Upload folder using huggingface_hub
Browse files
checkpoints/ablation_stream_graph_only/meta.json
CHANGED
|
@@ -1 +1 @@
|
|
| 1 |
-
{"epochs_done": 9, "history": [{"epoch": 0, "total": 2.191428791024746, "articulatory": 0.3237969840948398, "phoneme": 2.1752389414952353, "semantic": 3.8951404904707885, "encoder_grad_norm": 3.1820800354847543, "phoneme_head_grad_norm": Infinity, "nan_inf_batch_pct": 0.0, "blank_prediction_pct": 74.56272137612947, "mean_decoded_phoneme_len": 18.13179413179413, "mean_target_phoneme_len": 26.89738656405323, "train_per": 0.5938280509540472, "learning_rate": 0.0009935547237482068, "artic_target_positive_rate": 0.12211586766135998, "artic_pred_positive_rate": 0.1491869210337217, "weight_articulatory": 0.05, "weight_phoneme": 1.0, "weight_semantic": 0.0, "conf_warmup_alpha": 0.0, "gate_mean": 0.5111039886303355, "gate_min": 0.5040888786315918, "gate_max": 0.6839777827262878, "reliability_mean": 0.5895826248131024, "reliability_min": 9.999999747378752e-05, "reliability_max": 0.9999996423721313, "drift_severity_mean": 0.044740865274329626, "reference_initialized": true, "ctc_infeasible_pct_this_epoch": 0.0, "ctc_infeasible_pct_cumulative": 0.0, "val_per": 0.3007059205366078, "val_heuristic_wer": 0.9362265454126856, "val_heuristic_cer": 0.5118447340880982, "val_lexicon_wer": 0.5108783239323127, "val_lexicon_cer": 0.368257064028161, "val_wer": 0.5108783239323127, "val_cer": 0.368257064028161}, {"epoch": 1, "total": 1.5110528453802452, "articulatory": 0.2963853813898869, "phoneme": 1.1374421022641352, "semantic": 3.4397219303326727, "encoder_grad_norm": 2.0389815860833878, "phoneme_head_grad_norm": 1.612283854606824, "nan_inf_batch_pct": 0.0, "blank_prediction_pct": 65.87231737864735, "mean_decoded_phoneme_len": 25.31136764470098, "mean_target_phoneme_len": 26.89738656405323, "train_per": 0.34379675606077764, "learning_rate": 0.000942845842533964, "artic_target_positive_rate": 0.12211523922399069, "artic_pred_positive_rate": 0.12272187552581995, "weight_articulatory": 0.1, "weight_phoneme": 1.0, "weight_semantic": 0.09999999999999999, "conf_warmup_alpha": 0.3333333333333333, "gate_mean": 0.5116850419630888, "gate_min": 0.5044357180595398, "gate_max": 0.6760384440422058, "reliability_mean": 0.5829273608129568, "reliability_min": 9.999999747378752e-05, "reliability_max": 0.9999998807907104, "drift_severity_mean": 0.045853610191280994, "reference_initialized": true, "ctc_infeasible_pct_this_epoch": 0.0, "ctc_infeasible_pct_cumulative": 0.0, "val_per": 0.22586752495063966, "val_heuristic_wer": 0.9204558535743065, "val_heuristic_cer": 0.47424127104937686, "val_lexicon_wer": 0.38954759986186255, "val_lexicon_cer": 0.2811340500428123, "val_wer": 0.38954759986186255, "val_cer": 0.2811340500428123}, {"epoch": 2, "total": 1.5001678767876747, "articulatory": 0.29518899210752586, "phoneme": 0.8825265562687165, "semantic": 2.8668147881825763, "encoder_grad_norm": 2.4393460431160072, "phoneme_head_grad_norm": 1.5208116832451943, "nan_inf_batch_pct": 0.0, "blank_prediction_pct": 65.10615250987775, "mean_decoded_phoneme_len": 25.763187429854096, "mean_target_phoneme_len": 26.89738656405323, "train_per": 0.27794633968967386, "learning_rate": 0.0008468568807348963, "artic_target_positive_rate": 0.12211310711617653, "artic_pred_positive_rate": 0.12267070651436464, "weight_articulatory": 0.15, "weight_phoneme": 1.0, "weight_semantic": 