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": 0.36138334610559464,
|
|
"lr": 0.0009768611684372063,
|
|
"weight_decay": 0.0007819543421024561,
|
|
"backbone_lr_ratio": 0.1094875202574584,
|
|
"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": 10,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 83,
|
|
"mse": 1.5442239384585472,
|
|
"rmse": 1.2426680725191852,
|
|
"mae": 0.6065197844641754,
|
|
"r2": 0.05489693874226986
|
|
},
|
|
"delivery": {
|
|
"n_samples": 58,
|
|
"mse": 0.6228606286654659,
|
|
"rmse": 0.7892151979437965,
|
|
"mae": 0.645549583858971,
|
|
"r2": 0.2083065551159342
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7976190476190477,
|
|
"precision": 0.7473474801061009,
|
|
"recall": 0.7658588738417677,
|
|
"f1": 0.7551020408163265
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7023809523809523,
|
|
"precision": 0.6809523809523809,
|
|
"recall": 0.7176190476190477,
|
|
"f1": 0.6755966243457792
|
|
},
|
|
"toxic": {
|
|
"n_samples": 58,
|
|
"accuracy": 0.9655172413793104,
|
|
"precision": 0.75,
|
|
"recall": 0.9821428571428572,
|
|
"f1": 0.8242424242424242
|
|
},
|
|
"biodist": {
|
|
"n_samples": 58,
|
|
"kl_divergence": 0.29199224249606676,
|
|
"js_divergence": 0.052291809760011645
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 1,
|
|
"best_params": {
|
|
"dropout": 0.2617620972301631,
|
|
"lr": 0.0009708340161977452,
|
|
"weight_decay": 0.00876815717103087,
|
|
"backbone_lr_ratio": 0.07764666819777231,
|
|
"moe_n_experts": 2,
|
|
"moe_top_k": 2,
|
|
"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": 9,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.37954175519764655,
|
|
"rmse": 0.6160696025593589,
|
|
"mae": 0.4762687585538342,
|
|
"r2": -0.15898558467325086
|
|
},
|
|
"delivery": {
|
|
"n_samples": 61,
|
|
"mse": 0.9568238418858714,
|
|
"rmse": 0.9781737278652864,
|
|
"mae": 0.615501540392393,
|
|
"r2": 0.27575332810103925
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7023809523809523,
|
|
"precision": 0.5984375,
|
|
"recall": 0.5952380952380952,
|
|
"f1": 0.5966967543691185
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.5595238095238095,
|
|
"precision": 0.49682149682149684,
|
|
"recall": 0.5036154321868608,
|
|
"f1": 0.48889932920940665
|
|
},
|
|
"toxic": {
|
|
"n_samples": 61,
|
|
"accuracy": 0.9508196721311475,
|
|
"precision": 0.75,
|
|
"recall": 0.9741379310344828,
|
|
"f1": 0.8200589970501475
|
|
},
|
|
"biodist": {
|
|
"n_samples": 61,
|
|
"kl_divergence": 0.2650679664726965,
|
|
"js_divergence": 0.06721545599253585
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 2,
|
|
"best_params": {
|
|
"dropout": 0.3398640929006426,
|
|
"lr": 0.0006400879855817172,
|
|
"weight_decay": 0.00687535028340139,
|
|
"backbone_lr_ratio": 0.3085531888778921,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 1,
|
|
"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": 12,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.34299971144127045,
|
|
"rmse": 0.5856617722211946,
|
|
"mae": 0.4714168618721444,
|
|
"r2": 0.3225816982680323
|
|
},
|
|
"delivery": {
|
|
"n_samples": 60,
|
|
"mse": 0.6305334508551694,
|
|
"rmse": 0.7940613646659617,
|
|
"mae": 0.5496675532388812,
|
|
"r2": 0.16065069039431923
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.75,
|
|
"precision": 0.7014607425441266,
|
|
"recall": 0.7426686217008798,
|
|
"f1": 0.7105824446267432
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6547619047619048,
