mirror of
https://github.com/RYDE-WORK/lnp_ml.git
synced 2026-09-18 14:23:20 +08:00
7eff2b3d02
三个预训练分子编码器作为额外的 chemical token 接入（三选一，互斥）：
- MolFormer-XL 768d（SMILES 序列，11亿分子预训练）
- GraphMVP 300d（2D 图自监督，ICLR'22）
- GROVER 3200d（图 Transformer，NeurIPS'20）
实现：embedding 离线编码存 npy，模型内按 {smiles: vector} 查表注入，
不改动前向逻辑，默认全关、不影响既有实验。
- models.py: 新增三个 CHEM_KEYS_WITH_*、统一的 _load_offline_emb 查表、
proj_input_dims 按开关裁剪、forward 注入；子类同步转发参数
- nested_cv_optuna.py / pretrain.py: 新增 --{molformer,graphmvp,grover}-emb/-csv
- scripts_run/encode_embeddings/: 三个离线编码脚本 + README（含权重来源与
新旧环境兼容补丁说明）
- results/pretrained_encoders/: 五组 15 折结果 summary
(trials15/inner3/repeats3/seed42, 均带 external 预训练)
注：npy embedding 与预训练权重未入库（可由编码脚本复现）
977 lines
28 KiB
JSON
977 lines
28 KiB
JSON
{
|
|
"fold_results": [
|
|
{
|
|
"fold": 0,
|
|
"best_params": {
|
|
"dropout": 0.4884866103742062,
|
|
"lr": 0.0002617457281978971,
|
|
"weight_decay": 9.524902838811199e-05,
|
|
"backbone_lr_ratio": 0.9539957173265954,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 2,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 23,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 83,
|
|
"mse": 1.5279509180840005,
|
|
"rmse": 1.2361031179007682,
|
|
"mae": 0.578313840710255,
|
|
"r2": 0.0648564277702961
|
|
},
|
|
"delivery": {
|
|
"n_samples": 58,
|
|
"mse": 0.6655882063296256,
|
|
"rmse": 0.8158358942395374,
|
|
"mae": 0.674296624894286,
|
|
"r2": 0.15399722555537465
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7619047619047619,
|
|
"precision": 0.7122257053291536,
|
|
"recall": 0.7412687099073414,
|
|
"f1": 0.7214854111405835
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7142857142857143,
|
|
"precision": 0.6748583260680036,
|
|
"recall": 0.7142857142857144,
|
|
"f1": 0.6773495605138607
|
|
},
|
|
"toxic": {
|
|
"n_samples": 58,
|
|
"accuracy": 0.9655172413793104,
|
|
"precision": 0.75,
|
|
"recall": 0.9821428571428572,
|
|
"f1": 0.8242424242424242
|
|
},
|
|
"biodist": {
|
|
"n_samples": 58,
|
|
"kl_divergence": 0.2500928345559933,
|
|
"js_divergence": 0.05156067014803717
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 1,
|
|
"best_params": {
|
|
"dropout": 0.10335173439505978,
|
|
"lr": 0.000196757629557381,
|
|
"weight_decay": 0.0003725444839719432,
|
|
"backbone_lr_ratio": 0.05035847475715459,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 1,
|
|
"moe_expert_hidden_mult": 1,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 19,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.34213962908738765,
|
|
"rmse": 0.584927028856923,
|
|
"mae": 0.45376784687063526,
|
|
"r2": -0.044772788836473065
|
|
},
|
|
"delivery": {
|
|
"n_samples": 61,
|
|
"mse": 1.129722640308642,
|
|
"rmse": 1.0628841142423016,
|
|
"mae": 0.6647483138825561,
|
|
"r2": 0.1448814017846829
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6309523809523809,
|
|
"precision": 0.5535714285714286,
|
|
"recall": 0.5634920634920635,
|
|
"f1": 0.553421368547419
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.5238095238095238,
|
|
"precision": 0.46498599439775906,
|
|
"recall": 0.46436896436896435,
|
|
"f1": 0.4483749534478767
|
|
},
|
|
"toxic": {
|
|
"n_samples": 61,
|
|
"accuracy": 0.9508196721311475,
|
|
"precision": 0.75,
|
|
"recall": 0.9741379310344828,
