Add train_cv(without pretrain)

Makefile

@@ -74,6 +74,11 @@ data_pretrain: requirements
 	$(PYTHON_INTERPRETER) scripts/process_external.py
 
 ## Process CV data for cross-validation pretrain (external/all_amine_split_for_LiON -> processed/cv)
+.PHONY: data_pretrain_cv
+data_pretrain_cv: requirements
+	$(PYTHON_INTERPRETER) scripts/process_external_cv.py
+
+## Process internal data with amine-based CV splitting (interim -> processed/cv)
 .PHONY: data_cv
 data_cv: requirements
 	$(PYTHON_INTERPRETER) scripts/process_data_cv.py
@@ -106,8 +111,8 @@ pretrain_cv: requirements
 	$(PYTHON_INTERPRETER) -m lnp_ml.modeling.pretrain_cv main $(MPNN_FLAG) $(DEVICE_FLAG)
 
 ## Evaluate CV pretrain models on test sets (auto-detects MPNN from checkpoint)
-.PHONY: test_cv
-test_cv: requirements
+.PHONY: test_pretrain_cv
+test_pretrain_cv: requirements
 	$(PYTHON_INTERPRETER) -m lnp_ml.modeling.pretrain_cv test $(DEVICE_FLAG)
 
 ## Train model (multi-task, from scratch)
@@ -120,6 +125,22 @@ train: requirements
 finetune: requirements
 	$(PYTHON_INTERPRETER) -m lnp_ml.modeling.train --init-from-pretrain models/pretrain_delivery.pt $(FREEZE_FLAG) $(MPNN_FLAG) $(DEVICE_FLAG)
 
+## Finetune with cross-validation on internal data (5-fold, amine-based split) with pretrained weights
+.PHONY: finetune_cv
+finetune_cv: requirements
+	$(PYTHON_INTERPRETER) -m lnp_ml.modeling.train_cv main --init-from-pretrain models/pretrain_delivery.pt $(FREEZE_FLAG) $(MPNN_FLAG) $(DEVICE_FLAG)
+
+## Train with cross-validation on internal data only (5-fold, amine-based split)
+.PHONY: train_cv
+train_cv: requirements
+	$(PYTHON_INTERPRETER) -m lnp_ml.modeling.train_cv main $(FREEZE_FLAG) $(MPNN_FLAG) $(DEVICE_FLAG)
+
+
+## Evaluate CV finetuned models on test sets (auto-detects MPNN from checkpoint)
+.PHONY: test_cv
+test_cv: requirements
+	$(PYTHON_INTERPRETER) -m lnp_ml.modeling.train_cv test $(DEVICE_FLAG)
+
 ## Train with hyperparameter tuning
 .PHONY: tune
 tune: requirements

data/processed/cv/feature_columns.txt

@@ -1,9 +1,16 @@
+# Feature columns configuration
+
+# SMILES
 smiles
+
+# comp token [5]
 Cationic_Lipid_to_mRNA_weight_ratio
 Cationic_Lipid_Mol_Ratio
 Phospholipid_Mol_Ratio
 Cholesterol_Mol_Ratio
 PEG_Lipid_Mol_Ratio
+
+# phys token [12]
 Purity_Pure
 Purity_Crude
 Mix_type_Microfluidic
@@ -16,10 +23,14 @@ Target_or_delivered_gene_Peptide_barcode
 Target_or_delivered_gene_hEPO
 Target_or_delivered_gene_FVII
 Target_or_delivered_gene_GFP
+
+# help token [4]
 Helper_lipid_ID_DOPE
 Helper_lipid_ID_DOTAP
 Helper_lipid_ID_DSPC
 Helper_lipid_ID_MDOA
+
+# exp token [32]
 Model_type_A549
 Model_type_BDMC
 Model_type_BMDM
@@ -52,4 +63,27 @@ Value_name_hEPO
 Value_name_FVII_silencing
 Value_name_GFP_delivery
 Value_name_Discretized_luminescence
-quantified_delivery
+
+# Targets
+## Regression
+size
+quantified_delivery
+## PDI classification
+PDI_0_0to0_2
+PDI_0_2to0_3
+PDI_0_3to0_4
+PDI_0_4to0_5
+## EE classification
+Encapsulation_Efficiency_EE<50
+Encapsulation_Efficiency_50<=EE<80
+Encapsulation_Efficiency_80<EE<=100
+## Toxic
+toxic
+## Biodistribution
+Biodistribution_lymph_nodes
+Biodistribution_heart
+Biodistribution_liver
+Biodistribution_spleen
+Biodistribution_lung
+Biodistribution_kidney
+Biodistribution_muscle