0.19999999999999998, "conf_warmup_alpha": 0.6666666666666666, "gate_mean": 0.5121073152500218, "gate_min": 0.5045374035835266, "gate_max": 0.676649272441864, "reliability_mean": 0.5757737030523489, "reliability_min": 9.999999747378752e-05, "reliability_max": 0.9999998807907104, "drift_severity_mean": 0.04705526990702811, "reference_initialized": true, "ctc_infeasible_pct_this_epoch": 0.0, "ctc_infeasible_pct_cumulative": 0.0, "val_per": 0.1906255916479593, "val_heuristic_wer": 0.9109013468401059, "val_heuristic_cer": 0.456450385310627, "val_lexicon_wer": 0.3317601013007943, "val_lexicon_cer": 0.2409380648844068, "val_wer": 0.3317601013007943, "val_cer": 0.2409380648844068}, {"epoch": 3, "total": 1.544265055274352, "articulatory": 0.29447503044055057, "phoneme": 0.7268811824612128, "semantic": 2.5282961182105237, "encoder_grad_norm": 3.312517048151065, "phoneme_head_grad_norm": 1.4530611420289063, "nan_inf_batch_pct": 0.0, "blank_prediction_pct": 64.46674550621762, "mean_decoded_phoneme_len": 26.040243706910374, "mean_target_phoneme_len": 26.89738656405323, "train_per": 0.23517665222134132, "learning_rate": 0.0007159897317841356, "artic_target_positive_rate": 0.12211517855907097, "artic_pred_positive_rate": 0.12272970664959687, "weight_articulatory": 0.2, "weight_phoneme": 1.0, "weight_semantic": 0.3, "conf_warmup_alpha": 1.0, "gate_mean": 0.5125308890401581, "gate_min": 0.504912793636322, "gate_max": 0.6883471608161926, "reliability_mean": 0.5702482147134228, "reliability_min": 9.999999747378752e-05, "reliability_max": 0.9999992847442627, "drift_severity_mean": 0.04805035057722691, "reference_initialized": true, "ctc_infeasible_pct_this_epoch": 0.0, "ctc_infeasible_pct_cumulative": 0.0, "val_per": 0.17047575257620426, "val_heuristic_wer": 0.9052607344307586, "val_heuristic_cer": 0.4461516506516982, "val_lexicon_wer": 0.3012547484747324, "val_lexicon_cer": 0.2184616116449434, "val_wer": 0.3012547484747324, "val_cer": 0.2184616116449434}, {"epoch": 4, "total": 1.3305085860765897, "articulatory": 0.29369021245302296, "phoneme": 0.5981385341630532, "semantic": 2.245439947568453, "encoder_grad_norm": 3.3150538930526148, "phoneme_head_grad_norm": 1.3470729815654265, "nan_inf_batch_pct": 0.0, "blank_prediction_pct": 64.07249973712148, "mean_decoded_phoneme_len": 26.220458553791886, "mean_target_phoneme_len": 26.89738656405323, "train_per": 0.19898783373768322, "learning_rate": 0.0005644258818835341, "artic_target_positive_rate": 0.12211650454272062, "artic_pred_positive_rate": 0.12260943908148851, "weight_articulatory": 0.2, "weight_phoneme": 1.0, "weight_semantic": 0.3, "conf_warmup_alpha": 1.0, "gate_mean": 0.5127225981818305, "gate_min": 0.505071759223938, "gate_max": 0.6942287683486938, "reliability_mean": 0.5661500843935562, "reliability_min": 9.999999747378752e-05, "reliability_max": 0.9999996423721313, "drift_severity_mean": 0.04882526047932752, "reference_initialized": true, "ctc_infeasible_pct_this_epoch": 0.0, "ctc_infeasible_pct_cumulative": 0.0, "val_per": 0.15605982744164662, "val_heuristic_wer": 0.901577069183838, "val_heuristic_cer": 0.43865949957187705, "val_lexicon_wer": 0.276505122596984, "val_lexicon_cer": 0.20031395680715441, "val_wer": 0.276505122596984, "val_cer": 0.20031395680715441}, {"epoch": 5, "total": 1.1510480417655065, "articulatory": 0.2928615188369384, "phoneme": 0.49096781152945296, "semantic": 