |
+
"loss": 0.2328,
|
| 235 |
+
"step": 1550
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 1.9002375296912115,
|
| 239 |
+
"grad_norm": 7.211266994476318,
|
| 240 |
+
"learning_rate": 7.331749802058591e-06,
|
| 241 |
+
"loss": 0.2109,
|
| 242 |
+
"step": 1600
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 1.9596199524940618,
|
| 246 |
+
"grad_norm": 4.45063591003418,
|
| 247 |
+
"learning_rate": 6.935866983372922e-06,
|
| 248 |
+
"loss": 0.226,
|
| 249 |
+
"step": 1650
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 2.0,
|
| 253 |
+
"eval_accuracy": 0.875,
|
| 254 |
+
"eval_loss": 0.34283238649368286,
|
| 255 |
+
"eval_runtime": 2.2305,
|
| 256 |
+
"eval_samples_per_second": 390.945,
|
| 257 |
+
"eval_steps_per_second": 2.69,
|
| 258 |
+
"step": 1684
|
| 259 |
+
},
|
| 260 |
+
{
|
| 261 |
+
"epoch": 2.019002375296912,
|
| 262 |
+
"grad_norm": 4.772740364074707,
|
| 263 |
+
"learning_rate": 6.539984164687253e-06,
|
| 264 |
+
"loss": 0.2003,
|
| 265 |
+
"step": 1700
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 2.0783847980997625,
|
| 269 |
+
"grad_norm": 3.9535157680511475,
|
| 270 |
+
"learning_rate": 6.1441013460015846e-06,
|
| 271 |
+
"loss": 0.18,
|
| 272 |
+
"step": 1750
|
| 273 |
+
},
|
| 274 |
+
{
|
| 275 |
+
"epoch": 2.137767220902613,
|
| 276 |
+
"grad_norm": 5.986626148223877,
|
| 277 |
+
"learning_rate": 5.748218527315916e-06,
|
| 278 |
+
"loss": 0.1626,
|
| 279 |
+
"step": 1800
|
| 280 |
+
},
|
| 281 |
+
{
|
| 282 |
+
"epoch": 2.197149643705463,
|
| 283 |
+
"grad_norm": 5.395092487335205,
|
| 284 |
+
"learning_rate": 5.352335708630247e-06,
|
| 285 |
+
"loss": 0.175,
|
| 286 |
+
"step": 1850
|
| 287 |
+
},
|
| 288 |
+
{
|
| 289 |
+
"epoch": 2.2565320665083135,
|
| 290 |
+
"grad_norm": 6.2225847244262695,
|
| 291 |
+
"learning_rate": 4.956452889944577e-06,
|
| 292 |
+
"loss": 0.1567,
|
| 293 |
+
"step": 1900
|
| 294 |
+
},
|
| 295 |
+
{
|
| 296 |
+
"epoch": 2.315914489311164,
|
| 297 |
+
"grad_norm": 7.157998561859131,
|
| 298 |
+
"learning_rate": 4.560570071258908e-06,
|
| 299 |
+
"loss": 0.1729,
|
| 300 |
+
"step": 1950
|
| 301 |
+
},
|
| 302 |
+
{
|
| 303 |
+
"epoch": 2.375296912114014,
|
| 304 |
+
"grad_norm": 7.06791353225708,
|
| 305 |
+
"learning_rate": 4.164687252573238e-06,
|
| 306 |
+
"loss": 0.164,
|
| 307 |
+
"step": 2000
|
| 308 |
+
},
|
| 309 |
+
{
|
| 310 |
+
"epoch": 2.4346793349168645,
|
| 311 |
+
"grad_norm": 6.978614807128906,
|
| 312 |
+
"learning_rate": 3.7688044338875695e-06,
|
| 313 |
+
"loss": 0.1593,
|
| 314 |
+
"step": 2050
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"epoch": 2.494061757719715,
|
| 318 |
+
"grad_norm": 3.5602076053619385,
|
| 319 |
+
"learning_rate": 3.3729216152019006e-06,
|
| 320 |
+
"loss": 0.161,
|
| 321 |
+
"step": 2100
|
| 322 |
+
},
|
| 323 |
+
{
|
| 324 |
+
"epoch": 2.553444180522565,
|
| 325 |
+
"grad_norm": 7.145963668823242,
|
| 326 |
+
"learning_rate": 2.977038796516231e-06,
|
| 327 |
+
"loss": 0.1621,
|
| 328 |
+
"step": 2150
|
| 329 |
+
},
|
| 330 |
+
{
|
| 331 |
+
"epoch": 2.6128266033254155,