data/processed/pretrain_cv/feature_columns.txt

@@ -0,0 +1,55 @@
+smiles
+Cationic_Lipid_to_mRNA_weight_ratio
+Cationic_Lipid_Mol_Ratio
+Phospholipid_Mol_Ratio
+Cholesterol_Mol_Ratio
+PEG_Lipid_Mol_Ratio
+Purity_Pure
+Purity_Crude
+Mix_type_Microfluidic
+Mix_type_Pipetting
+Cargo_type_mRNA
+Cargo_type_pDNA
+Cargo_type_siRNA
+Target_or_delivered_gene_FFL
+Target_or_delivered_gene_Peptide_barcode
+Target_or_delivered_gene_hEPO
+Target_or_delivered_gene_FVII
+Target_or_delivered_gene_GFP
+Helper_lipid_ID_DOPE
+Helper_lipid_ID_DOTAP
+Helper_lipid_ID_DSPC
+Helper_lipid_ID_MDOA
+Model_type_A549
+Model_type_BDMC
+Model_type_BMDM
+Model_type_HBEC_ALI
+Model_type_HEK293T
+Model_type_HeLa
+Model_type_IGROV1
+Model_type_Mouse
+Model_type_RAW264p7
+Delivery_target_body
+Delivery_target_dendritic_cell
+Delivery_target_generic_cell
+Delivery_target_liver
+Delivery_target_lung
+Delivery_target_lung_epithelium
+Delivery_target_macrophage
+Delivery_target_muscle
+Delivery_target_spleen
+Route_of_administration_in_vitro
+Route_of_administration_intramuscular
+Route_of_administration_intratracheal
+Route_of_administration_intravenous
+Batch_or_individual_or_barcoded_Barcoded
+Batch_or_individual_or_barcoded_Individual
+Value_name_log_luminescence
+Value_name_luminescence
+Value_name_FFL_silencing
+Value_name_Peptide_abundance
+Value_name_hEPO
+Value_name_FVII_silencing
+Value_name_GFP_delivery
+Value_name_Discretized_luminescence
+quantified_delivery

data/processed/test_results.json

@@ -1,37 +1,51 @@
 {
   "loss_metrics": {
-    "loss": 2.8661977450052896,
-    "loss_size": 0.44916408757368725,
-    "loss_pdi": 0.5041926403840383,
-    "loss_ee": 0.9021427234013876,
-    "loss_delivery": 0.5761533578236898,
-    "loss_biodist": 0.4019051690896352,
-    "loss_toxic": 0.03263980595511384,
-    "acc_pdi": 0.7633587786259542,
-    "acc_ee": 0.6641221374045801,
-    "acc_toxic": 0.9702970297029703
+    "loss": 2.5374555587768555,
+    "loss_size": 0.1886825958887736,
+    "loss_pdi": 0.45798932512601215,
+    "loss_ee": 0.829658567905426,
+    "loss_delivery": 0.4857304096221924,
+    "loss_biodist": 0.5346279243628184,
+    "loss_toxic": 0.04076674363265435,
+    "acc_pdi": 0.7862595419847328,
+    "acc_ee": 0.6793893129770993,
+    "acc_toxic": 0.9801980198019802
   },
   "detailed_metrics": {
     "size": {
-      "mse": 0.41126506251447736,
-      "rmse": 0.6412995107704959,
-      "mae": 0.41415552388095633,
-      "r2": -0.9333718010891026
+      "mse": 0.1669999969286325,
+      "rmse": 0.4086563310761654,
+      "mae": 0.26111859684375066,
+      "r2": 0.2149270281561566
     },
     "delivery": {
-      "mse": 0.6277965050686476,
-      "rmse": 0.7923361061245711,
-      "mae": 0.5387302115022443,
-      "r2": 0.24206702565575944
+      "mse": 0.5193460523366603,
+      "rmse": 0.7206566813238189,
+      "mae": 0.4828052782115008,
+      "r2": 0.37299826459145
     },
     "pdi": {
-      "accuracy": 0.7633587786259542
+      "accuracy": 0.7862595419847328,
+      "precision": 0.7282763532763532,
+      "recall": 0.6907738095238095,
+      "f1": 0.7041935483870968
     },
     "ee": {
-      "accuracy": 0.6641221374045801
+      "accuracy": 0.6793893129770993,
+      "precision": 0.612247574088644,
+      "recall": 0.6062951496388029,
+      "f1": 0.6069449904342585
     },
     "toxic": {
-      "accuracy": 0.9702970297029703
+      "accuracy": 0.9801980198019802,
+      "precision": 0.5,
+      "recall": 0.4900990099009901,
+      "f1": 0.495
+    },
+    "biodist": {
+      "n_samples": 101,
+      "kl_divergence": 0.2931957937514963,
+      "js_divergence": 0.07706768601895059
     }
   }
 }