2.0050263490432347, "encoder_grad_norm": Infinity, "phoneme_head_grad_norm": 1.236970381400524, "nan_inf_batch_pct": 0.0, "blank_prediction_pct": 63.62711869289594, "mean_decoded_phoneme_len": 26.368286034952703, "mean_target_phoneme_len": 26.89738656405323, "train_per": 0.16764525301176092, "learning_rate": 0.00040858962592526263, "artic_target_positive_rate": 0.12211424843049966, "artic_pred_positive_rate": 0.12267035835255415, "weight_articulatory": 0.2, "weight_phoneme": 1.0, "weight_semantic": 0.3, "conf_warmup_alpha": 1.0, "gate_mean": 0.5130395900750172, "gate_min": 0.5050415396690369, "gate_max": 0.6816018223762512, "reliability_mean": 0.5633640362319224, "reliability_min": 9.999999747378752e-05, "reliability_max": 0.9999991059303284, "drift_severity_mean": 0.04937787890835643, "reference_initialized": true, "ctc_infeasible_pct_this_epoch": 0.0, "ctc_infeasible_pct_cumulative": 0.0, "val_per": 0.14689097449490168, "val_heuristic_wer": 0.89904454932658, "val_heuristic_cer": 0.4352345162211017, "val_lexicon_wer": 0.2632669506158628, "val_lexicon_cer": 0.18723242317572067, "val_wer": 0.2632669506158628, "val_cer": 0.18723242317572067}, {"epoch": 6, "total": 0.9985782916729267, "articulatory": 0.29202397137116165, "phoneme": 0.39978370250035555, "semantic": 1.8012992427899288, "encoder_grad_norm": 3.321632634523587, "phoneme_head_grad_norm": 1.1109671927415408, "nan_inf_batch_pct": 0.0, "blank_prediction_pct": 63.28760927464252, "mean_decoded_phoneme_len": 26.493987493987493, "mean_target_phoneme_len": 26.89738656405323, "train_per": 0.1399746064294613, "learning_rate": 0.00026536824029105174, "artic_target_positive_rate": 0.12211624505237127, "artic_pred_positive_rate": 0.12254943211491291, "weight_articulatory": 0.2, "weight_phoneme": 1.0, "weight_semantic": 0.3, "conf_warmup_alpha": 1.0, "gate_mean": 0.5130120666244578, "gate_min": 0.5051751136779785, "gate_max": 0.6850141286849976, "reliability_mean": 0.5614773051677959, "reliability_min": 9.999999747378752e-05, "reliability_max": 0.9999997615814209, "drift_severity_mean": 0.049691303319726134, "reference_initialized": true, "ctc_infeasible_pct_this_epoch": 0.0, "ctc_infeasible_pct_cumulative": 0.0, "val_per": 0.14269872609742246, "val_heuristic_wer": 0.8969724876251871, "val_heuristic_cer": 0.4324279326419941, "val_lexicon_wer": 0.2560147346609877, "val_lexicon_cer": 0.1827609171344306, "val_wer": 0.2560147346609877, "val_cer": 0.1827609171344306}, {"epoch": 7, "total": 0.8778667448422848, "articulatory": 0.29142207339023934, "phoneme": 0.32959854457622917, "semantic": 1.6332792113988828, "encoder_grad_norm": Infinity, "phoneme_head_grad_norm": 0.9909259447684655, "nan_inf_batch_pct": 0.0, "blank_prediction_pct": 63.37864446867568, "mean_decoded_phoneme_len": 26.56854256854257, "mean_target_phoneme_len": 26.89738656405323, "train_per": 0.11808010300490585, "learning_rate": 0.00015028198442313846, "artic_target_positive_rate": 0.12211496022840342, "artic_pred_positive_rate": 0.12251577814802145, "weight_articulatory": 0.2, "weight_phoneme": 1.0, "weight_semantic": 0.3, "conf_warmup_alpha": 1.0, "gate_mean": 0.5130539723308074, "gate_min": 0.5049024224281311, "gate_max": 0.685222864151001, "reliability_mean": 0.5606571273656922, "reliability_min": 9.999999747378752e-05, "reliability_max": 1.0, "drift_severity_mean": 0.04986832595385667, "reference_initialized": true, "ctc_infeasible_pct_this_epoch": 