|
| 332 |
+
"grad_norm": 4.337043762207031,
|
| 333 |
+
"learning_rate": 2.581155977830562e-06,
|
| 334 |
+
"loss": 0.1742,
|
| 335 |
+
"step": 2200
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 2.6722090261282663,
|
| 339 |
+
"grad_norm": 5.760071754455566,
|
| 340 |
+
"learning_rate": 2.1852731591448932e-06,
|
| 341 |
+
"loss": 0.1659,
|
| 342 |
+
"step": 2250
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 2.731591448931116,
|
| 346 |
+
"grad_norm": 5.518279552459717,
|
| 347 |
+
"learning_rate": 1.789390340459224e-06,
|
| 348 |
+
"loss": 0.1817,
|
| 349 |
+
"step": 2300
|
| 350 |
+
},
|
| 351 |
+
{
|
| 352 |
+
"epoch": 2.790973871733967,
|
| 353 |
+
"grad_norm": 5.907181262969971,
|
| 354 |
+
"learning_rate": 1.3935075217735553e-06,
|
| 355 |
+
"loss": 0.1624,
|
| 356 |
+
"step": 2350
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 2.850356294536817,
|
| 360 |
+
"grad_norm": 6.645471572875977,
|
| 361 |
+
"learning_rate": 9.976247030878861e-07,
|
| 362 |
+
"loss": 0.1743,
|
| 363 |
+
"step": 2400
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"epoch": 2.9097387173396676,
|
| 367 |
+
"grad_norm": 5.510932445526123,
|
| 368 |
+
"learning_rate": 6.01741884402217e-07,
|
| 369 |
+
"loss": 0.1611,
|
| 370 |
+
"step": 2450
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"epoch": 2.969121140142518,
|
| 374 |
+
"grad_norm": 6.247923851013184,
|
| 375 |
+
"learning_rate": 2.0585906571654793e-07,
|
| 376 |
+
"loss": 0.1632,
|
| 377 |
+
"step": 2500
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 3.0,
|
| 381 |
+
"eval_accuracy": 0.8772935779816514,
|
| 382 |
+
"eval_loss": 0.37144818902015686,
|
| 383 |
+
"eval_runtime": 2.2019,
|
| 384 |
+
"eval_samples_per_second": 396.017,
|
| 385 |
+
"eval_steps_per_second": 2.725,
|
| 386 |
+
"step": 2526
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 3.0,
|
| 390 |
+
"step": 2526,
|
| 391 |
+
"total_flos": 6694669755293622.0,
|
| 392 |
+
"train_loss": 0.2535441892069668,
|
| 393 |
+
"train_runtime": 275.9521,
|
| 394 |
+
"train_samples_per_second": 732.181,
|
| 395 |
+
"train_steps_per_second": 9.154
|
| 396 |
+
}
|
| 397 |
+
],
|
| 398 |
+
"logging_steps": 50,
|
| 399 |
+
"max_steps": 2526,
|
| 400 |
+
"num_input_tokens_seen": 0,
|
| 401 |
+
"num_train_epochs": 3,
|
| 402 |
+
"save_steps": 500,
|
| 403 |
+
"stateful_callbacks": {
|
| 404 |
+
"TrainerControl": {
|
| 405 |
+
"args": {
|
| 406 |
+
"should_epoch_stop": false,
|
| 407 |
+
"should_evaluate": false,
|
| 408 |
+
"should_log": false,
|
| 409 |
+
"should_save": false,
|
| 410 |
+
"should_training_stop": false
|
| 411 |
+
},
|
| 412 |
+
"attributes": {}
|
| 413 |
+
}
|
| 414 |
+
},
|
| 415 |
+
"total_flos": 6694669755293622.0,
|
| 416 |
+
"train_batch_size": 80,
|
| 417 |
+
"trial_name": null,
|
| 418 |
+
"trial_params": null
|
| 419 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/multi_branch_average/seed_42/all_results.json
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.8704128440366973,
|
| 4 |
+
"eval_loss": 0.3955666124820709,
|
| 5 |
+