lnp_ml/modeling/predict.py

@@ -217,15 +217,31 @@ def test(
     """
     import json
     import numpy as np
+    from scipy.special import rel_entr
     from sklearn.metrics import (
         mean_squared_error,
         mean_absolute_error,
         r2_score,
         accuracy_score,
-        classification_report,
+        precision_score,
+        recall_score,
+        f1_score,
     )
     from lnp_ml.modeling.trainer import validate
     
+    def kl_divergence(p: np.ndarray, q: np.ndarray, eps: float = 1e-10) -> float:
+        """计算 KL 散度 KL(p || q)"""
+        p = np.clip(p, eps, 1.0)
+        q = np.clip(q, eps, 1.0)
+        return float(np.sum(rel_entr(p, q), axis=-1).mean())
+    
+    def js_divergence(p: np.ndarray, q: np.ndarray, eps: float = 1e-10) -> float:
+        """计算 JS 散度"""
+        p = np.clip(p, eps, 1.0)
+        q = np.clip(q, eps, 1.0)
+        m = 0.5 * (p + q)
+        return float(0.5 * (np.sum(rel_entr(p, m), axis=-1) + np.sum(rel_entr(q, m), axis=-1)).mean())
+    
     logger.info(f"Using device: {device}")
     device_obj = torch.device(device)
     
@@ -287,6 +303,9 @@ def test(
             y_pred = np.array(predictions["pdi"])[mask]
             results["detailed_metrics"]["pdi"] = {
                 "accuracy": float(accuracy_score(y_true, y_pred)),
+                "precision": float(precision_score(y_true, y_pred, average="macro", zero_division=0)),
+                "recall": float(recall_score(y_true, y_pred, average="macro", zero_division=0)),
+                "f1": float(f1_score(y_true, y_pred, average="macro", zero_division=0)),
             }
     
     # 分类指标：EE
@@ -299,6 +318,9 @@ def test(
             y_pred = np.array(predictions["ee"])[mask]
             results["detailed_metrics"]["ee"] = {
                 "accuracy": float(accuracy_score(y_true, y_pred)),
+                "precision": float(precision_score(y_true, y_pred, average="macro", zero_division=0)),
+                "recall": float(recall_score(y_true, y_pred, average="macro", zero_division=0)),
+                "f1": float(f1_score(y_true, y_pred, average="macro", zero_division=0)),
             }
     
