mirror of
https://github.com/RYDE-WORK/lnp_ml.git
synced 2026-07-24 23:11:52 +08:00
- Updated ablation_summary.csv with full 6-variant comparison - Updated run_all.sh to run 6 variants (added BioT5 variants) - Added per-fold json artifacts (best_params/test_metrics/splits/history/summary) - Model weights (.pt) excluded per .gitignore
992 lines
28 KiB
JSON
992 lines
28 KiB
JSON
{
|
|
"fold_results": [
|
|
{
|
|
"fold": 0,
|
|
"best_params": {
|
|
"dropout": 0.38118503457897024,
|
|
"lr": 0.0009770749364761302,
|
|
"weight_decay": 0.0008416934971427061,
|
|
"backbone_lr_ratio": 0.048484014956722714,
|
|
"moe_n_experts": 2,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 1,
|
|
"llm_lora_r": 16,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 15,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 83,
|
|
"mse": 1.2775270638012577,
|
|
"rmse": 1.1302774278031291,
|
|
"mae": 0.5309888495648482,
|
|
"r2": 0.21812199074999694
|
|
},
|
|
"delivery": {
|
|
"n_samples": 58,
|
|
"mse": 0.5813553680199423,
|
|
"rmse": 0.7624666340371506,
|
|
"mae": 0.5806423342176552,
|
|
"r2": 0.26106224598641015
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7142857142857143,
|
|
"precision": 0.6664637857577602,
|
|
"recall": 0.6949394155381325,
|
|
"f1": 0.672514619883041
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6666666666666666,
|
|
"precision": 0.6453567538126362,
|
|
"recall": 0.6947619047619048,
|
|
"f1": 0.6466727716727717
|
|
},
|
|
"toxic": {
|
|
"n_samples": 58,
|
|
"accuracy": 0.9655172413793104,
|
|
"precision": 0.75,
|
|
"recall": 0.9821428571428572,
|
|
"f1": 0.8242424242424242
|
|
},
|
|
"biodist": {
|
|
"n_samples": 58,
|
|
"kl_divergence": 0.30543536337189214,
|
|
"js_divergence": 0.055401807245651365
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 1,
|
|
"best_params": {
|
|
"dropout": 0.12715228141755225,
|
|
"lr": 0.00038210169217516986,
|
|
"weight_decay": 0.0032858044759097593,
|
|
"backbone_lr_ratio": 0.01707341286908778,
|
|
"moe_n_experts": 2,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 1,
|
|
"llm_lora_r": 16,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 18,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.3226448578157207,
|
|
"rmse": 0.5680183604565268,
|
|
"mae": 0.43127859149882125,
|
|
"r2": 0.014757311793952343
|
|
},
|
|
"delivery": {
|
|
"n_samples": 61,
|
|
"mse": 0.9731257240529179,
|
|
"rmse": 0.9864713498388678,
|
|
"mae": 0.6065680303717734,
|
|
"r2": 0.26341397848585624
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6666666666666666,
|
|
"precision": 0.5783050847457627,
|
|
"recall": 0.5873015873015873,
|
|
"f1": 0.5808980755523877
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.5238095238095238,
|
|
"precision": 0.4563973063973064,
|
|
"recall": 0.4560201703058846,
|
|
"f1": 0.4443340972752738
|
|
},
|
|
"toxic": {
|
|
"n_samples": 61,
|
|
"accuracy": 0.9508196721311475,
|
|
"precision": 0.75,
|
|
"recall": 0.9741379310344828,
|
|
"f1": 0.8200589970501475
|
|
},
|
|
"biodist": {
|
|
"n_samples": 61,
|
|
"kl_divergence": 0.3699852731241905,
|
|
"js_divergence": 0.0921670637504171
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 2,
|
|
"best_params": {
|
|
"dropout": 0.33774333325981976,