"eval_runtime": 2.4397,
|
| 6 |
+
"eval_samples": 872,
|
| 7 |
+
"eval_samples_per_second": 357.416,
|
| 8 |
+
"eval_steps_per_second": 2.459,
|
| 9 |
+
"total_flos": 6934924671671304.0,
|
| 10 |
+
"train_loss": 0.2517735588106954,
|
| 11 |
+
"train_runtime": 273.9564,
|
| 12 |
+
"train_samples": 67349,
|
| 13 |
+
"train_samples_per_second": 737.515,
|
| 14 |
+
"train_steps_per_second": 9.22
|
| 15 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/multi_branch_average/seed_42/eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.8704128440366973,
|
| 4 |
+
"eval_loss": 0.3955666124820709,
|
| 5 |
+
"eval_runtime": 2.4397,
|
| 6 |
+
"eval_samples": 872,
|
| 7 |
+
"eval_samples_per_second": 357.416,
|
| 8 |
+
"eval_steps_per_second": 2.459
|
| 9 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/multi_branch_average/seed_42/run_command.txt
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
/usr/local/bin/python /workspace/run_glue_MHA_gated.py --model_name_or_path bert-base-multilingual-cased --dataset_name tmnam20/ViGLUE --task_name sst2 --max_seq_length 256 --do_train --do_eval --evaluation_strategy epoch --save_strategy no --logging_strategy steps --logging_steps 50 --report_to none --num_train_epochs 3 --learning_rate 2e-05 --per_device_train_batch_size 80 --per_device_eval_batch_size 168 --gradient_accumulation_steps 1 --dataloader_num_workers 2 --seed 42 --output_dir result_ablation_mbert_paper/sst2/mBERT/multi_branch_average/seed_42 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save --use_mha true --pooling_strategy multi_branch_average --n_heads 8 --head_dim 32 --classifier_hidden_dim 512 --head_dropout 0.05
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/multi_branch_average/seed_42/train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 6934924671671304.0,
|
| 4 |
+
"train_loss": 0.2517735588106954,
|
| 5 |
+
"train_runtime": 273.9564,
|
| 6 |
+
"train_samples": 67349,
|
| 7 |
+
"train_samples_per_second": 737.515,
|
| 8 |
+
"train_steps_per_second": 9.22
|
| 9 |
+
}
|
mbert_paper_metrics/result_ablation_mbert_paper/sst2/mBERT/multi_branch_average/seed_42/trainer_state.json
ADDED
|
@@ -0,0 +1,419 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 3.0,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 2526,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.05938242280285035,
|
| 13 |
+
"grad_norm": 7.919157028198242,
|
| 14 |
+
"learning_rate": 1.9604117181314332e-05,
|
| 15 |
+
"loss": 0.5549,
|
| 16 |
+
"step": 50
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 0.1187648456057007,
|
| 20 |
+
"grad_norm": 3.3355274200439453,
|
| 21 |
+
"learning_rate": 1.9208234362628665e-05,
|
| 22 |
+
"loss": 0.4602,
|
| 23 |
+
"step": 100
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.17814726840855108,
|
| 27 |
+
"grad_norm": 4.203125953674316,
|
| 28 |
+
"learning_rate": 1.8812351543942996e-05,
|
| 29 |
+
"loss": 0.411,
|
| 30 |
+
"step": 150
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.2375296912114014,
|
| 34 |
+
"grad_norm": 5.096848011016846,