|
|
"f1": 0.8200589970501475
|
|
},
|
|
"biodist": {
|
|
"n_samples": 61,
|
|
"kl_divergence": 0.49848379935596204,
|
|
"js_divergence": 0.1305599268889464
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 2,
|
|
"best_params": {
|
|
"dropout": 0.2514825943283291,
|
|
"lr": 0.00029260650717186565,
|
|
"weight_decay": 0.0001495891705646636,
|
|
"backbone_lr_ratio": 0.2434609849039314,
|
|
"moe_n_experts": 2,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 1,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 20,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.3851340701995523,
|
|
"rmse": 0.6205917097412375,
|
|
"mae": 0.4943395415542736,
|
|
"r2": 0.2393670925336251
|
|
},
|
|
"delivery": {
|
|
"n_samples": 60,
|
|
"mse": 0.7088597182801071,
|
|
"rmse": 0.8419380727108776,
|
|
"mae": 0.5629216438857838,
|
|
"r2": 0.0563848526375017
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6904761904761905,
|
|
"precision": 0.6505882352941177,
|
|
"recall": 0.6876832844574781,
|
|
"f1": 0.6517857142857143
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6666666666666666,
|
|
"precision": 0.6139316239316239,
|
|
"recall": 0.6785714285714285,
|
|
"f1": 0.6277576489533011
|
|
},
|
|
"toxic": {
|
|
"n_samples": 61,
|
|
"accuracy": 0.9344262295081968,
|
|
"precision": 0.7142857142857143,
|
|
"recall": 0.9655172413793103,
|
|
"f1": 0.7821428571428571
|
|
},
|
|
"biodist": {
|
|
"n_samples": 60,
|
|
"kl_divergence": 0.2665180519269797,
|
|
"js_divergence": 0.06450481128744694
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 3,
|
|
"best_params": {
|
|
"dropout": 0.40463841601213524,
|
|
"lr": 0.0002572978619380859,
|
|
"weight_decay": 2.0354315009822035e-05,
|
|
"backbone_lr_ratio": 0.8823918523568081,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 2,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 19,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 83,
|
|
"mse": 1.820237956048593,
|
|
"rmse": 1.3491619458199202,
|
|
"mae": 0.6053206579400653,
|
|
"r2": -0.029540461621649028
|
|
},
|
|
"delivery": {
|
|
"n_samples": 59,
|
|
"mse": 0.6693395316760985,
|
|
"rmse": 0.8181317324710603,
|
|
"mae": 0.5448598891843931,
|
|
"r2": 0.34932245037470455
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6785714285714286,
|
|
"precision": 0.6159874608150471,
|
|
"recall": 0.6356304985337243,
|
|
"f1": 0.6199095022624435
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6309523809523809,
|
|
"precision": 0.5653410228683567,
|
|
"recall": 0.5826043237807944,
|
|
"f1": 0.5650366426763943
|
|
},
|
|
"toxic": {
|
|
"n_samples": 60,
|
|
"accuracy": 0.9333333333333333,
|
|
"precision": 0.7142857142857143,
|
|
"recall": 0.9649122807017544,
|
|
"f1": 0.7818181818181817
|
|
},
|
|
"biodist": {
|
|
"n_samples": 60,
|
|
"kl_divergence": 0.29215591110500194,
|
|
"js_divergence": 0.07269076599614506
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 4,
|
|
"best_params": {
|
|
"dropout": 0.3730299609537786,
|
|
"lr": 0.0007646220605180808,
|
|
"weight_decay": 0.004588258639584098,
|
|
"backbone_lr_ratio": 0.17129190112240483,
|
|
"moe_n_experts": 2,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 2,
|
|
"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": 84,
|
|
"mse": 0.7676733172863143,