|
|
"precision": 0.5981481481481481,
|
|
"recall": 0.6507936507936507,
|
|
"f1": 0.6112591899825942
|
|
},
|
|
"toxic": {
|
|
"n_samples": 61,
|
|
"accuracy": 0.9344262295081968,
|
|
"precision": 0.7142857142857143,
|
|
"recall": 0.9655172413793103,
|
|
"f1": 0.7821428571428571
|
|
},
|
|
"biodist": {
|
|
"n_samples": 60,
|
|
"kl_divergence": 0.22453620358096357,
|
|
"js_divergence": 0.05481360908460922
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"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": 1.7015599299013842,
|
|
"rmse": 1.30443854968388,
|
|
"mae": 0.5289686592071351,
|
|
"r2": 0.037584734519839236
|
|
},
|
|
"delivery": {
|
|
"n_samples": 59,
|
|
"mse": 0.6980658831768289,
|
|
"rmse": 0.8355033711343294,
|
|
"mae": 0.5028807028843942,
|
|
"r2": 0.32139702371214895
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6666666666666666,
|
|
"precision": 0.6074074074074074,
|
|
"recall": 0.6275659824046921,
|
|
"f1": 0.6100795755968169
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.5833333333333334,
|
|
"precision": 0.5110428786899375,
|
|
"recall": 0.5272247360482655,
|
|
"f1": 0.5083333333333333
|
|
},
|
|
"toxic": {
|
|
"n_samples": 60,
|
|
"accuracy": 0.9333333333333333,
|
|
"precision": 0.7142857142857143,
|
|
"recall": 0.9649122807017544,
|
|
"f1": 0.7818181818181817
|
|
},
|
|
"biodist": {
|
|
"n_samples": 60,
|
|
"kl_divergence": 0.21258712972173607,
|
|
"js_divergence": 0.05148357428738361
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 4,
|
|
"best_params": {
|
|
"dropout": 0.20496607857127885,
|
|
"lr": 0.0009573095972231165,
|
|
"weight_decay": 0.00038069869878401156,
|
|
"backbone_lr_ratio": 0.04121720375781922,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 1,
|
|
"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": 18,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.7454307837607543,
|
|
"rmse": 0.8633833353503844,
|
|
"mae": 0.4983095005099174,
|
|
"r2": 0.002521016964725842
|
|
},
|
|
"delivery": {
|
|
"n_samples": 58,
|
|
"mse": 1.1305003453510372,
|
|
"rmse": 1.0632498978843294,
|
|
"mae": 0.660637865806448,
|
|
"r2": -0.09317517153949795
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7142857142857143,
|
|
"precision": 0.6592592592592592,
|
|
"recall": 0.6891495601173021,
|
|
"f1": 0.6657824933687002
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6785714285714286,
|
|
"precision": 0.6144079123229084,
|
|
"recall": 0.6371794871794871,
|
|
"f1": 0.6167919799498747
|
|
},
|
|
"toxic": {
|
|
"n_samples": 59,
|
|
"accuracy": 0.9830508474576272,
|
|
"precision": 0.875,
|
|
"recall": 0.9910714285714286,
|
|
"f1": 0.924066924066924
|
|
},
|
|
"biodist": {
|
|
"n_samples": 58,
|
|
"kl_divergence": 0.22363313091427037,
|
|
"js_divergence": 0.05554485020959375
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 0,
|
|
"best_params": {
|
|
"dropout": 0.36138334610559464,
|
|
"lr": 0.0009768611684372063,
|
|
"weight_decay": 0.0007819543421024561,
|
|
"backbone_lr_ratio": 0.1094875202574584,
|
|
"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": 10,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 83,
|
|
"mse": 1.4913369439691264,
|
|
"rmse": 1.2212030723713097,
|
|
"mae": 0.5824022564184235,
|
|
"r2": 0.08726508117798826
|
|
},
|
|
"delivery": {
|
|
"n_samples": 58,
|
|
"mse": 0.6888913417632222,
|
|
"rmse": 0.8299947841783237,
|
|
"mae": 0.6565480354274141,
|
|
"r2": 0.1243775342167962