     # 分类指标：toxic
@@ -309,6 +331,28 @@ def test(
             y_pred = np.array(predictions["toxic"])[mask.values]
             results["detailed_metrics"]["toxic"] = {
                 "accuracy": float(accuracy_score(y_true, y_pred)),
+                "precision": float(precision_score(y_true, y_pred, average="macro", zero_division=0)),
+                "recall": float(recall_score(y_true, y_pred, average="macro", zero_division=0)),
+                "f1": float(f1_score(y_true, y_pred, average="macro", zero_division=0)),
+            }
+    
+    # 分布指标：biodist
+    biodist_cols = [
+        "Biodistribution_lymph_nodes", "Biodistribution_heart", "Biodistribution_liver",
+        "Biodistribution_spleen", "Biodistribution_lung", "Biodistribution_kidney", "Biodistribution_muscle"
+    ]
+    if all(c in test_df.columns for c in biodist_cols):
+        biodist_true = test_df[biodist_cols].values
+        biodist_pred = np.array(predictions["biodist"])
+        # mask: 有效样本是 sum > 0 且无 NaN
+        mask = (biodist_true.sum(axis=1) > 0) & (~np.isnan(biodist_true).any(axis=1))
+        if mask.any():
+            y_true = biodist_true[mask]
+            y_pred = biodist_pred[mask]
+            results["detailed_metrics"]["biodist"] = {
+                "n_samples": int(mask.sum()),
+                "kl_divergence": kl_divergence(y_true, y_pred),
+                "js_divergence": js_divergence(y_true, y_pred),
             }
     
     # 打印结果

lnp_ml/modeling/pretrain_cv.py

@@ -271,7 +271,7 @@ def create_model(
 
 @app.command()
 def main(
-    data_dir: Path = PROCESSED_DATA_DIR / "cv",
+    data_dir: Path = PROCESSED_DATA_DIR / "pretrain_cv",
     output_dir: Path = MODELS_DIR / "pretrain_cv",
     # 模型参数
     d_model: int = 256,
@@ -322,7 +322,7 @@ def main(
     
     if not fold_dirs:
         logger.error(f"No fold_* directories found in {data_dir}")
-        logger.info("Please run 'make data_cv' first to process CV data.")
+        logger.info("Please run 'make data_pretrain_cv' first to process CV data.")
         raise typer.Exit(1)
     
     logger.info(f"Found {len(fold_dirs)} folds: {[d.name for d in fold_dirs]}")
@@ -464,7 +464,7 @@ def main(
 
 @app.command()
 def test(
-    data_dir: Path = PROCESSED_DATA_DIR / "cv",
+    data_dir: Path = PROCESSED_DATA_DIR / "pretrain_cv",
     model_dir: Path = MODELS_DIR / "pretrain_cv",
     output_path: Path = MODELS_DIR / "pretrain_cv" / "test_results.json",
     batch_size: int = 64,