0.0, "ctc_infeasible_pct_cumulative": 0.0, "val_per": 0.13858761799150732, "val_heuristic_wer": 0.8959364567744906, "val_heuristic_cer": 0.4307630101798116, "val_lexicon_wer": 0.24876251870611257, "val_lexicon_cer": 0.17657691941775283, "val_wer": 0.24876251870611257, "val_cer": 0.17657691941775283}, {"epoch": 8, "total": 0.796430523158648, "articulatory": 0.29111607395685635, "phoneme": 0.284761623140329, "semantic": 1.5114855631803854, "encoder_grad_norm": 3.0681250845774626, "phoneme_head_grad_norm": 0.9236487403512001, "nan_inf_batch_pct": 0.0, "blank_prediction_pct": 62.95299731935924, "mean_decoded_phoneme_len": 26.61792528459195, "mean_target_phoneme_len": 26.89738656405323, "train_per": 0.10338640549836373, "learning_rate": 7.580223989206306e-05, "artic_target_positive_rate": 0.12211636625803433, "artic_pred_positive_rate": 0.12267417807418567, "weight_articulatory": 0.2, "weight_phoneme": 1.0, "weight_semantic": 0.3, "conf_warmup_alpha": 1.0, "gate_mean": 0.5130912013709096, "gate_min": 0.5054836869239807, "gate_max": 0.6825555562973022, "reliability_mean": 0.5601125739793165, "reliability_min": 9.999999747378752e-05, "reliability_max": 0.9999997615814209, "drift_severity_mean": 0.0500266769910176, "reference_initialized": true, "ctc_infeasible_pct_this_epoch": 0.0, "ctc_infeasible_pct_cumulative": 0.0, "val_per": 0.13645092364698563, "val_heuristic_wer": 0.895245769540693, "val_heuristic_cer": 0.4297640567025021, "val_lexicon_wer": 0.24277656267986647, "val_lexicon_cer": 0.1726286747217201, "val_wer": 0.24277656267986647, "val_cer": 0.1726286747217201}], "cfg": {"d_model": 256, "n_heads": 8, "graph_dim": 32, "graph_heads": 4, "n_graph_layers": 2, "trunk_layers": 2, "trunk_kind": "gru", "patch_bins": 8, "semantic_embed_dim": 256, "baseline_d_model": 256, "graph_patch_chunk_size": 16, "dropout": 0.1, "sinkhorn_iters": 50, "batch_size": 8, "accumulation_steps": 1, "epochs": 30, "ablation_epochs": 10, "lr": 0.001, "weight_decay": 0.01, "warmup_frac": 0.05, "min_lr_frac": 0.05, "aug_noise_std": 0.3, "aug_offset_std": 0.1, "aug_channel_drop": 0.08, "aug_time_mask_n": 2, "aug_time_mask_max": 10, "ema_decay": 0.999, "use_session_affine": true, "collapse_patience_epochs": 4, "log_every": 100, "cl_lr": 0.0001, "cl_base_epochs": 3, "cl_task_epochs": 3, "cl_adapt_epochs": 8, "cl_lora_rank": 8, "cl_ot_shrink": 0.05, "cl_replay_ratio": 0.5, "cl_ewc_lambda": 100.0}, "pipeline_version": "chgft_v6_strict_cl_ema_affine_ot"}
|
|
|
|
| 1 |
+
{"epochs_done": 10, "history": [{"epoch": 0, "total": 2.191428791024746, "articulatory": 0.3237969840948398, "phoneme": 2.1752389414952353, "semantic": 3.8951404904707885, "encoder_grad_norm": 3.1820800354847543, "phoneme_head_grad_norm": Infinity, "nan_inf_batch_pct": 0.0, "blank_prediction_pct": 74.56272137612947, "mean_decoded_phoneme_len": 18.13179413179413, "mean_target_phoneme_len": 26.89738656405323, "train_per": 0.5938280509540472, "learning_rate": 0.0009935547237482068, "artic_target_positive_rate": 0.12211586766135998, "artic_pred_positive_rate": 0.1491869210337217, "weight_articulatory": 0.05, "weight_phoneme": 1.0, "weight_semantic": 0.0, "conf_warmup_alpha": 0.0, "gate_mean": 0.5111039886303355, "gate_min": 0.5040888786315918, "gate_max": 0.6839777827262878, "reliability_mean": 0.5895826248131024, "reliability_min": 9.999999747378752e-05, "reliability_max": 0.9999996423721313, "drift_severity_mean": 