lnp_ml/models/finetune_cv/cv_results.json
2026-01-23 13:40:22 +08:00

54 lines
1.2 KiB
JSON

{
"fold_results": [
{
"fold_idx": 0,
"best_val_loss": 0.9860520362854004,
"epochs_trained": 40,
"final_train_loss": 0.5008097920152876
},
{
"fold_idx": 1,
"best_val_loss": 2.4599782625834146,
"epochs_trained": 38,
"final_train_loss": 0.564177993271086
},
{
"fold_idx": 2,
"best_val_loss": 0.7660132050514221,
"epochs_trained": 43,
"final_train_loss": 0.6722757054699792
},
{
"fold_idx": 3,
"best_val_loss": 1.065057098865509,
"epochs_trained": 31,
"final_train_loss": 0.7323974437183804
},
{
"fold_idx": 4,
"best_val_loss": 1.321769932905833,
"epochs_trained": 36,
"final_train_loss": 0.5991987817817264
}
],
"summary": {
"val_loss_mean": 1.3197741071383158,
"val_loss_std": 0.5971552245392587
},
"config": {
"d_model": 256,
"num_heads": 8,
"n_attn_layers": 4,
"fusion_strategy": "attention",
"head_hidden_dim": 128,
"dropout": 0.1,
"use_mpnn": true,
"lr": 0.0001,
"weight_decay": 1e-05,
"batch_size": 32,
"epochs": 100,
"patience": 15,
"init_from_pretrain": "models/pretrain_delivery.pt",
"freeze_backbone": false
}
}