|
| 35 |
+
"learning_rate": 1.8416468725257326e-05,
|
| 36 |
+
"loss": 0.4103,
|
| 37 |
+
"step": 200
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.29691211401425177,
|
| 41 |
+
"grad_norm": 5.2811384201049805,
|
| 42 |
+
"learning_rate": 1.8020585906571656e-05,
|
| 43 |
+
"loss": 0.3842,
|
| 44 |
+
"step": 250
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.35629453681710216,
|
| 48 |
+
"grad_norm": 4.095509052276611,
|
| 49 |
+
"learning_rate": 1.762470308788599e-05,
|
| 50 |
+
"loss": 0.3794,
|
| 51 |
+
"step": 300
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.4156769596199525,
|
| 55 |
+
"grad_norm": 4.775886535644531,
|
| 56 |
+
"learning_rate": 1.7228820269200316e-05,
|
| 57 |
+
"loss": 0.3501,
|
| 58 |
+
"step": 350
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.4750593824228028,
|
| 62 |
+
"grad_norm": 6.607873439788818,
|
| 63 |
+
"learning_rate": 1.6832937450514647e-05,
|
| 64 |
+
"loss": 0.3407,
|
| 65 |
+
"step": 400
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.5344418052256532,
|
| 69 |
+
"grad_norm": 5.063623428344727,
|
| 70 |
+
"learning_rate": 1.643705463182898e-05,
|
| 71 |
+
"loss": 0.3419,
|
| 72 |
+
"step": 450
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 0.5938242280285035,
|
| 76 |
+
"grad_norm": 3.4853391647338867,
|
| 77 |
+
"learning_rate": 1.604117181314331e-05,
|
| 78 |
+
"loss": 0.3246,
|
| 79 |
+
"step": 500
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.6532066508313539,
|
| 83 |
+
"grad_norm": 3.693147659301758,
|
| 84 |
+
"learning_rate": 1.564528899445764e-05,
|
| 85 |
+
"loss": 0.3261,
|
| 86 |
+
"step": 550
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 0.7125890736342043,
|
| 90 |
+
"grad_norm": 3.6549038887023926,
|
| 91 |
+
"learning_rate": 1.5249406175771972e-05,
|
| 92 |
+
"loss": 0.3275,
|
| 93 |
+
"step": 600
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.7719714964370546,
|
| 97 |
+
"grad_norm": 4.986508846282959,
|
| 98 |
+
"learning_rate": 1.4853523357086304e-05,
|
| 99 |
+
"loss": 0.3129,
|
| 100 |
+
"step": 650
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.831353919239905,
|
| 104 |
+
"grad_norm": 3.98590087890625,
|
| 105 |
+
"learning_rate": 1.4457640538400634e-05,
|
| 106 |
+
"loss": 0.3077,
|
| 107 |
+
"step": 700
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.8907363420427553,
|
| 111 |
+
"grad_norm": 6.5343732833862305,
|
| 112 |
+
"learning_rate": 1.4061757719714966e-05,
|
| 113 |
+
"loss": 0.3019,
|
| 114 |
+
"step": 750
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.9501187648456056,
|
| 118 |
+
"grad_norm": 3.427091598510742,
|
| 119 |
+
"learning_rate": 1.3665874901029297e-05,
|
| 120 |
+
"loss": 0.2831,
|
| 121 |
+
"step": 800
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 1.0,
|
| 125 |
+
"eval_accuracy": 0.8772935779816514,
|
| 126 |
+
"eval_loss": 0.3290562033653259,
|
| 127 |
+
"eval_runtime": 2.5766,
|
| 128 |
+
"eval_samples_per_second": 338.433,
|
| 129 |
+
"eval_steps_per_second": 2.329,
|
| 130 |
+
"step": 842