|
|
"rmse": 0.8761696852130381,
|
|
"mae": 0.5270063721386361,
|
|
"r2": -0.02724225576902284
|
|
},
|
|
"delivery": {
|
|
"n_samples": 58,
|
|
"mse": 0.9040350302083842,
|
|
"rmse": 0.9508075673912068,
|
|
"mae": 0.6287906430019387,
|
|
"r2": 0.12581304969093698
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7261904761904762,
|
|
"precision": 0.6836007130124777,
|
|
"recall": 0.7265395894428153,
|
|
"f1": 0.6891391794046662
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7023809523809523,
|
|
"precision": 0.6645240097721175,
|
|
"recall": 0.7038461538461539,
|
|
"f1": 0.6728729539789446
|
|
},
|
|
"toxic": {
|
|
"n_samples": 59,
|
|
"accuracy": 0.9661016949152542,
|
|
"precision": 0.8,
|
|
"recall": 0.9821428571428572,
|
|
"f1": 0.865909090909091
|
|
},
|
|
"biodist": {
|
|
"n_samples": 58,
|
|
"kl_divergence": 0.20431460217894573,
|
|
"js_divergence": 0.05043386132176378
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 0,
|
|
"best_params": {
|
|
"dropout": 0.4884866103742062,
|
|
"lr": 0.0002617457281978971,
|
|
"weight_decay": 9.524902838811199e-05,
|
|
"backbone_lr_ratio": 0.9539957173265954,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 2,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 23,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 83,
|
|
"mse": 1.5550563846692986,
|
|
"rmse": 1.2470189993217018,
|
|
"mae": 0.5954494658094572,
|
|
"r2": 0.04826721502168674
|
|
},
|
|
"delivery": {
|
|
"n_samples": 58,
|
|
"mse": 0.704182837608261,
|
|
"rmse": 0.8391560269748772,
|
|
"mae": 0.6780449980548744,
|
|
"r2": 0.10494112024899149
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7142857142857143,
|
|
"precision": 0.6737967914438503,
|
|
"recall": 0.7084818246614397,
|
|
"f1": 0.6785714285714286
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7261904761904762,
|
|
"precision": 0.6931286549707601,
|
|
"recall": 0.7380952380952381,
|
|
"f1": 0.7007070707070707
|
|
},
|
|
"toxic": {
|
|
"n_samples": 58,
|
|
"accuracy": 0.9655172413793104,
|
|
"precision": 0.75,
|
|
"recall": 0.9821428571428572,
|
|
"f1": 0.8242424242424242
|
|
},
|
|
"biodist": {
|
|
"n_samples": 58,
|
|
"kl_divergence": 0.2453488369022721,
|
|
"js_divergence": 0.052406074597242516
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 1,
|
|
"best_params": {
|
|
"dropout": 0.10335173439505978,
|
|
"lr": 0.000196757629557381,
|
|
"weight_decay": 0.0003725444839719432,
|
|
"backbone_lr_ratio": 0.05035847475715459,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 1,
|
|
"moe_expert_hidden_mult": 1,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 19,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.3727706424788934,
|
|
"rmse": 0.610549459486202,
|
|
"mae": 0.48057969708350445,
|
|
"r2": -0.13830901371429016
|
|
},
|
|
"delivery": {
|
|
"n_samples": 61,
|
|
"mse": 1.0803452340552668,
|
|
"rmse": 1.0393965720817377,
|
|
"mae": 0.6485996645432516,
|
|
"r2": 0.18225653875401804
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6309523809523809,
|
|
"precision": 0.5535714285714286,
|
|
"recall": 0.5634920634920635,
|
|
"f1": 0.553421368547419
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.5476190476190477,
|
|
"precision": 0.49737237237237236,
|
|
"recall": 0.5073022215879358,
|
|