lnp_ml/modeling/train_cv.py

@@ -0,0 +1,720 @@
+"""Cross-Validation 训练脚本：在 5-fold 内部数据上进行多任务训练"""
+
+import json
+from pathlib import Path
+from typing import Dict, List, Optional, Union
+
+import numpy as np
+import pandas as pd
+import torch
+import torch.nn as nn
+from torch.utils.data import DataLoader
+from loguru import logger
+from tqdm import tqdm
+import typer
+
+from lnp_ml.config import MODELS_DIR, PROCESSED_DATA_DIR
+from lnp_ml.dataset import LNPDataset, collate_fn
+from lnp_ml.modeling.models import LNPModel, LNPModelWithoutMPNN
+from lnp_ml.modeling.trainer import (
+    train_epoch,
+    validate,
+    EarlyStopping,
+    LossWeights,
+)
+
+
+# MPNN ensemble 默认路径
+DEFAULT_MPNN_ENSEMBLE_DIR = MODELS_DIR / "mpnn" / "all_amine_split_for_LiON"
+
+
+def find_mpnn_ensemble_paths(base_dir: Path = DEFAULT_MPNN_ENSEMBLE_DIR) -> List[str]:
+    """自动查找 MPNN ensemble 的 model.pt 文件。"""
+    model_paths = sorted(base_dir.glob("cv_*/fold_*/model_*/model.pt"))
+    if not model_paths:
+        raise FileNotFoundError(f"No model.pt files found in {base_dir}")
+    return [str(p) for p in model_paths]
+
+
+app = typer.Typer()
+
+
+def create_model(
+    d_model: int = 256,
+    num_heads: int = 8,
+    n_attn_layers: int = 4,
+    fusion_strategy: str = "attention",
+    head_hidden_dim: int = 128,
+    dropout: float = 0.1,
+    mpnn_checkpoint: Optional[str] = None,
+    mpnn_ensemble_paths: Optional[List[str]] = None,
+    mpnn_device: str = "cpu",
+) -> Union[LNPModel, LNPModelWithoutMPNN]:
+    """创建模型（支持可选的 MPNN encoder）"""
+    use_mpnn = mpnn_checkpoint is not None or mpnn_ensemble_paths is not None
+    
+    if use_mpnn:
+        return LNPModel(
+            d_model=d_model,
+            num_heads=num_heads,
+            n_attn_layers=n_attn_layers,
+            fusion_strategy=fusion_strategy,
+            head_hidden_dim=head_hidden_dim,
+            dropout=dropout,
+            mpnn_checkpoint=mpnn_checkpoint,
+            mpnn_ensemble_paths=mpnn_ensemble_paths,
+            mpnn_device=mpnn_device,
+        )
+    else:
+        return LNPModelWithoutMPNN(
+            d_model=d_model,
+            num_heads=num_heads,
+            n_attn_layers=n_attn_layers,
+            fusion_strategy=fusion_strategy,
+            head_hidden_dim=head_hidden_dim,
+            dropout=dropout,
+        )
+
+
+def train_fold(
+    fold_idx: int,
+    train_loader: DataLoader,
+    val_loader: DataLoader,
+    model: nn.Module,
+    device: torch.device,
+    output_dir: Path,
+    lr: float = 1e-4,
+    weight_decay: float = 1e-5,
+    epochs: int = 100,
+    patience: int = 15,
+    loss_weights: Optional[LossWeights] = None,
+    config: Optional[Dict] = None,
+) -> Dict:
+    """训练单个 fold"""
+    logger.info(f"\n{'='*60}")
+    logger.info(f"Training Fold {fold_idx}")
+    logger.info(f"{'='*60}")
+    
+    model = model.to(device)
+    optimizer = torch.optim.AdamW(model.parameters(), lr=lr, weight_decay=weight_decay)
+    scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(
+        optimizer, mode="min", factor=0.5, patience=5
+    )
+    early_stopping = EarlyStopping(patience=patience)
+    
+    history = {"train": [], "val": []}
+    best_val_loss = float("inf")
+    best_state = None
+    
+    for epoch in range(epochs):
+        # Train
+        train_metrics = train_epoch(model, train_loader, optimizer, device, loss_weights)
+        
+        # Validate
+        val_metrics = validate(model, val_loader, device, loss_weights)
+        
+        current_lr = optimizer.param_groups[0]["lr"]
+        
+        # Log
+        logger.info(
+            f"Fold {fold_idx} Epoch {epoch+1}/{epochs} | "
+            f"Train Loss: {train_metrics['loss']:.4f} | "
+            f"Val Loss: {val_metrics['loss']:.4f} | "
+            f"LR: {current_lr:.2e}"
+        )
+        
+        history["train"].append(train_metrics)