|
|
"lr": 0.0008937362022799393,
|
|
"weight_decay": 0.0797179350056365,
|
|
"backbone_lr_ratio": 0.012019950560462185,
|
|
"moe_n_experts": 2,
|
|
"moe_top_k": 1,
|
|
"moe_expert_hidden_mult": 2,
|
|
"llm_lora_r": 8,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 24,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.43468907454487105,
|
|
"rmse": 0.6593095437993227,
|
|
"mae": 0.5041968602287982,
|
|
"r2": 0.14149684434925025
|
|
},
|
|
"delivery": {
|
|
"n_samples": 60,
|
|
"mse": 0.6609372574877708,
|
|
"rmse": 0.8129804779253748,
|
|
"mae": 0.5105325942936664,
|
|
"r2": 0.12017795405995413
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7142857142857143,
|
|
"precision": 0.6592592592592592,
|
|
"recall": 0.6891495601173021,
|
|
"f1": 0.6657824933687002
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6785714285714286,
|
|
"precision": 0.6277327935222673,
|
|
"recall": 0.6944444444444443,
|
|
"f1": 0.6407632082237629
|
|
},
|
|
"toxic": {
|
|
"n_samples": 61,
|
|
"accuracy": 0.9344262295081968,
|
|
"precision": 0.7142857142857143,
|
|
"recall": 0.9655172413793103,
|
|
"f1": 0.7821428571428571
|
|
},
|
|
"biodist": {
|
|
"n_samples": 60,
|
|
"kl_divergence": 0.2188401979219755,
|
|
"js_divergence": 0.052433191262446874
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 3,
|
|
"best_params": {
|
|
"dropout": 0.10301838844131589,
|
|
"lr": 0.0008648998582653286,
|
|
"weight_decay": 0.0011644667702360517,
|
|
"backbone_lr_ratio": 0.9869864342312137,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 2,
|
|
"llm_lora_r": 32,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 15,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 83,
|
|
"mse": 1.550640342858784,
|
|
"rmse": 1.245247101124425,
|
|
"mae": 0.5501430786989001,
|
|
"r2": 0.12294600324587124
|
|
},
|
|
"delivery": {
|
|
"n_samples": 59,
|
|
"mse": 0.748405480052253,
|
|
"rmse": 0.8651043174393785,
|
|
"mae": 0.5687398803657142,
|
|
"r2": 0.2724609546561282
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7142857142857143,
|
|
"precision": 0.6445623342175066,
|
|
"recall": 0.6598240469208211,
|
|
"f1": 0.65
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6428571428571429,
|
|
"precision": 0.5892509050403788,
|
|
"recall": 0.6273504273504273,
|
|
"f1": 0.5933913253594297
|
|
},
|
|
"toxic": {
|
|
"n_samples": 60,
|
|
"accuracy": 0.9666666666666667,
|
|
"precision": 0.8,
|
|
"recall": 0.9824561403508771,
|
|
"f1": 0.8660714285714286
|
|
},
|
|
"biodist": {
|
|
"n_samples": 60,
|
|
"kl_divergence": 0.16217964699830698,
|
|
"js_divergence": 0.0385441230143625
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 4,
|
|
"best_params": {
|
|
"dropout": 0.303142745975858,
|
|
"lr": 0.0004649241012312578,
|
|
"weight_decay": 3.7624434569437365e-05,
|
|
"backbone_lr_ratio": 0.17310516991766065,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 1,
|
|
"llm_lora_r": 8,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 16,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.6548873107621247,
|
|
"rmse": 0.8092510801735916,
|
|
"mae": 0.47477718863436685,
|
|
"r2": 0.12367943077681431
|
|
},
|
|
"delivery": {
|
|
"n_samples": 58,
|
|
"mse": 0.8409562737928621,
|
|
"rmse": 0.9170366807237659,