0.044740865274329626, "reference_initialized": true, "ctc_infeasible_pct_this_epoch": 0.0, "ctc_infeasible_pct_cumulative": 0.0, "val_per": 0.3007059205366078, "val_heuristic_wer": 0.9362265454126856, "val_heuristic_cer": 0.5118447340880982, "val_lexicon_wer": 0.5108783239323127, "val_lexicon_cer": 0.368257064028161, "val_wer": 0.5108783239323127, "val_cer": 0.368257064028161}, {"epoch": 1, "total": 1.5110528453802452, "articulatory": 0.2963853813898869, "phoneme": 1.1374421022641352, "semantic": 3.4397219303326727, "encoder_grad_norm": 2.0389815860833878, "phoneme_head_grad_norm": 1.612283854606824, "nan_inf_batch_pct": 0.0, "blank_prediction_pct": 65.87231737864735, "mean_decoded_phoneme_len": 25.31136764470098, "mean_target_phoneme_len": 26.89738656405323, "train_per": 0.34379675606077764, "learning_rate": 0.000942845842533964, "artic_target_positive_rate": 0.12211523922399069, "artic_pred_positive_rate": 0.12272187552581995, "weight_articulatory": 0.1, "weight_phoneme": 1.0, "weight_semantic": 0.09999999999999999, "conf_warmup_alpha": 0.3333333333333333, "gate_mean": 0.5116850419630888, "gate_min": 0.5044357180595398, "gate_max": 0.6760384440422058, "reliability_mean": 0.5829273608129568, "reliability_min": 9.999999747378752e-05, "reliability_max": 0.9999998807907104, "drift_severity_mean": 0.045853610191280994, "reference_initialized": true, "ctc_infeasible_pct_this_epoch": 0.0, "ctc_infeasible_pct_cumulative": 0.0, "val_per": 0.22586752495063966, "val_heuristic_wer": 0.9204558535743065, "val_heuristic_cer": 0.47424127104937686, "val_lexicon_wer": 0.38954759986186255, "val_lexicon_cer": 0.2811340500428123, "val_wer": 0.38954759986186255, "val_cer": 0.2811340500428123}, {"epoch": 2, "total": 1.5001678767876747, "articulatory": 0.29518899210752586, "phoneme": 0.8825265562687165, "semantic": 2.8668147881825763, "encoder_grad_norm": 2.4393460431160072, "phoneme_head_grad_norm": 1.5208116832451943, "nan_inf_batch_pct": 0.0, "blank_prediction_pct": 65.10615250987775, "mean_decoded_phoneme_len": 25.763187429854096, "mean_target_phoneme_len": 26.89738656405323, "train_per": 0.27794633968967386, "learning_rate": 0.0008468568807348963, "artic_target_positive_rate": 0.12211310711617653, "artic_pred_positive_rate": 0.12267070651436464, "weight_articulatory": 0.15, "weight_phoneme": 1.0, "weight_semantic": 0.19999999999999998, "conf_warmup_alpha": 0.6666666666666666, "gate_mean": 0.5121073152500218, "gate_min": 0.5045374035835266, "gate_max": 0.676649272441864, "reliability_mean": 0.5757737030523489, "reliability_min": 9.999999747378752e-05, "reliability_max": 0.9999998807907104, "drift_severity_mean": 0.04705526990702811, "reference_initialized": true, "ctc_infeasible_pct_this_epoch": 0.0, "ctc_infeasible_pct_cumulative": 0.0, "val_per": 0.1906255916479593, "val_heuristic_wer": 0.9109013468401059, "val_heuristic_cer": 0.456450385310627, "val_lexicon_wer": 0.3317601013007943, "val_lexicon_cer": 0.2409380648844068, "val_wer": 0.3317601013007943, "val_cer": 0.2409380648844068}, {"epoch": 3, "total": 1.544265055274352, "articulatory": 0.29447503044055057, "phoneme": 0.7268811824612128, "semantic": 2.5282961182105237, "encoder_grad_norm": 3.312517048151065, "phoneme_head_grad_norm": 1.4530611420289063, "nan_inf_batch_pct": 0.0, "blank_prediction_pct": 64.46674550621762, "mean_decoded_phoneme_len": 26.040243706910374, "mean_target_phoneme_len": 