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 1.009501187648456,
|
| 134 |
+
"grad_norm": 5.138549327850342,
|
| 135 |
+
"learning_rate": 1.3269992082343628e-05,
|
| 136 |
+
"loss": 0.2992,
|
| 137 |
+
"step": 850
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.0688836104513064,
|
| 141 |
+
"grad_norm": 3.9386653900146484,
|
| 142 |
+
"learning_rate": 1.2874109263657959e-05,
|
| 143 |
+
"loss": 0.2598,
|
| 144 |
+
"step": 900
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 1.1282660332541568,
|
| 148 |
+
"grad_norm": 4.604132652282715,
|
| 149 |
+
"learning_rate": 1.247822644497229e-05,
|
| 150 |
+
"loss": 0.2351,
|
| 151 |
+
"step": 950
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"epoch": 1.187648456057007,
|
| 155 |
+
"grad_norm": 9.215094566345215,
|
| 156 |
+
"learning_rate": 1.208234362628662e-05,
|
| 157 |
+
"loss": 0.2301,
|
| 158 |
+
"step": 1000
|
| 159 |
+
},
|
| 160 |
+
{
|
| 161 |
+
"epoch": 1.2470308788598574,
|
| 162 |
+
"grad_norm": 4.37424898147583,
|
| 163 |
+
"learning_rate": 1.168646080760095e-05,
|
| 164 |
+
"loss": 0.2425,
|
| 165 |
+
"step": 1050
|
| 166 |
+
},
|
| 167 |
+
{
|
| 168 |
+
"epoch": 1.3064133016627077,
|
| 169 |
+
"grad_norm": 4.911233901977539,
|
| 170 |
+
"learning_rate": 1.1290577988915281e-05,
|
| 171 |
+
"loss": 0.2383,
|
| 172 |
+
"step": 1100
|
| 173 |
+
},
|
| 174 |
+
{
|
| 175 |
+
"epoch": 1.365795724465558,
|
| 176 |
+
"grad_norm": 4.536508560180664,
|
| 177 |
+
"learning_rate": 1.0894695170229613e-05,
|
| 178 |
+
"loss": 0.2223,
|
| 179 |
+
"step": 1150
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"epoch": 1.4251781472684084,
|
| 183 |
+
"grad_norm": 7.142017364501953,
|
| 184 |
+
"learning_rate": 1.0498812351543943e-05,
|
| 185 |
+
"loss": 0.2173,
|
| 186 |
+
"step": 1200
|
| 187 |
+
},
|
| 188 |
+
{
|
| 189 |
+
"epoch": 1.484560570071259,
|
| 190 |
+
"grad_norm": 5.40737247467041,
|
| 191 |
+
"learning_rate": 1.0102929532858275e-05,
|
| 192 |
+
"loss": 0.2345,
|
| 193 |
+
"step": 1250
|
| 194 |
+
},
|
| 195 |
+
{
|
| 196 |
+
"epoch": 1.5439429928741093,
|
| 197 |
+
"grad_norm": 7.336752414703369,
|
| 198 |
+
"learning_rate": 9.707046714172605e-06,
|
| 199 |
+
"loss": 0.2366,
|
| 200 |
+
"step": 1300
|
| 201 |
+
},
|
| 202 |
+
{
|
| 203 |
+
"epoch": 1.6033254156769596,
|
| 204 |
+
"grad_norm": 10.44306468963623,
|
| 205 |
+
"learning_rate": 9.311163895486937e-06,
|
| 206 |
+
"loss": 0.2085,
|
| 207 |
+
"step": 1350
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 1.66270783847981,
|
| 211 |
+
"grad_norm": 4.404452323913574,
|
| 212 |
+
"learning_rate": 8.915281076801267e-06,
|
| 213 |
+
"loss": 0.2437,
|
| 214 |
+
"step": 1400
|
| 215 |
+
},
|
| 216 |
+
{
|
| 217 |
+
"epoch": 1.7220902612826603,
|
| 218 |
+
"grad_norm": 5.115004062652588,
|
| 219 |
+
"learning_rate": 8.5193982581156e-06,
|
| 220 |
+
"loss": 0.2059,
|
| 221 |
+
"step": 1450
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 1.7814726840855108,
|
| 225 |
+
"grad_norm": 5.785116672515869,
|
| 226 |
+
"learning_rate": 8.12351543942993e-06,
|
| 227 |
+
"loss": 0.2108,