|
|
"mae": 0.6096879987624185,
|
|
"r2": 0.18680916583420515
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7380952380952381,
|
|
"precision": 0.6724137931034483,
|
|
"recall": 0.6906158357771262,
|
|
"f1": 0.6791666666666667
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7142857142857143,
|
|
"precision": 0.6853395646499095,
|
|
"recall": 0.7205128205128205,
|
|
"f1": 0.6928875379939209
|
|
},
|
|
"toxic": {
|
|
"n_samples": 59,
|
|
"accuracy": 0.9661016949152542,
|
|
"precision": 0.8,
|
|
"recall": 0.9821428571428572,
|
|
"f1": 0.865909090909091
|
|
},
|
|
"biodist": {
|
|
"n_samples": 58,
|
|
"kl_divergence": 0.32346840560601176,
|
|
"js_divergence": 0.08737275181528921
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 0,
|
|
"best_params": {
|
|
"dropout": 0.38118503457897024,
|
|
"lr": 0.0009770749364761302,
|
|
"weight_decay": 0.0008416934971427061,
|
|
"backbone_lr_ratio": 0.048484014956722714,
|
|
"moe_n_experts": 2,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 1,
|
|
"llm_lora_r": 16,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 15,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 83,
|
|
"mse": 1.375775554685643,
|
|
"rmse": 1.1729345909664541,
|
|
"mae": 0.5610226254506283,
|
|
"r2": 0.15799149595176643
|
|
},
|
|
"delivery": {
|
|
"n_samples": 58,
|
|
"mse": 0.6578320017621242,
|
|
"rmse": 0.811068432231291,
|
|
"mae": 0.5792412295744851,
|
|
"r2": 0.16385582960043577
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7857142857142857,
|
|
"precision": 0.7305773342836779,
|
|
"recall": 0.7305773342836779,
|
|
"f1": 0.7305773342836779
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7142857142857143,
|
|
"precision": 0.6695636813283872,
|
|
"recall": 0.7142857142857144,
|
|
"f1": 0.6750809541465826
|
|
},
|
|
"toxic": {
|
|
"n_samples": 58,
|
|
"accuracy": 0.9655172413793104,
|
|
"precision": 0.75,
|
|
"recall": 0.9821428571428572,
|
|
"f1": 0.8242424242424242
|
|
},
|
|
"biodist": {
|
|
"n_samples": 58,
|
|
"kl_divergence": 0.2226351478005787,
|
|
"js_divergence": 0.04298447371604951
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 1,
|
|
"best_params": {
|
|
"dropout": 0.12715228141755225,
|
|
"lr": 0.00038210169217516986,
|
|
"weight_decay": 0.0032858044759097593,
|
|
"backbone_lr_ratio": 0.01707341286908778,
|
|
"moe_n_experts": 2,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 1,
|
|
"llm_lora_r": 16,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 18,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.34095375976281117,
|
|
"rmse": 0.5839124589891974,
|
|
"mae": 0.44237262724588317,
|
|
"r2": -0.04115156552265198
|
|
},
|
|
"delivery": {
|
|
"n_samples": 61,
|
|
"mse": 1.0561289903263111,
|
|
"rmse": 1.0276813661472661,
|
|
"mae": 0.6424658659112746,
|
|
"r2": 0.20058648953369573
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6785714285714286,
|
|
"precision": 0.6358823529411765,
|
|
"recall": 0.6746031746031746,
|
|
"f1": 0.635076427996782
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.5595238095238095,
|
|
"precision": 0.49252136752136755,
|
|
"recall": 0.49526663812378097,
|
|
"f1": 0.4855607355607356
|
|
},
|
|
"toxic": {
|
|
"n_samples": 61,
|
|
"accuracy": 0.9508196721311475,