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7976190476190477,
|
|
"precision": 0.7473474801061009,
|
|
"recall": 0.7658588738417677,
|
|
"f1": 0.7551020408163265
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7142857142857143,
|
|
"precision": 0.6825338404285772,
|
|
"recall": 0.6971428571428572,
|
|
"f1": 0.6715159985600238
|
|
},
|
|
"toxic": {
|
|
"n_samples": 58,
|
|
"accuracy": 0.9655172413793104,
|
|
"precision": 0.75,
|
|
"recall": 0.9821428571428572,
|
|
"f1": 0.8242424242424242
|
|
},
|
|
"biodist": {
|
|
"n_samples": 58,
|
|
"kl_divergence": 0.25535786486716,
|
|
"js_divergence": 0.047258401575109285
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 1,
|
|
"best_params": {
|
|
"dropout": 0.2617620972301631,
|
|
"lr": 0.0009708340161977452,
|
|
"weight_decay": 0.00876815717103087,
|
|
"backbone_lr_ratio": 0.07764666819777231,
|
|
"moe_n_experts": 2,
|
|
"moe_top_k": 2,
|
|
"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": 9,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.4102990857495265,
|
|
"rmse": 0.6405459279001987,
|
|
"mae": 0.48430168065464213,
|
|
"r2": -0.25290753725023607
|
|
},
|
|
"delivery": {
|
|
"n_samples": 61,
|
|
"mse": 1.141709732617368,
|
|
"rmse": 1.0685081808846237,
|
|
"mae": 0.6904619923930188,
|
|
"r2": 0.13580803704365674
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7142857142857143,
|
|
"precision": 0.6597687157638467,
|
|
"recall": 0.6984126984126984,
|
|
"f1": 0.6657824933687002
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.5595238095238095,
|
|
"precision": 0.48331733625851275,
|
|
"recall": 0.4847771276342705,
|
|
"f1": 0.472943722943723
|
|
},
|
|
"toxic": {
|
|
"n_samples": 61,
|
|
"accuracy": 0.9508196721311475,
|
|
"precision": 0.75,
|
|
"recall": 0.9741379310344828,
|
|
"f1": 0.8200589970501475
|
|
},
|
|
"biodist": {
|
|
"n_samples": 61,
|
|
"kl_divergence": 0.27289092812520727,
|
|
"js_divergence": 0.06838914596052793
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 2,
|
|
"best_params": {
|
|
"dropout": 0.3398640929006426,
|
|
"lr": 0.0006400879855817172,
|
|
"weight_decay": 0.00687535028340139,
|
|
"backbone_lr_ratio": 0.3085531888778921,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 1,
|
|
"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": 12,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.5255951670587833,
|
|
"rmse": 0.7249794252658369,
|
|
"mae": 0.5467459253566285,
|
|
"r2": -0.038041064149567205
|
|
},
|
|
"delivery": {
|
|
"n_samples": 60,
|
|
"mse": 0.6715002048252938,
|
|
"rmse": 0.8194511607321658,
|
|
"mae": 0.5382574244169518,
|
|
"r2": 0.10611684034247182
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7142857142857143,
|
|
"precision": 0.6670673076923077,
|
|
"recall": 0.7038123167155426,
|
|
"f1": 0.672514619883041
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6904761904761905,
|
|
"precision": 0.636892026815005,
|
|
"recall": 0.7009803921568626,
|
|
"f1": 0.6545789829099787
|
|
},
|
|
"toxic": {
|
|
"n_samples": 61,
|
|
"accuracy": 0.9344262295081968,
|
|
"precision": 0.7142857142857143,
|
|
"recall": 0.9655172413793103,
|
|
"f1": 0.7821428571428571
|
|
},
|
|
"biodist": {
|
|
"n_samples": 60,
|
|
"kl_divergence": 0.21845003637708307,
|
|
"js_divergence": 0.05363676365676617
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"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": 1.7821711302703098,
|
|
"rmse": 1.3349798239188149,