"f1": 0.4920019687461548
|
|
},
|
|
"toxic": {
|
|
"n_samples": 61,
|
|
"accuracy": 0.9508196721311475,
|
|
"precision": 0.75,
|
|
"recall": 0.9741379310344828,
|
|
"f1": 0.8200589970501475
|
|
},
|
|
"biodist": {
|
|
"n_samples": 61,
|
|
"kl_divergence": 0.5055220562600806,
|
|
"js_divergence": 0.1300376337593102
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 2,
|
|
"best_params": {
|
|
"dropout": 0.2514825943283291,
|
|
"lr": 0.00029260650717186565,
|
|
"weight_decay": 0.0001495891705646636,
|
|
"backbone_lr_ratio": 0.2434609849039314,
|
|
"moe_n_experts": 2,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 1,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 20,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.501474748290053,
|
|
"rmse": 0.708148817897801,
|
|
"mae": 0.5538181106981245,
|
|
"r2": 0.009596331959947801
|
|
},
|
|
"delivery": {
|
|
"n_samples": 60,
|
|
"mse": 0.739867919930154,
|
|
"rmse": 0.8601557532971305,
|
|
"mae": 0.5739390922787909,
|
|
"r2": 0.015107561778814449
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7380952380952381,
|
|
"precision": 0.6923076923076923,
|
|
"recall": 0.7346041055718475,
|
|
"f1": 0.6998050682261208
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6785714285714286,
|
|
"precision": 0.6228947368421053,
|
|
"recall": 0.6851073762838468,
|
|
"f1": 0.63836444060848
|
|
},
|
|
"toxic": {
|
|
"n_samples": 61,
|
|
"accuracy": 0.9344262295081968,
|
|
"precision": 0.7142857142857143,
|
|
"recall": 0.9655172413793103,
|
|
"f1": 0.7821428571428571
|
|
},
|
|
"biodist": {
|
|
"n_samples": 60,
|
|
"kl_divergence": 0.2914659417557568,
|
|
"js_divergence": 0.07149076417555128
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 3,
|
|
"best_params": {
|
|
"dropout": 0.40463841601213524,
|
|
"lr": 0.0002572978619380859,
|
|
"weight_decay": 2.0354315009822035e-05,
|
|
"backbone_lr_ratio": 0.8823918523568081,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 2,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 19,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 83,
|
|
"mse": 1.7508482794444888,
|
|
"rmse": 1.3231962361813492,
|
|
"mae": 0.5995979247704507,
|
|
"r2": 0.009706868347147068
|
|
},
|
|
"delivery": {
|
|
"n_samples": 59,
|
|
"mse": 0.6800398997076931,
|
|
"rmse": 0.824645317520019,
|
|
"mae": 0.5383539610084588,
|
|
"r2": 0.33892042132757505
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7023809523809523,
|
|
"precision": 0.6423197492163009,
|
|
"recall": 0.6664222873900294,
|
|
"f1": 0.6480643539467068
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6666666666666666,
|
|
"precision": 0.5898730898730898,
|
|
"recall": 0.6022121669180492,
|
|
"f1": 0.5907255864702673
|
|
},
|
|
"toxic": {
|
|
"n_samples": 60,
|
|
"accuracy": 0.9166666666666666,
|
|
"precision": 0.6875,
|
|
"recall": 0.956140350877193,
|
|
"f1": 0.749791492910759
|
|
},
|
|
"biodist": {
|
|
"n_samples": 60,
|
|
"kl_divergence": 0.22697074487882685,
|
|
"js_divergence": 0.05572372821276913
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 4,
|
|
"best_params": {
|
|
"dropout": 0.3730299609537786,
|
|
"lr": 0.0007646220605180808,
|
|
"weight_decay": 0.004588258639584098,
|
|
"backbone_lr_ratio": 0.17129190112240483,
|
|