|
|
"precision": 0.75,
|
|
"recall": 0.9741379310344828,
|
|
"f1": 0.8200589970501475
|
|
},
|
|
"biodist": {
|
|
"n_samples": 61,
|
|
"kl_divergence": 0.45483073434111365,
|
|
"js_divergence": 0.11671493720560924
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 2,
|
|
"best_params": {
|
|
"dropout": 0.33774333325981976,
|
|
"lr": 0.0008937362022799393,
|
|
"weight_decay": 0.0797179350056365,
|
|
"backbone_lr_ratio": 0.012019950560462185,
|
|
"moe_n_experts": 2,
|
|
"moe_top_k": 1,
|
|
"moe_expert_hidden_mult": 2,
|
|
"llm_lora_r": 8,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 24,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.5561159310353995,
|
|
"rmse": 0.7457318090543004,
|
|
"mae": 0.5436848102891374,
|
|
"r2": -0.09831902768989997
|
|
},
|
|
"delivery": {
|
|
"n_samples": 60,
|
|
"mse": 0.7353416379416455,
|
|
"rmse": 0.8575206341200459,
|
|
"mae": 0.5434921078461533,
|
|
"r2": 0.021132827618370764
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7380952380952381,
|
|
"precision": 0.7,
|
|
"recall": 0.749266862170088,
|
|
"f1": 0.7053571428571428
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6904761904761905,
|
|
"precision": 0.6361111111111112,
|
|
"recall": 0.7009803921568626,
|
|
"f1": 0.6513176638176638
|
|
},
|
|
"toxic": {
|
|
"n_samples": 61,
|
|
"accuracy": 0.9344262295081968,
|
|
"precision": 0.7142857142857143,
|
|
"recall": 0.9655172413793103,
|
|
"f1": 0.7821428571428571
|
|
},
|
|
"biodist": {
|
|
"n_samples": 60,
|
|
"kl_divergence": 0.2349928026224927,
|
|
"js_divergence": 0.05641933268258879
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 3,
|
|
"best_params": {
|
|
"dropout": 0.10301838844131589,
|
|
"lr": 0.0008648998582653286,
|
|
"weight_decay": 0.0011644667702360517,
|
|
"backbone_lr_ratio": 0.9869864342312137,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 2,
|
|
"llm_lora_r": 32,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 15,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 83,
|
|
"mse": 1.7393398374573046,
|
|
"rmse": 1.3188403381218305,
|
|
"mae": 0.5609516930620534,
|
|
"r2": 0.016216130851348765
|
|
},
|
|
"delivery": {
|
|
"n_samples": 59,
|
|
"mse": 0.7428648348736053,
|
|
"rmse": 0.8618960696473823,
|
|
"mae": 0.5506586653922322,
|
|
"r2": 0.2778471200588467
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7023809523809523,
|
|
"precision": 0.6190306486101211,
|
|
"recall": 0.6224340175953079,
|
|
"f1": 0.6205962059620596
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.5952380952380952,
|
|
"precision": 0.5305408193339228,
|
|
"recall": 0.5427350427350427,
|
|
"f1": 0.525837308103318
|
|
},
|
|
"toxic": {
|
|
"n_samples": 60,
|
|
"accuracy": 0.95,
|
|
"precision": 0.75,
|
|
"recall": 0.9736842105263157,
|
|
"f1": 0.8198198198198199
|
|
},
|
|
"biodist": {
|
|
"n_samples": 60,
|
|
"kl_divergence": 0.15141505126503543,
|
|
"js_divergence": 0.037430317852708686
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 4,
|
|
"best_params": {
|
|
"dropout": 0.303142745975858,
|
|
"lr": 0.0004649241012312578,
|
|
"weight_decay": 3.7624434569437365e-05,
|
|
"backbone_lr_ratio": 0.17310516991766065,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 1,
|
|