26.89738656405323, "train_per": 0.23517665222134132, "learning_rate": 0.0007159897317841356, "artic_target_positive_rate": 0.12211517855907097, "artic_pred_positive_rate": 0.12272970664959687, "weight_articulatory": 0.2, "weight_phoneme": 1.0, "weight_semantic": 0.3, "conf_warmup_alpha": 1.0, "gate_mean": 0.5125308890401581, "gate_min": 0.504912793636322, "gate_max": 0.6883471608161926, "reliability_mean": 0.5702482147134228, "reliability_min": 9.999999747378752e-05, "reliability_max": 0.9999992847442627, "drift_severity_mean": 0.04805035057722691, "reference_initialized": true, "ctc_infeasible_pct_this_epoch": 0.0, "ctc_infeasible_pct_cumulative": 0.0, "val_per": 0.17047575257620426, "val_heuristic_wer": 0.9052607344307586, "val_heuristic_cer": 0.4461516506516982, "val_lexicon_wer": 0.3012547484747324, "val_lexicon_cer": 0.2184616116449434, "val_wer": 0.3012547484747324, "val_cer": 0.2184616116449434}, {"epoch": 4, "total": 1.3305085860765897, "articulatory": 0.29369021245302296, "phoneme": 0.5981385341630532, "semantic": 2.245439947568453, "encoder_grad_norm": 3.3150538930526148, "phoneme_head_grad_norm": 1.3470729815654265, "nan_inf_batch_pct": 0.0, "blank_prediction_pct": 64.07249973712148, "mean_decoded_phoneme_len": 26.220458553791886, "mean_target_phoneme_len": 26.89738656405323, "train_per": 0.19898783373768322, "learning_rate": 0.0005644258818835341, "artic_target_positive_rate": 0.12211650454272062, "artic_pred_positive_rate": 0.12260943908148851, "weight_articulatory": 0.2, "weight_phoneme": 1.0, "weight_semantic": 0.3, "conf_warmup_alpha": 1.0, "gate_mean": 0.5127225981818305, "gate_min": 0.505071759223938, "gate_max": 0.6942287683486938, "reliability_mean": 0.5661500843935562, "reliability_min": 9.999999747378752e-05, "reliability_max": 0.9999996423721313, "drift_severity_mean": 0.04882526047932752, "reference_initialized": true, "ctc_infeasible_pct_this_epoch": 0.0, "ctc_infeasible_pct_cumulative": 0.0, "val_per": 0.15605982744164662, "val_heuristic_wer": 0.901577069183838, "val_heuristic_cer": 0.43865949957187705, "val_lexicon_wer": 0.276505122596984, "val_lexicon_cer": 0.20031395680715441, "val_wer": 0.276505122596984, "val_cer": 0.20031395680715441}, {"epoch": 5, "total": 1.1510480417655065, "articulatory": 0.2928615188369384, "phoneme": 0.49096781152945296, "semantic": 2.0050263490432347, "encoder_grad_norm": Infinity, "phoneme_head_grad_norm": 1.236970381400524, "nan_inf_batch_pct": 0.0, "blank_prediction_pct": 63.62711869289594, "mean_decoded_phoneme_len": 26.368286034952703, "mean_target_phoneme_len": 26.89738656405323, "train_per": 0.16764525301176092, "learning_rate": 0.00040858962592526263, "artic_target_positive_rate": 0.12211424843049966, "artic_pred_positive_rate": 0.12267035835255415, "weight_articulatory": 0.2, "weight_phoneme": 1.0, "weight_semantic": 0.3, "conf_warmup_alpha": 1.0, "gate_mean": 0.5130395900750172, "gate_min": 0.5050415396690369, "gate_max": 0.6816018223762512, "reliability_mean": 0.5633640362319224, "reliability_min": 9.999999747378752e-05, "reliability_max": 0.9999991059303284, "drift_severity_mean": 0.04937787890835643, "reference_initialized": true, "ctc_infeasible_pct_this_epoch": 0.0, "ctc_infeasible_pct_cumulative": 0.0, "val_per": 0.14689097449490168, "val_heuristic_wer": 0.89904454932658, "val_heuristic_cer": 0.4352345162211017, "val_lexicon_wer": 0.2632669506158628, "val_lexicon_cer": 