+        history["val"].append(val_metrics)
+        
+        # Learning rate scheduling
+        scheduler.step(val_metrics["loss"])
+        
+        # Save best model
+        if val_metrics["loss"] < best_val_loss:
+            best_val_loss = val_metrics["loss"]
+            best_state = {k: v.cpu().clone() for k, v in model.state_dict().items()}
+            logger.info(f"  -> New best model (val_loss={best_val_loss:.4f})")
+        
+        # Early stopping
+        if early_stopping(val_metrics["loss"]):
+            logger.info(f"Early stopping at epoch {epoch+1}")
+            break
+    
+    # 保存最佳模型
+    fold_output_dir = output_dir / f"fold_{fold_idx}"
+    fold_output_dir.mkdir(parents=True, exist_ok=True)
+    
+    checkpoint_path = fold_output_dir / "model.pt"
+    torch.save({
+        "model_state_dict": best_state,
+        "config": config,
+        "best_val_loss": best_val_loss,
+        "fold_idx": fold_idx,
+    }, checkpoint_path)
+    logger.success(f"Saved fold {fold_idx} model to {checkpoint_path}")
+    
+    # 保存训练历史
+    history_path = fold_output_dir / "history.json"
+    with open(history_path, "w") as f:
+        json.dump(history, f, indent=2)
+    
+    return {
+        "fold_idx": fold_idx,
+        "best_val_loss": best_val_loss,
+        "epochs_trained": len(history["train"]),
+        "final_train_loss": history["train"][-1]["loss"] if history["train"] else 0,
+    }
+
+
+@app.command()
+def main(
+    data_dir: Path = PROCESSED_DATA_DIR / "cv",
+    output_dir: Path = MODELS_DIR / "finetune_cv",
+    # 模型参数
+    d_model: int = 256,
+    num_heads: int = 8,
+    n_attn_layers: int = 4,
+    fusion_strategy: str = "attention",
+    head_hidden_dim: int = 128,
+    dropout: float = 0.1,
+    # MPNN 参数（可选）
+    use_mpnn: bool = False,
+    mpnn_checkpoint: Optional[str] = None,
+    mpnn_ensemble_paths: Optional[str] = None,
+    mpnn_device: str = "cpu",
+    # 训练参数
+    batch_size: int = 32,
+    lr: float = 1e-4,
+    weight_decay: float = 1e-5,
+    epochs: int = 100,
+    patience: int = 15,
+    # 预训练权重加载
+    init_from_pretrain: Optional[Path] = None,
+    load_delivery_head: bool = True,
+    freeze_backbone: bool = False,
+    # 设备
+    device: str = "cuda" if torch.cuda.is_available() else "cpu",
+):
+    """
+    基于 Cross-Validation 训练 LNP 模型（多任务）。
+    
+    在 5-fold 内部数据上训练 5 个模型。
+    
+    使用 --use-mpnn 启用 MPNN encoder。
+    使用 --init-from-pretrain 从预训练 checkpoint 初始化。
+    使用 --freeze-backbone 冻结 backbone，只训练 heads。
+    """
+    logger.info(f"Using device: {device}")
+    device = torch.device(device)
+    
+    # 查找所有 fold 目录
+    fold_dirs = sorted([d for d in data_dir.iterdir() if d.is_dir() and d.name.startswith("fold_")])
+    
+    if not fold_dirs:
+        logger.error(f"No fold_* directories found in {data_dir}")
+        logger.info("Please run 'make data_cv' first to process CV data.")
+        raise typer.Exit(1)
+    
+    logger.info(f"Found {len(fold_dirs)} folds: {[d.name for d in fold_dirs]}")
+    
+    output_dir.mkdir(parents=True, exist_ok=True)
+    
+    # 解析 MPNN 配置
+    ensemble_paths_list = None
+    if mpnn_ensemble_paths:
+        ensemble_paths_list = mpnn_ensemble_paths.split(",")
+    elif use_mpnn and mpnn_checkpoint is None:
+        logger.info(f"Auto-detecting MPNN ensemble from {DEFAULT_MPNN_ENSEMBLE_DIR}")
+        ensemble_paths_list = find_mpnn_ensemble_paths()
+        logger.info(f"Found {len(ensemble_paths_list)} MPNN models")
+    
+    enable_mpnn = mpnn_checkpoint is not None or ensemble_paths_list is not None
+    
+    # 模型配置
+    config = {
+        "d_model": d_model,
+        "num_heads": num_heads,
+        "n_attn_layers": n_attn_layers,
+        "fusion_strategy": fusion_strategy,
+        "head_hidden_dim": head_hidden_dim,
+        "dropout": dropout,
+        "use_mpnn": enable_mpnn,
+        "lr": lr,