"moe_n_experts": 2,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 2,
|
|
"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": 84,
|
|
"mse": 0.7388933186856751,
|
|
"rmse": 0.8595890405802502,
|
|
"mae": 0.4949821270059883,
|
|
"r2": 0.01126895729235744
|
|
},
|
|
"delivery": {
|
|
"n_samples": 58,
|
|
"mse": 1.0735857604041814,
|
|
"rmse": 1.0361398363175607,
|
|
"mae": 0.6398222603160759,
|
|
"r2": -0.038139707447659266
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7023809523809523,
|
|
"precision": 0.6345029239766082,
|
|
"recall": 0.6517595307917888,
|
|
"f1": 0.6398559423769508
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7261904761904762,
|
|
"precision": 0.662962962962963,
|
|
"recall": 0.7064102564102565,
|
|
"f1": 0.6789544933874829
|
|
},
|
|
"toxic": {
|
|
"n_samples": 59,
|
|
"accuracy": 0.9661016949152542,
|
|
"precision": 0.8,
|
|
"recall": 0.9821428571428572,
|
|
"f1": 0.865909090909091
|
|
},
|
|
"biodist": {
|
|
"n_samples": 58,
|
|
"kl_divergence": 0.2849032950499658,
|
|
"js_divergence": 0.0742269697587313
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 0,
|
|
"best_params": {
|
|
"dropout": 0.4884866103742062,
|
|
"lr": 0.0002617457281978971,
|
|
"weight_decay": 9.524902838811199e-05,
|
|
"backbone_lr_ratio": 0.9539957173265954,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 2,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 23,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 83,
|
|
"mse": 1.575393168842647,
|
|
"rmse": 1.255146672242988,
|
|
"mae": 0.5922032207340361,
|
|
"r2": 0.035820602519645584
|
|
},
|
|
"delivery": {
|
|
"n_samples": 58,
|
|
"mse": 0.6805529919450708,
|
|
"rmse": 0.8249563576002495,
|
|
"mae": 0.6910147559026192,
|
|
"r2": 0.13497608000435835
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7976190476190477,
|
|
"precision": 0.75,
|
|
"recall": 0.7794012829650748,
|
|
"f1": 0.7606837606837606
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7380952380952381,
|
|
"precision": 0.7064443475733798,
|
|
"recall": 0.7547619047619047,
|
|
"f1": 0.7126436781609194
|
|
},
|
|
"toxic": {
|
|
"n_samples": 58,
|
|
"accuracy": 0.9655172413793104,
|
|
"precision": 0.75,
|
|
"recall": 0.9821428571428572,
|
|
"f1": 0.8242424242424242
|
|
},
|
|
"biodist": {
|
|
"n_samples": 58,
|
|
"kl_divergence": 0.23590056396282402,
|
|
"js_divergence": 0.047894235451583785
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 1,
|
|
"best_params": {
|
|
"dropout": 0.10335173439505978,
|
|
"lr": 0.000196757629557381,
|
|
"weight_decay": 0.0003725444839719432,
|
|
"backbone_lr_ratio": 0.05035847475715459,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 1,
|
|
"moe_expert_hidden_mult": 1,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 19,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.3579244858231685,
|
|
"rmse": 0.5982679047242703,
|
|
"mae": 0.46379138462777647,
|
|
"r2": -0.09297412943304506
|
|
},
|
|
"delivery": {
|
|
"n_samples": 61,
|
|
"mse": 1.1096558552074713,
|
|
"rmse": 1.0534020387332992,
|
|
"mae": 0.6546663998885721,
|
|
"r2": 0.1600705115131673
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6547619047619048,
|
|
"precision": 0.5696286472148542,
|
|
"recall": 0.5793650793650793,
|
|
"f1": 0.571654650958326