"llm_lora_r": 8,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 16,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.6032289134459601,
|
|
"rmse": 0.7766781273126984,
|
|
"mae": 0.48499917273875326,
|
|
"r2": 0.19280478318069083
|
|
},
|
|
"delivery": {
|
|
"n_samples": 58,
|
|
"mse": 1.0297388737573114,
|
|
"rmse": 1.01476050068837,
|
|
"mae": 0.6067489293114893,
|
|
"r2": 0.004259508110989074
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7261904761904762,
|
|
"precision": 0.6686520376175549,
|
|
"recall": 0.6972140762463344,
|
|
"f1": 0.6762192056309704
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6904761904761905,
|
|
"precision": 0.626984126984127,
|
|
"recall": 0.6871794871794873,
|
|
"f1": 0.6442295293359123
|
|
},
|
|
"toxic": {
|
|
"n_samples": 59,
|
|
"accuracy": 0.9661016949152542,
|
|
"precision": 0.8,
|
|
"recall": 0.9821428571428572,
|
|
"f1": 0.865909090909091
|
|
},
|
|
"biodist": {
|
|
"n_samples": 58,
|
|
"kl_divergence": 0.2809668697133116,
|
|
"js_divergence": 0.07302986860825746
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 0,
|
|
"best_params": {
|
|
"dropout": 0.38118503457897024,
|
|
"lr": 0.0009770749364761302,
|
|
"weight_decay": 0.0008416934971427061,
|
|
"backbone_lr_ratio": 0.048484014956722714,
|
|
"moe_n_experts": 2,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 1,
|
|
"llm_lora_r": 16,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 15,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 83,
|
|
"mse": 1.351653676835711,
|
|
"rmse": 1.1626064152737636,
|
|
"mae": 0.5484207456774381,
|
|
"r2": 0.1727546789535871
|
|
},
|
|
"delivery": {
|
|
"n_samples": 58,
|
|
"mse": 0.5919714070454296,
|
|
"rmse": 0.7693967812809134,
|
|
"mae": 0.5870344878399166,
|
|
"r2": 0.2475686197716349
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7261904761904762,
|
|
"precision": 0.69,
|
|
"recall": 0.7302209550962224,
|
|
"f1": 0.6946420104314841
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6785714285714286,
|
|
"precision": 0.6549209197235513,
|
|
"recall": 0.7014285714285715,
|
|
"f1": 0.6513560067776936
|
|
},
|
|
"toxic": {
|
|
"n_samples": 58,
|
|
"accuracy": 0.9655172413793104,
|
|
"precision": 0.75,
|
|
"recall": 0.9821428571428572,
|
|
"f1": 0.8242424242424242
|
|
},
|
|
"biodist": {
|
|
"n_samples": 58,
|
|
"kl_divergence": 0.2953526247170886,
|
|
"js_divergence": 0.053149365074416116
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 1,
|
|
"best_params": {
|
|
"dropout": 0.12715228141755225,
|
|
"lr": 0.00038210169217516986,
|
|
"weight_decay": 0.0032858044759097593,
|
|
"backbone_lr_ratio": 0.01707341286908778,
|
|
"moe_n_experts": 2,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 1,
|
|
"llm_lora_r": 16,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 18,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.3606658162369632,
|
|
"rmse": 0.6005545905552327,
|
|
"mae": 0.46134645698059884,
|
|
"r2": -0.10134517791165032
|
|
},
|
|
"delivery": {
|
|
"n_samples": 61,
|
|
"mse": 1.066438762627893,
|
|
"rmse": 1.0326852195262084,
|
|
"mae": 0.643426902744858,
|
|
"r2": 0.19278273512187016
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6547619047619048,
|
|
"precision": 0.5803571428571429,
|
|