0.18723242317572067, "val_wer": 0.2632669506158628, "val_cer": 0.18723242317572067}, {"epoch": 6, "total": 0.9985782916729267, "articulatory": 0.29202397137116165, "phoneme": 0.39978370250035555, "semantic": 1.8012992427899288, "encoder_grad_norm": 3.321632634523587, "phoneme_head_grad_norm": 1.1109671927415408, "nan_inf_batch_pct": 0.0, "blank_prediction_pct": 63.28760927464252, "mean_decoded_phoneme_len": 26.493987493987493, "mean_target_phoneme_len": 26.89738656405323, "train_per": 0.1399746064294613, "learning_rate": 0.00026536824029105174, "artic_target_positive_rate": 0.12211624505237127, "artic_pred_positive_rate": 0.12254943211491291, "weight_articulatory": 0.2, "weight_phoneme": 1.0, "weight_semantic": 0.3, "conf_warmup_alpha": 1.0, "gate_mean": 0.5130120666244578, "gate_min": 0.5051751136779785, "gate_max": 0.6850141286849976, "reliability_mean": 0.5614773051677959, "reliability_min": 9.999999747378752e-05, "reliability_max": 0.9999997615814209, "drift_severity_mean": 0.049691303319726134, "reference_initialized": true, "ctc_infeasible_pct_this_epoch": 0.0, "ctc_infeasible_pct_cumulative": 0.0, "val_per": 0.14269872609742246, "val_heuristic_wer": 0.8969724876251871, "val_heuristic_cer": 0.4324279326419941, "val_lexicon_wer": 0.2560147346609877, "val_lexicon_cer": 0.1827609171344306, "val_wer": 0.2560147346609877, "val_cer": 0.1827609171344306}, {"epoch": 7, "total": 0.8778667448422848, "articulatory": 0.29142207339023934, "phoneme": 0.32959854457622917, "semantic": 1.6332792113988828, "encoder_grad_norm": Infinity, "phoneme_head_grad_norm": 0.9909259447684655, "nan_inf_batch_pct": 0.0, "blank_prediction_pct": 63.37864446867568, "mean_decoded_phoneme_len": 26.56854256854257, "mean_target_phoneme_len": 26.89738656405323, "train_per": 0.11808010300490585, "learning_rate": 0.00015028198442313846, "artic_target_positive_rate": 0.12211496022840342, "artic_pred_positive_rate": 0.12251577814802145, "weight_articulatory": 0.2, "weight_phoneme": 1.0, "weight_semantic": 0.3, "conf_warmup_alpha": 1.0, "gate_mean": 0.5130539723308074, "gate_min": 0.5049024224281311, "gate_max": 0.685222864151001, "reliability_mean": 0.5606571273656922, "reliability_min": 9.999999747378752e-05, "reliability_max": 1.0, "drift_severity_mean": 0.04986832595385667, "reference_initialized": true, "ctc_infeasible_pct_this_epoch": 0.0, "ctc_infeasible_pct_cumulative": 0.0, "val_per": 0.13858761799150732, "val_heuristic_wer": 0.8959364567744906, "val_heuristic_cer": 0.4307630101798116, "val_lexicon_wer": 0.24876251870611257, "val_lexicon_cer": 0.17657691941775283, "val_wer": 0.24876251870611257, "val_cer": 0.17657691941775283}, {"epoch": 8, "total": 0.796430523158648, "articulatory": 0.29111607395685635, "phoneme": 0.284761623140329, "semantic": 1.5114855631803854, "encoder_grad_norm": 3.0681250845774626, "phoneme_head_grad_norm": 0.9236487403512001, "nan_inf_batch_pct": 0.0, "blank_prediction_pct": 62.95299731935924, "mean_decoded_phoneme_len": 26.61792528459195, "mean_target_phoneme_len": 26.89738656405323, "train_per": 0.10338640549836373, "learning_rate": 7.580223989206306e-05, "artic_target_positive_rate": 0.12211636625803433, "artic_pred_positive_rate": 0.12267417807418567, "weight_articulatory": 0.2, "weight_phoneme": 1.0, "weight_semantic": 0.3, "conf_warmup_alpha": 1.0, "gate_mean": 0.5130912013709096, "gate_min": 0.5054836869239807, "gate_max": 