|
| 228 |
+
"step": 1500
|
| 229 |
+
},
|
| 230 |
+
{
|
| 231 |
+
"epoch": 1.8408551068883612,
|
| 232 |
+
"grad_norm": 5.2329630851745605,
|
| 233 |
+
"learning_rate": 7.72763262074426e-06,
|
| 234 |
+
"loss": 0.2227,
|
| 235 |
+
"step": 1550
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 1.9002375296912115,
|
| 239 |
+
"grad_norm": 4.726866245269775,
|
| 240 |
+
"learning_rate": 7.331749802058591e-06,
|
| 241 |
+
"loss": 0.2181,
|
| 242 |
+
"step": 1600
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 1.9596199524940618,
|
| 246 |
+
"grad_norm": 7.0507283210754395,
|
| 247 |
+
"learning_rate": 6.935866983372922e-06,
|
| 248 |
+
"loss": 0.2143,
|
| 249 |
+
"step": 1650
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 2.0,
|
| 253 |
+
"eval_accuracy": 0.8715596330275229,
|
| 254 |
+
"eval_loss": 0.3729800581932068,
|
| 255 |
+
"eval_runtime": 2.4343,
|
| 256 |
+
"eval_samples_per_second": 358.212,
|
| 257 |
+
"eval_steps_per_second": 2.465,
|
| 258 |
+
"step": 1684
|
| 259 |
+
},
|
| 260 |
+
{
|
| 261 |
+
"epoch": 2.019002375296912,
|
| 262 |
+
"grad_norm": 4.931351184844971,
|
| 263 |
+
"learning_rate": 6.539984164687253e-06,
|
| 264 |
+
"loss": 0.1928,
|
| 265 |
+
"step": 1700
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 2.0783847980997625,
|
| 269 |
+
"grad_norm": 6.169066429138184,
|
| 270 |
+
"learning_rate": 6.1441013460015846e-06,
|
| 271 |
+
"loss": 0.1756,
|
| 272 |
+
"step": 1750
|
| 273 |
+
},
|
| 274 |
+
{
|
| 275 |
+
"epoch": 2.137767220902613,
|
| 276 |
+
"grad_norm": 2.093853712081909,
|
| 277 |
+
"learning_rate": 5.748218527315916e-06,
|
| 278 |
+
"loss": 0.163,
|
| 279 |
+
"step": 1800
|
| 280 |
+
},
|
| 281 |
+
{
|
| 282 |
+
"epoch": 2.197149643705463,
|
| 283 |
+
"grad_norm": 4.834881782531738,
|
| 284 |
+
"learning_rate": 5.352335708630247e-06,
|
| 285 |
+
"loss": 0.1706,
|
| 286 |
+
"step": 1850
|
| 287 |
+
},
|
| 288 |
+
{
|
| 289 |
+
"epoch": 2.2565320665083135,
|
| 290 |
+
"grad_norm": 7.731045722961426,
|
| 291 |
+
"learning_rate": 4.956452889944577e-06,
|
| 292 |
+
"loss": 0.1774,
|
| 293 |
+
"step": 1900
|
| 294 |
+
},
|
| 295 |
+
{
|
| 296 |
+
"epoch": 2.315914489311164,
|
| 297 |
+
"grad_norm": 6.053497791290283,
|
| 298 |
+
"learning_rate": 4.560570071258908e-06,
|
| 299 |
+
"loss": 0.1682,
|
| 300 |
+
"step": 1950
|
| 301 |
+
},
|
| 302 |
+
{
|
| 303 |
+
"epoch": 2.375296912114014,
|
| 304 |
+
"grad_norm": 7.780023574829102,
|
| 305 |
+
"learning_rate": 4.164687252573238e-06,
|
| 306 |
+
"loss": 0.1761,
|
| 307 |
+
"step": 2000
|
| 308 |
+
},
|
| 309 |
+
{
|
| 310 |
+
"epoch": 2.4346793349168645,
|
| 311 |
+
"grad_norm": 5.648860931396484,
|
| 312 |
+
"learning_rate": 3.7688044338875695e-06,
|
| 313 |
+
"loss": 0.1742,
|
| 314 |
+
"step": 2050
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"epoch": 2.494061757719715,
|
| 318 |
+
"grad_norm": 6.062436103820801,
|
| 319 |
+
"learning_rate": 3.3729216152019006e-06,
|
| 320 |
+
"loss": 0.1464,
|
| 321 |
+
"step": 2100
|
| 322 |
+
},
|
| 323 |
+
{
|
| 324 |
+
"epoch": 2.553444180522565,