+        "weight_decay": weight_decay,
+        "batch_size": batch_size,
+        "epochs": epochs,
+        "patience": patience,
+        "init_from_pretrain": str(init_from_pretrain) if init_from_pretrain else None,
+        "freeze_backbone": freeze_backbone,
+    }
+    
+    # 保存配置
+    config_path = output_dir / "config.json"
+    with open(config_path, "w") as f:
+        json.dump(config, f, indent=2)
+    logger.info(f"Saved config to {config_path}")
+    
+    # 加载预训练权重（如果指定）
+    pretrain_state = None
+    if init_from_pretrain is not None:
+        logger.info(f"Loading pretrain weights from {init_from_pretrain}")
+        checkpoint = torch.load(init_from_pretrain, map_location="cpu")
+        pretrain_config = checkpoint.get("config", {})
+        if pretrain_config.get("d_model") != d_model:
+            logger.warning(
+                f"d_model mismatch: pretrain={pretrain_config.get('d_model')}, "
+                f"current={d_model}. Skipping pretrain loading."
+            )
+        else:
+            pretrain_state = checkpoint["model_state_dict"]
+    
+    # 训练每个 fold
+    fold_results = []
+    
+    for fold_dir in tqdm(fold_dirs, desc="Training folds"):
+        fold_idx = int(fold_dir.name.split("_")[1])
+        
+        # 加载数据
+        train_df = pd.read_parquet(fold_dir / "train.parquet")
+        val_df = pd.read_parquet(fold_dir / "val.parquet")
+        
+        logger.info(f"\nFold {fold_idx}: train={len(train_df)}, val={len(val_df)}")
+        
+        # 创建 Dataset 和 DataLoader
+        train_dataset = LNPDataset(train_df)
+        val_dataset = LNPDataset(val_df)
+        
+        train_loader = DataLoader(
+            train_dataset, batch_size=batch_size, shuffle=True, collate_fn=collate_fn
+        )
+        val_loader = DataLoader(
+            val_dataset, batch_size=batch_size, shuffle=False, collate_fn=collate_fn
+        )
+        
+        # 创建新模型（每个 fold 独立初始化）
+        model = create_model(
+            d_model=d_model,
+            num_heads=num_heads,
+            n_attn_layers=n_attn_layers,
+            fusion_strategy=fusion_strategy,
+            head_hidden_dim=head_hidden_dim,
+            dropout=dropout,
+            mpnn_checkpoint=mpnn_checkpoint,
+            mpnn_ensemble_paths=ensemble_paths_list,
+            mpnn_device=device.type,
+        )
+        
+        # 加载预训练权重
+        if pretrain_state is not None:
+            model.load_pretrain_weights(
+                pretrain_state_dict=pretrain_state,
+                load_delivery_head=load_delivery_head,
+                strict=False,
+            )
+            logger.info(f"Loaded pretrain weights (backbone + delivery_head={load_delivery_head})")
+        
+        # 冻结 backbone（如果指定）
+        if freeze_backbone:
+            frozen_count = 0
+            for name, param in model.named_parameters():
+                if name.startswith(("token_projector.", "cross_attention.", "fusion.")):
+                    param.requires_grad = False
+                    frozen_count += 1
+            logger.info(f"Frozen {frozen_count} parameter tensors")
+        
+        # 打印模型信息（仅第一个 fold）
+        if fold_idx == 0:
+            n_params_total = sum(p.numel() for p in model.parameters())
+            n_params_trainable = sum(p.numel() for p in model.parameters() if p.requires_grad)
+            logger.info(f"Model parameters: {n_params_total:,} total, {n_params_trainable:,} trainable")
+        
+        # 训练
+        result = train_fold(
+            fold_idx=fold_idx,
+            train_loader=train_loader,
+            val_loader=val_loader,
+            model=model,
+            device=device,
+            output_dir=output_dir,
+            lr=lr,
+            weight_decay=weight_decay,
+            epochs=epochs,
+            patience=patience,
+            config=config,
+        )
+        fold_results.append(result)