0.6825555562973022, "reliability_mean": 0.5601125739793165, "reliability_min": 9.999999747378752e-05, "reliability_max": 0.9999997615814209, "drift_severity_mean": 0.0500266769910176, "reference_initialized": true, "ctc_infeasible_pct_this_epoch": 0.0, "ctc_infeasible_pct_cumulative": 0.0, "val_per": 0.13645092364698563, "val_heuristic_wer": 0.895245769540693, "val_heuristic_cer": 0.4297640567025021, "val_lexicon_wer": 0.24277656267986647, "val_lexicon_cer": 0.1726286747217201, "val_wer": 0.24277656267986647, "val_cer": 0.1726286747217201}, {"epoch": 9, "total": 0.7461338589588801, "articulatory": 0.2908713075595024, "phoneme": 0.2556014254832497, "semantic": 1.4411938537389803, "encoder_grad_norm": 2.9719681767316963, "phoneme_head_grad_norm": 0.8670198319432063, "nan_inf_batch_pct": 0.0, "blank_prediction_pct": 62.7859074062604, "mean_decoded_phoneme_len": 26.659611992945326, "mean_target_phoneme_len": 26.89738656405323, "train_per": 0.09402178124571559, "learning_rate": 5.0000042690076865e-05, "artic_target_positive_rate": 0.1221164602213181, "artic_pred_positive_rate": 0.12247573817387605, "weight_articulatory": 0.2, "weight_phoneme": 1.0, "weight_semantic": 0.3, "conf_warmup_alpha": 1.0, "gate_mean": 0.5131067244196542, "gate_min": 0.5053651332855225, "gate_max": 0.6937165856361389, "reliability_mean": 0.5594649456700626, "reliability_min": 9.999999747378752e-05, "reliability_max": 0.9999988079071045, "drift_severity_mean": 0.05010952550248303, "reference_initialized": true, "ctc_infeasible_pct_this_epoch": 0.0, "ctc_infeasible_pct_cumulative": 0.0, "val_per": 0.13534200632894275, "val_heuristic_wer": 0.8946701968458617, "val_heuristic_cer": 0.42878888783179525, "val_lexicon_wer": 0.2402440428226085, "val_lexicon_cer": 0.17203405955665493, "val_wer": 0.2402440428226085, "val_cer": 0.17203405955665493}], "cfg": {"d_model": 256, "n_heads": 8, "graph_dim": 32, "graph_heads": 4, "n_graph_layers": 2, "trunk_layers": 2, "trunk_kind": "gru", "patch_bins": 8, "semantic_embed_dim": 256, "baseline_d_model": 256, "graph_patch_chunk_size": 16, "dropout": 0.1, "sinkhorn_iters": 50, "batch_size": 8, "accumulation_steps": 1, "epochs": 30, "ablation_epochs": 10, "lr": 0.001, "weight_decay": 0.01, "warmup_frac": 0.05, "min_lr_frac": 0.05, "aug_noise_std": 0.3, "aug_offset_std": 0.1, "aug_channel_drop": 0.08, "aug_time_mask_n": 2, "aug_time_mask_max": 10, "ema_decay": 0.999, "use_session_affine": true, "collapse_patience_epochs": 4, "log_every": 100, "cl_lr": 0.0001, "cl_base_epochs": 3, "cl_task_epochs": 3, "cl_adapt_epochs": 8, "cl_lora_rank": 8, "cl_ot_shrink": 0.05, "cl_replay_ratio": 0.5, "cl_ewc_lambda": 100.0}, "pipeline_version": "chgft_v6_strict_cl_ema_affine_ot"}
|
checkpoints/ablation_stream_graph_only/model.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 36956519
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8251e9787dd74af3d1b2abf9c44cd7e4e017910188e21b0ed5a7264222c1a644
|
| 3 |
size 36956519
|
checkpoints/ablation_stream_graph_only/optimizer.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 36449995
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:170067172c47f50e62ad4646cc40e708e86f5b5999f9a1f5ce069e125197ff23
|
| 3 |
size 36449995
|
checkpoints/ablation_stream_graph_only/scaler.pt
CHANGED
|
Binary files a/checkpoints/ablation_stream_graph_only/scaler.pt and b/checkpoints/ablation_stream_graph_only/scaler.pt differ
|
|
|