"recall": 0.5952380952380952,
|
|
"f1": 0.5822328931572629
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.5714285714285714,
|
|
"precision": 0.5202020202020202,
|
|
"recall": 0.5292564578278864,
|
|
"f1": 0.5119174942704355
|
|
},
|
|
"toxic": {
|
|
"n_samples": 61,
|
|
"accuracy": 0.9508196721311475,
|
|
"precision": 0.75,
|
|
"recall": 0.9741379310344828,
|
|
"f1": 0.8200589970501475
|
|
},
|
|
"biodist": {
|
|
"n_samples": 61,
|
|
"kl_divergence": 0.46767746325655735,
|
|
"js_divergence": 0.11852224057626487
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 2,
|
|
"best_params": {
|
|
"dropout": 0.33774333325981976,
|
|
"lr": 0.0008937362022799393,
|
|
"weight_decay": 0.0797179350056365,
|
|
"backbone_lr_ratio": 0.012019950560462185,
|
|
"moe_n_experts": 2,
|
|
"moe_top_k": 1,
|
|
"moe_expert_hidden_mult": 2,
|
|
"llm_lora_r": 8,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 24,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.6726206869952828,
|
|
"rmse": 0.8201345542014937,
|
|
"mae": 0.558441653248987,
|
|
"r2": -0.3284138391240339
|
|
},
|
|
"delivery": {
|
|
"n_samples": 60,
|
|
"mse": 0.7104755031729033,
|
|
"rmse": 0.8428970893133416,
|
|
"mae": 0.5519123785896227,
|
|
"r2": 0.05423396289104898
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7261904761904762,
|
|
"precision": 0.6836007130124777,
|
|
"recall": 0.7265395894428153,
|
|
"f1": 0.6891391794046662
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6666666666666666,
|
|
"precision": 0.6064814814814815,
|
|
"recall": 0.6573295985060691,
|
|
"f1": 0.6185851973086015
|
|
},
|
|
"toxic": {
|
|
"n_samples": 61,
|
|
"accuracy": 0.9344262295081968,
|
|
"precision": 0.7142857142857143,
|
|
"recall": 0.9655172413793103,
|
|
"f1": 0.7821428571428571
|
|
},
|
|
"biodist": {
|
|
"n_samples": 60,
|
|
"kl_divergence": 0.22423699727633165,
|
|
"js_divergence": 0.05516182797349107
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 3,
|
|
"best_params": {
|
|
"dropout": 0.10301838844131589,
|
|
"lr": 0.0008648998582653286,
|
|
"weight_decay": 0.0011644667702360517,
|
|
"backbone_lr_ratio": 0.9869864342312137,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 2,
|
|
"llm_lora_r": 32,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 15,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 83,
|
|
"mse": 1.7462471475632266,
|
|
"rmse": 1.3214564493630605,
|
|
"mae": 0.5399843071621045,
|
|
"r2": 0.012309303608578626
|
|
},
|
|
"delivery": {
|
|
"n_samples": 59,
|
|
"mse": 0.6793796847444405,
|
|
"rmse": 0.8242449179366776,
|
|
"mae": 0.4859620340136906,
|
|
"r2": 0.33956222871259445
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6785714285714286,
|
|
"precision": 0.6072124756335282,
|
|
"recall": 0.6209677419354839,
|
|
"f1": 0.6110444177671068
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.5714285714285714,
|
|
"precision": 0.48659611992945323,
|
|
"recall": 0.5015837104072398,
|
|
"f1": 0.4878256038173678
|
|
},
|
|
"toxic": {
|
|
"n_samples": 60,
|
|
"accuracy": 0.95,
|
|
"precision": 0.75,
|
|
"recall": 0.9736842105263157,
|
|
"f1": 0.8198198198198199
|
|
},
|
|
"biodist": {
|
|
"n_samples": 60,
|
|
"kl_divergence": 0.155532256253434,
|
|
"js_divergence": 0.0368257614890642