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.5476190476190477,
|
|
"precision": 0.49253034547152197,
|
|
"recall": 0.49681271109842545,
|
|
"f1": 0.4801587301587302
|
|
},
|
|
"toxic": {
|
|
"n_samples": 61,
|
|
"accuracy": 0.9344262295081968,
|
|
"precision": 0.7142857142857143,
|
|
"recall": 0.9655172413793103,
|
|
"f1": 0.7821428571428571
|
|
},
|
|
"biodist": {
|
|
"n_samples": 61,
|
|
"kl_divergence": 0.502867985249319,
|
|
"js_divergence": 0.13123665503825419
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 2,
|
|
"best_params": {
|
|
"dropout": 0.2514825943283291,
|
|
"lr": 0.00029260650717186565,
|
|
"weight_decay": 0.0001495891705646636,
|
|
"backbone_lr_ratio": 0.2434609849039314,
|
|
"moe_n_experts": 2,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 1,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 20,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 84,
|
|
"mse": 0.618457534400238,
|
|
"rmse": 0.7864207108159335,
|
|
"mae": 0.5679304331563235,
|
|
"r2": -0.22144258048008303
|
|
},
|
|
"delivery": {
|
|
"n_samples": 60,
|
|
"mse": 0.7362245815365097,
|
|
"rmse": 0.8580353031994137,
|
|
"mae": 0.5702473803190514,
|
|
"r2": 0.019957476658378215
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6785714285714286,
|
|
"precision": 0.6523864289821737,
|
|
"recall": 0.6942815249266863,
|
|
"f1": 0.6473332296687918
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6785714285714286,
|
|
"precision": 0.6228947368421053,
|
|
"recall": 0.6851073762838468,
|
|
"f1": 0.63836444060848
|
|
},
|
|
"toxic": {
|
|
"n_samples": 61,
|
|
"accuracy": 0.9344262295081968,
|
|
"precision": 0.7142857142857143,
|
|
"recall": 0.9655172413793103,
|
|
"f1": 0.7821428571428571
|
|
},
|
|
"biodist": {
|
|
"n_samples": 60,
|
|
"kl_divergence": 0.26356343521196823,
|
|
"js_divergence": 0.06378606399643424
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 3,
|
|
"best_params": {
|
|
"dropout": 0.40463841601213524,
|
|
"lr": 0.0002572978619380859,
|
|
"weight_decay": 2.0354315009822035e-05,
|
|
"backbone_lr_ratio": 0.8823918523568081,
|
|
"moe_n_experts": 4,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 2,
|
|
"d_model": 256,
|
|
"num_heads": 8,
|
|
"n_attn_layers": 4,
|
|
"fusion_strategy": "attention",
|
|
"head_hidden_dim": 128,
|
|
"set_transformer_block": "sab"
|
|
},
|
|
"epoch_mean": 19,
|
|
"test_metrics": {
|
|
"size": {
|
|
"n_samples": 83,
|
|
"mse": 1.8408361971730496,
|
|
"rmse": 1.3567741879815702,
|
|
"mae": 0.6257544057668153,
|
|
"r2": -0.04119098379947639
|
|
},
|
|
"delivery": {
|
|
"n_samples": 59,
|
|
"mse": 0.737776343129472,
|
|
"rmse": 0.8589390799873249,
|
|
"mae": 0.5497218715026975,
|
|
"r2": 0.28279373860245827
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6904761904761905,
|
|
"precision": 0.6333333333333333,
|
|
"recall": 0.6583577712609971,
|
|
"f1": 0.6379310344827587
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.6428571428571429,
|
|
"precision": 0.5640697674418605,
|
|
"recall": 0.5891402714932127,
|
|
"f1": 0.5710279611967066
|
|
},
|
|
"toxic": {
|
|
"n_samples": 60,
|
|
"accuracy": 0.9333333333333333,
|
|
"precision": 0.7142857142857143,
|
|
"recall": 0.9649122807017544,
|
|
"f1": 0.7818181818181817
|
|
},
|
|
"biodist": {
|
|
"n_samples": 60,
|
|
"kl_divergence": 0.24488500401793606,