|
| 325 |
+
"grad_norm": 4.422060966491699,
|
| 326 |
+
"learning_rate": 2.977038796516231e-06,
|
| 327 |
+
"loss": 0.1714,
|
| 328 |
+
"step": 2150
|
| 329 |
+
},
|
| 330 |
+
{
|
| 331 |
+
"epoch": 2.6128266033254155,
|
| 332 |
+
"grad_norm": 4.258304119110107,
|
| 333 |
+
"learning_rate": 2.581155977830562e-06,
|
| 334 |
+
"loss": 0.1731,
|
| 335 |
+
"step": 2200
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 2.6722090261282663,
|
| 339 |
+
"grad_norm": 5.822784900665283,
|
| 340 |
+
"learning_rate": 2.1852731591448932e-06,
|
| 341 |
+
"loss": 0.1735,
|
| 342 |
+
"step": 2250
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 2.731591448931116,
|
| 346 |
+
"grad_norm": 6.498347282409668,
|
| 347 |
+
"learning_rate": 1.789390340459224e-06,
|
| 348 |
+
"loss": 0.1625,
|
| 349 |
+
"step": 2300
|
| 350 |
+
},
|
| 351 |
+
{
|
| 352 |
+
"epoch": 2.790973871733967,
|
| 353 |
+
"grad_norm": 7.230003356933594,
|
| 354 |
+
"learning_rate": 1.3935075217735553e-06,
|
| 355 |
+
"loss": 0.1724,
|
| 356 |
+
"step": 2350
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 2.850356294536817,
|
| 360 |
+
"grad_norm": 5.901012897491455,
|
| 361 |
+
"learning_rate": 9.976247030878861e-07,
|
| 362 |
+
"loss": 0.1631,
|
| 363 |
+
"step": 2400
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"epoch": 2.9097387173396676,
|
| 367 |
+
"grad_norm": 6.808001518249512,
|
| 368 |
+
"learning_rate": 6.01741884402217e-07,
|
| 369 |
+
"loss": 0.1557,
|
| 370 |
+
"step": 2450
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"epoch": 2.969121140142518,
|
| 374 |
+
"grad_norm": 6.5908684730529785,
|
| 375 |
+
"learning_rate": 2.0585906571654793e-07,
|
| 376 |
+
"loss": 0.1598,
|
| 377 |
+
"step": 2500
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 3.0,
|
| 381 |
+
"eval_accuracy": 0.8704128440366973,
|
| 382 |
+
"eval_loss": 0.3955666124820709,
|
| 383 |
+
"eval_runtime": 2.4422,
|
| 384 |
+
"eval_samples_per_second": 357.049,
|
| 385 |
+
"eval_steps_per_second": 2.457,
|
| 386 |
+
"step": 2526
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 3.0,
|
| 390 |
+
"step": 2526,
|
| 391 |
+
"total_flos": 6934924671671304.0,
|
| 392 |
+
"train_loss": 0.2517735588106954,
|
| 393 |
+
"train_runtime": 273.9564,
|
| 394 |
+
"train_samples_per_second": 737.515,
|
| 395 |
+
"train_steps_per_second": 9.22
|
| 396 |
+
}
|
| 397 |
+
],
|
| 398 |
+
"logging_steps": 50,
|
| 399 |
+
"max_steps": 2526,
|
| 400 |
+
"num_input_tokens_seen": 0,
|
| 401 |
+
"num_train_epochs": 3,
|
| 402 |
+
"save_steps": 500,
|
| 403 |
+
"stateful_callbacks": {
|
| 404 |
+
"TrainerControl": {
|
| 405 |
+
"args": {
|
| 406 |
+
"should_epoch_stop": false,
|
| 407 |
+
"should_evaluate": false,
|
| 408 |
+
"should_log": false,
|
| 409 |
+
"should_save": false,
|
| 410 |
+
"should_training_stop": false
|
| 411 |
+
},
|
| 412 |
+
"attributes": {}
|
| 413 |
+
}
|
| 414 |
+
},
|
| 415 |
+
"total_flos": 6934924671671304.0,
|
| 416 |
+
"train_batch_size": 80,
|
| 417 |
+
"trial_name": null,
|
| 418 |
+
"trial_params": null
|
| 419 |
+
}
|