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 4,
|
|
"best_params": {
|
|
"dropout": 0.303142745975858,
|
|
"lr": 0.0004649241012312578,
|
|
"weight_decay": 3.7624434569437365e-05,
|
|
"backbone_lr_ratio": 0.17310516991766065,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 1,
|
|
"llm_lora_r": 8,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 16,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.6349645228657972,
|
|
"rmse": 0.796846611880729,
|
|
"mae": 0.5075787665887869,
|
|
"r2": 0.15033859570933517
|
|
},
|
|
"delivery": {
|
|
"n_samples": 58,
|
|
"mse": 0.953887574019918,
|
|
"rmse": 0.9766716817948179,
|
|
"mae": 0.6102629316778019,
|
|
"r2": 0.07760646279605954
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7619047619047619,
|
|
"precision": 0.7002652519893899,
|
|
"recall": 0.7214076246334311,
|
|
"f1": 0.7083333333333334
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7023809523809523,
|
|
"precision": 0.6404663314601203,
|
|
"recall": 0.6508547008547009,
|
|
"f1": 0.6445173587126155
|
|
},
|
|
"toxic": {
|
|
"n_samples": 59,
|
|
"accuracy": 0.9661016949152542,
|
|
"precision": 0.8,
|
|
"recall": 0.9821428571428572,
|
|
"f1": 0.865909090909091
|
|
},
|
|
"biodist": {
|
|
"n_samples": 58,
|
|
"kl_divergence": 0.2842203287465849,
|
|
"js_divergence": 0.07675843428767629
|
|
}
|
|
}
|
|
}
|
|
],
|
|
"summary_stats": {
|
|
"size": {
|
|
"mse_mean": 0.9081302997777905,
|
|
"mse_std": 0.5136450803217497,
|
|
"rmse_mean": 0.9141199639383836,
|
|
"rmse_std": 0.2692860770761804,
|
|
"mae_mean": 0.5133458284713402,
|
|
"mae_std": 0.04319170563122826,
|
|
"r2_mean": 0.050279130594863726,
|
|
"r2_std": 0.14254606826389926
|
|
},
|
|
"delivery": {
|
|
"mse_mean": 0.8019226249118218,
|
|
"mse_std": 0.16491413379339875,
|
|
"rmse_mean": 0.8908588101767234,
|
|
"rmse_std": 0.09106703707892613,
|
|
"mae_mean": 0.5784917580608503,
|
|
"mae_std": 0.04315723112464039,
|
|
"r2_mean": 0.17889067221587332,
|
|
"r2_std": 0.09951788455210314
|
|
},
|
|
"pdi": {
|
|
"accuracy_mean": 0.7150793650793652,
|
|
"accuracy_std": 0.03408081040251411,
|
|
"precision_mean": 0.6557721476019204,
|
|
"precision_std": 0.043177655691081666,
|
|
"recall_mean": 0.6793533277933066,
|
|
"recall_std": 0.04998713373151071,
|
|
"f1_mean": 0.660105333753019,
|
|
"f1_std": 0.044135656309097604
|
|
},
|
|
"ee": {
|
|
"accuracy_mean": 0.6444444444444444,
|
|
"accuracy_std": 0.060837358402339045,
|
|
"precision_mean": 0.591231020166536,
|
|
"precision_std": 0.07158423002562818,
|
|
"recall_mean": 0.6249326720587225,
|
|
"recall_std": 0.08964169776063406,
|
|
"f1_mean": 0.594285119491739,
|
|
"f1_std": 0.07762446019388265
|
|
},
|
|
"toxic": {
|
|
"accuracy_mean": 0.9544840786978929,
|
|
"accuracy_std": 0.012127005670401548,
|
|
"precision_mean": 0.7561904761904762,
|
|
"precision_std": 0.02972283149395665,
|
|
"recall_mean": 0.976109814766802,
|
|
"recall_std": 0.006422363902381704,
|
|
"f1_mean": 0.8255180784163084,
|
|
"f1_std": 0.028684876656301855
|
|
},
|
|
"biodist": {
|
|
"kl_divergence_mean": 0.2767846108676604,
|
|
"kl_divergence_std": 0.09472632342725851,
|
|
"js_divergence_mean": 0.06619436643695288,
|
|
"js_divergence_std": 0.02615132013321954
|
|
}
|
|
}
|
|
} |