|
|
"mae": 0.550809847426612,
|
|
"r2": -0.008009574819726861
|
|
},
|
|
"delivery": {
|
|
"n_samples": 59,
|
|
"mse": 0.6972863344754117,
|
|
"rmse": 0.8350367264231027,
|
|
"mae": 0.513030981155649,
|
|
"r2": 0.3221548376688139
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7261904761904762,
|
|
"precision": 0.6552542372881356,
|
|
"recall": 0.6678885630498533,
|
|
"f1": 0.6602778266221205
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.5833333333333334,
|
|
"precision": 0.5107142857142857,
|
|
"recall": 0.5069632981397687,
|
|
"f1": 0.4998670465337132
|
|
},
|
|
"toxic": {
|
|
"n_samples": 60,
|
|
"accuracy": 0.95,
|
|
"precision": 0.75,
|
|
"recall": 0.9736842105263157,
|
|
"f1": 0.8198198198198199
|
|
},
|
|
"biodist": {
|
|
"n_samples": 60,
|
|
"kl_divergence": 0.16841687531678426,
|
|
"js_divergence": 0.04110119757993926
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 4,
|
|
"best_params": {
|
|
"dropout": 0.20496607857127885,
|
|
"lr": 0.0009573095972231165,
|
|
"weight_decay": 0.00038069869878401156,
|
|
"backbone_lr_ratio": 0.04121720375781922,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 1,
|
|
"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": 18,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.5699890881870557,
|
|
"rmse": 0.7549762169678299,
|
|
"mae": 0.4506003191816576,
|
|
"r2": 0.23728379829225832
|
|
},
|
|
"delivery": {
|
|
"n_samples": 58,
|
|
"mse": 0.8941534147821184,
|
|
"rmse": 0.9455968563727982,
|
|
"mae": 0.5503344679700917,
|
|
"r2": 0.1353684086814212
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7738095238095238,
|
|
"precision": 0.7122033898305085,
|
|
"recall": 0.7294721407624634,
|
|
"f1": 0.719359943731317
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6428571428571429,
|
|
"precision": 0.6250750750750751,
|
|
"recall": 0.6393162393162393,
|
|
"f1": 0.6113324253773692
|
|
},
|
|
"toxic": {
|
|
"n_samples": 59,
|
|
"accuracy": 0.9830508474576272,
|
|
"precision": 0.875,
|
|
"recall": 0.9910714285714286,
|
|
"f1": 0.924066924066924
|
|
},
|
|
"biodist": {
|
|
"n_samples": 58,
|
|
"kl_divergence": 0.15337617795379077,
|
|
"js_divergence": 0.03951820037181606
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 0,
|
|
"best_params": {
|
|
"dropout": 0.36138334610559464,
|
|
"lr": 0.0009768611684372063,
|
|
"weight_decay": 0.0007819543421024561,
|
|
"backbone_lr_ratio": 0.1094875202574584,
|
|
"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": 10,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 83,
|
|
"mse": 1.5319067584729609,
|
|
"rmse": 1.2377022091250225,
|
|
"mae": 0.5937045352555901,
|
|
"r2": 0.06243535607962858
|
|
},
|
|
"delivery": {
|
|
"n_samples": 58,
|
|
"mse": 0.6125216869516746,
|
|
"rmse": 0.7826376472874753,
|
|
"mae": 0.6350975606975884,
|
|
"r2": 0.22144797392640647
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7976190476190477,
|
|
"precision": 0.75,
|
|
"recall": 0.7794012829650748,
|
|
"f1": 0.7606837606837606
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7261904761904762,
|
|
"precision": 0.6886574074074074,
|
|
"recall": 0.730952380952381,
|
|
"f1": 0.6892665474060823
|
|
},
|
|
"toxic": {
|
|
"n_samples": 58,
|
|
"accuracy": 0.9655172413793104,
|
|
"precision": 0.75,
|
|
"recall": 0.9821428571428572,
|
|
"f1": 0.8242424242424242
|
|
},
|
|
"biodist": {
|
|
"n_samples": 58,
|
|
"kl_divergence": 0.2734443230053972,
|
|
"js_divergence": 0.049328419061986935