|
|
"js_divergence": 0.060842241588791174
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"fold": 4,
|
|
"best_params": {
|
|
"dropout": 0.3730299609537786,
|
|
"lr": 0.0007646220605180808,
|
|
"weight_decay": 0.004588258639584098,
|
|
"backbone_lr_ratio": 0.17129190112240483,
|
|
"moe_n_experts": 2,
|
|
"moe_top_k": 2,
|
|
"moe_expert_hidden_mult": 2,
|
|
"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": 84,
|
|
"mse": 0.712297885673984,
|
|
"rmse": 0.8439774201209319,
|
|
"mae": 0.5084991393599748,
|
|
"r2": 0.046856950237922024
|
|
},
|
|
"delivery": {
|
|
"n_samples": 58,
|
|
"mse": 1.1055967310671657,
|
|
"rmse": 1.0514735997956228,
|
|
"mae": 0.6599975297677105,
|
|
"r2": -0.06909378763839769
|
|
},
|
|
"pdi": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7619047619047619,
|
|
"precision": 0.711111111111111,
|
|
"recall": 0.750733137829912,
|
|
"f1": 0.7214854111405835
|
|
},
|
|
"ee": {
|
|
"n_samples": 84,
|
|
"accuracy": 0.7023809523809523,
|
|
"precision": 0.6536363636363637,
|
|
"recall": 0.7038461538461539,
|
|
"f1": 0.6717592592592593
|
|
},
|
|
"toxic": {
|
|
"n_samples": 59,
|
|
"accuracy": 0.9661016949152542,
|
|
"precision": 0.8,
|
|
"recall": 0.9821428571428572,
|
|
"f1": 0.865909090909091
|
|
},
|
|
"biodist": {
|
|
"n_samples": 58,
|
|
"kl_divergence": 0.1782104997564988,
|
|
"js_divergence": 0.04297224929274232
|
|
}
|
|
}
|
|
}
|
|
],
|
|
"summary_stats": {
|
|
"size": {
|
|
"mse_mean": 0.9911392357458229,
|
|
"mse_std": 0.5817120308358389,
|
|
"rmse_mean": 0.9504028624589923,
|
|
"rmse_std": 0.29643487442535593,
|
|
"mae_mean": 0.542756944548421,
|
|
"mae_std": 0.054981488862431024,
|
|
"r2_mean": -0.008648784531427449,
|
|
"r2_std": 0.09954199193808574
|
|
},
|
|
"delivery": {
|
|
"mse_mean": 0.8483582187596068,
|
|
"mse_std": 0.18619352158793398,
|
|
"rmse_mean": 0.9157264844374813,
|
|
"rmse_std": 0.09901123400593626,
|
|
"mae_mean": 0.6186683352287373,
|
|
"mae_std": 0.05324906806818642,
|
|
"r2_mean": 0.13081259558966032,
|
|
"r2_std": 0.1207826593867956
|
|
},
|
|
"pdi": {
|
|
"accuracy_mean": 0.703968253968254,
|
|
"accuracy_std": 0.04618202071738462,
|
|
"precision_mean": 0.6485954432786384,
|
|
"precision_std": 0.056485137303304574,
|
|
"recall_mean": 0.6761008502725561,
|
|
"recall_std": 0.06599285309996766,
|
|
"f1_mean": 0.6529698282829115,
|
|
"f1_std": 0.05951390665637187
|
|
},
|
|
"ee": {
|
|
"accuracy_mean": 0.6595238095238095,
|
|
"accuracy_std": 0.0668930397006523,
|
|
"precision_mean": 0.6059632236682923,
|
|
"precision_std": 0.07313293968503662,
|
|
"recall_mean": 0.640831484108795,
|
|
"recall_std": 0.0904387508717925,
|
|
"f1_mean": 0.6110732925915952,
|
|
"f1_std": 0.08105368908168387
|
|
},
|
|
"toxic": {
|
|
"accuracy_mean": 0.947835626967474,
|
|
"accuracy_std": 0.01646827276703399,
|
|
"precision_mean": 0.741547619047619,
|
|
"precision_std": 0.034629149636327315,
|
|
"recall_mean": 0.9726111255149369,
|
|
"recall_std": 0.008721274087923442,
|
|
"f1_mean": 0.810171454978226,
|
|
"f1_std": 0.03512408564460891
|
|
},
|
|
"biodist": {
|
|
"kl_divergence_mean": 0.299413570811222,
|
|
"kl_divergence_std": 0.1056856517385418,
|
|
"js_divergence_mean": 0.0733577767675833,
|
|
"js_divergence_std": 0.029986885602095156
|
|
}
|
|
}
|
|
} |