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 1,
|
|
"best_params": {
|
|
"dropout": 0.2617620972301631,
|
|
"lr": 0.0009708340161977452,
|
|
"weight_decay": 0.00876815717103087,
|
|
"backbone_lr_ratio": 0.07764666819777231,
|
|
"moe_n_experts": 2,
|
|
"moe_top_k": 2,
|
|
"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": 9,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.46951935155806995,
|
|
"rmse": 0.6852148214670126,
|
|
"mae": 0.5178531104077896,
|
|
"r2": -0.43374517488216946
|
|
},
|
|
"delivery": {
|
|
"n_samples": 61,
|
|
"mse": 1.0067166172516437,
|
|
"rmse": 1.0033526883661814,
|
|
"mae": 0.6208895779230067,
|
|
"r2": 0.23798809386602826
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7142857142857143,
|
|
"precision": 0.6105263157894737,
|
|
"recall": 0.6031746031746031,
|
|
"f1": 0.60625
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.5595238095238095,
|
|
"precision": 0.5069444444444445,
|
|
"recall": 0.5141049426763712,
|
|
"f1": 0.4983847945308392
|
|
},
|
|
"toxic": {
|
|
"n_samples": 61,
|
|
"accuracy": 0.9508196721311475,
|
|
"precision": 0.75,
|
|
"recall": 0.9741379310344828,
|
|
"f1": 0.8200589970501475
|
|
},
|
|
"biodist": {
|
|
"n_samples": 61,
|
|
"kl_divergence": 0.23066686287196406,
|
|
"js_divergence": 0.05783347774650594
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 2,
|
|
"best_params": {
|
|
"dropout": 0.3398640929006426,
|
|
"lr": 0.0006400879855817172,
|
|
"weight_decay": 0.00687535028340139,
|
|
"backbone_lr_ratio": 0.3085531888778921,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 1,
|
|
"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": 12,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.6330731519688326,
|
|
"rmse": 0.7956589419901171,
|
|
"mae": 0.5650630322218474,
|
|
"r2": -0.2503081640413001
|
|
},
|
|
"delivery": {
|
|
"n_samples": 60,
|
|
"mse": 0.7394687840051654,
|
|
"rmse": 0.8599237082469382,
|
|
"mae": 0.5835061956274633,
|
|
"r2": 0.015638880874769834
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7261904761904762,
|
|
"precision": 0.6686520376175549,
|
|
"recall": 0.6972140762463344,
|
|
"f1": 0.6762192056309704
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6666666666666666,
|
|
"precision": 0.6151596941070625,
|
|
"recall": 0.6785714285714285,
|
|
"f1": 0.6297974020691934
|
|
},
|
|
"toxic": {
|
|
"n_samples": 61,
|
|
"accuracy": 0.9344262295081968,
|
|
"precision": 0.7142857142857143,
|
|
"recall": 0.9655172413793103,
|
|
"f1": 0.7821428571428571
|
|
},
|
|
"biodist": {
|
|
"n_samples": 60,
|
|
"kl_divergence": 0.2593635999587868,
|
|
"js_divergence": 0.06293953369328302
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"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": 1.7226714632498719,
|
|
"rmse": 1.3125057955109654,
|
|
"mae": 0.5719237275960216,
|
|
"r2": 0.02564389034783543
|
|
},
|
|
"delivery": {
|
|
"n_samples": 59,
|
|
"mse": 0.618400296452231,
|
|
"rmse": 0.7863843185442033,
|
|
"mae": 0.48686958986760703,
|
|
"r2": 0.3988414391489876
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6785714285714286,
|
|
"precision": 0.5983050847457627,
|
|
"recall": 0.6063049853372434,
|
|
"f1": 0.6011957095129242
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.5952380952380952,
|
|
"precision": 0.5286355080299572,
|
|
"recall": 0.5540221216691805,
|
|
"f1": 0.532211596393791
|
|
},
|
|
"toxic": {
|
|
"n_samples": 60,
|
|
"accuracy": 0.95,
|
|
"precision": 0.75,
|
|
"recall": 0.9736842105263157,
|
|
"f1": 0.8198198198198199
|
|
},
|
|
"biodist": {
|
|
"n_samples": 60,
|
|
"kl_divergence": 0.13535815823438127,
|
|
"js_divergence": 0.032927490587265504
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 4,
|
|
"best_params": {
|
|
"dropout": 0.20496607857127885,
|
|
"lr": 0.0009573095972231165,
|
|
"weight_decay": 0.00038069869878401156,
|
|
"backbone_lr_ratio": 0.04121720375781922,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 1,
|
|
"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": 18,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.6138974653353957,
|
|
"rmse": 0.7835160913059768,
|
|
"mae": 0.4787395654545565,
|
|
"r2": 0.17852893554874139
|
|
},
|
|
"delivery": {
|
|
"n_samples": 58,
|
|
"mse": 0.918933858582442,
|
|
"rmse": 0.9586103789248487,
|
|
"mae": 0.5818251890355143,
|
|
"r2": 0.11140612860460075
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7142857142857143,
|
|
"precision": 0.6592592592592592,
|
|
"recall": 0.6891495601173021,
|
|
"f1": 0.6657824933687002
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6309523809523809,
|
|
"precision": 0.5510869565217392,
|
|
"recall": 0.5807692307692308,
|
|
"f1": 0.5608224589284287
|
|
},
|
|
"toxic": {
|
|
"n_samples": 59,
|
|
"accuracy": 0.9830508474576272,
|
|
"precision": 0.875,
|
|
"recall": 0.9910714285714286,
|
|
"f1": 0.924066924066924
|
|
},
|
|
"biodist": {
|
|
"n_samples": 58,
|
|
"kl_divergence": 0.1877719140441793,
|
|
"js_divergence": 0.0491918504902384
|
|
}
|
|
}
|
|
}
|
|
],
|
|
"summary_stats": {
|
|
"size": {
|
|
"mse_mean": 0.9642810483053024,
|
|
"mse_std": 0.555690014600843,
|
|
"rmse_mean": 0.9402335772104724,
|
|
"rmse_std": 0.28326995709270125,
|
|
"mae_mean": 0.5282418376387317,
|
|
"mae_std": 0.048170441970642804,
|
|
"r2_mean": -0.008883709991662088,
|
|
"r2_std": 0.1913450489288403
|
|
},
|
|
"delivery": {
|
|
"mse_mean": 0.801891094776063,
|
|
"mse_std": 0.18220205643295026,
|
|
"rmse_mean": 0.8899800006302908,
|
|
"rmse_std": 0.09912967897743918,
|
|
"mae_mean": 0.5887372173796929,
|
|
"mae_std": 0.0612901937191995,
|
|
"r2_mean": 0.17880537334385976,
|
|
"r2_std": 0.12174784172665178
|
|
},
|
|
"pdi": {
|
|
"accuracy_mean": 0.7325396825396825,
|
|
"accuracy_std": 0.04051507624793722,
|
|
"precision_mean": 0.6694864144939896,
|
|
"precision_std": 0.05125867646022508,
|
|
"recall_mean": 0.6907446822617079,
|
|
"recall_std": 0.058907119543823376,
|
|
"f1_mean": 0.6747607601597044,
|
|
"f1_std": 0.05440981896386903
|
|
},
|
|
"ee": {
|
|
"accuracy_mean": 0.6365079365079365,
|
|
"accuracy_std": 0.05748355226105338,
|
|
"precision_mean": 0.5820259594491293,
|
|
"precision_std": 0.07075806198131898,
|
|
"recall_mean": 0.6082688248570601,
|
|
"recall_std": 0.08471623019407899,
|
|
"f1_mean": 0.5814400954982754,
|
|
"f1_std": 0.07366890243489375
|
|
},
|
|
"toxic": {
|
|
"accuracy_mean": 0.9556516869841453,
|
|
"accuracy_std": 0.017373786295358143,
|
|
"precision_mean": 0.7654761904761904,
|
|
"precision_std": 0.05679457353274553,
|
|
"recall_mean": 0.9767259384092415,
|
|
"recall_std": 0.009166871618238513,
|
|
"f1_mean": 0.831532761930992,
|
|
"f1_std": 0.04919763950001062
|
|
},
|
|
"biodist": {
|
|
"kl_divergence_mean": 0.22486089426269779,
|
|
"kl_divergence_std": 0.045328690570765884,
|
|
"js_divergence_mean": 0.05223158533717151,
|
|
"js_divergence_std": 0.009495161673733467
|
|
}
|
|
}
|
|
} |