+    
+    # 汇总结果
+    logger.info("\n" + "=" * 60)
+    logger.info("CROSS-VALIDATION TRAINING COMPLETE")
+    logger.info("=" * 60)
+    
+    val_losses = [r["best_val_loss"] for r in fold_results]
+    
+    logger.info(f"\n[Per-Fold Results]")
+    for r in fold_results:
+        logger.info(
+            f"  Fold {r['fold_idx']}: "
+            f"Val Loss={r['best_val_loss']:.4f}, "
+            f"Epochs={r['epochs_trained']}"
+        )
+    
+    logger.info(f"\n[Summary Statistics]")
+    logger.info(f"  Val Loss: {np.mean(val_losses):.4f} ± {np.std(val_losses):.4f}")
+    
+    # 保存 CV 结果
+    cv_results = {
+        "fold_results": fold_results,
+        "summary": {
+            "val_loss_mean": float(np.mean(val_losses)),
+            "val_loss_std": float(np.std(val_losses)),
+        },
+        "config": config,
+    }
+    
+    results_path = output_dir / "cv_results.json"
+    with open(results_path, "w") as f:
+        json.dump(cv_results, f, indent=2)
+    logger.success(f"Saved CV results to {results_path}")
+
+
+@app.command()
+def test(
+    data_dir: Path = PROCESSED_DATA_DIR / "cv",
+    model_dir: Path = MODELS_DIR / "finetune_cv",
+    output_path: Path = MODELS_DIR / "finetune_cv" / "test_results.json",
+    batch_size: int = 64,
+    device: str = "cuda" if torch.cuda.is_available() else "cpu",
+):
+    """
+    在测试集上评估 CV 训练的模型。
+    
+    使用每个 fold 的模型在对应的测试集上评估，然后汇总结果。
+    """
+    from scipy.special import rel_entr
+    from sklearn.metrics import (
+        mean_squared_error,
+        mean_absolute_error,
+        r2_score,
+        accuracy_score,
+        precision_score,
+        recall_score,
+        f1_score,
+    )
+    
+    def kl_divergence(p: np.ndarray, q: np.ndarray, eps: float = 1e-10) -> float:
+        """计算 KL 散度 KL(p || q)"""
+        p = np.clip(p, eps, 1.0)
+        q = np.clip(q, eps, 1.0)
+        return float(np.sum(rel_entr(p, q), axis=-1).mean())
+    
+    def js_divergence(p: np.ndarray, q: np.ndarray, eps: float = 1e-10) -> float:
+        """计算 JS 散度"""
+        p = np.clip(p, eps, 1.0)
+        q = np.clip(q, eps, 1.0)
+        m = 0.5 * (p + q)
+        return float(0.5 * (np.sum(rel_entr(p, m), axis=-1) + np.sum(rel_entr(q, m), axis=-1)).mean())
+    
+    logger.info(f"Using device: {device}")
+    device = torch.device(device)
+    
+    # 查找所有 fold 目录
+    fold_dirs = sorted([d for d in data_dir.iterdir() if d.is_dir() and d.name.startswith("fold_")])
+    
+    if not fold_dirs:
+        logger.error(f"No fold_* directories found in {data_dir}")
+        raise typer.Exit(1)
+    
+    logger.info(f"Found {len(fold_dirs)} folds")
+    
+    fold_results = []
+    # 用于汇总所有 fold 的预测
+    all_preds = {
+        "size": [], "delivery": [], "pdi": [], "ee": [], "toxic": [], "biodist": []
+    }
+    all_targets = {
+        "size": [], "delivery": [], "pdi": [], "ee": [], "toxic": [], "biodist": []
+    }
+    
+    for fold_dir in tqdm(fold_dirs, desc="Evaluating folds"):
+        fold_idx = int(fold_dir.name.split("_")[1])
+        model_path = model_dir / f"fold_{fold_idx}" / "model.pt"
+        test_path = fold_dir / "test.parquet"
+        
+        if not model_path.exists():
+            logger.warning(f"Fold {fold_idx}: model not found at {model_path}, skipping")
+            continue
+        
+        if not test_path.exists():
+            logger.warning(f"Fold {fold_idx}: test data not found at {test_path}, skipping")
+            continue
+        
+        # 加载模型
+        checkpoint = torch.load(model_path, map_location=device)
+        config = checkpoint["config"]
+        
+        use_mpnn = config.get("use_mpnn", False)
+        
+        # 总是重新查找 MPNN 路径
+        if use_mpnn:
+            mpnn_paths = find_mpnn_ensemble_paths()
+        else:
+            mpnn_paths = None
+        
+        model = create_model(
+            d_model=config["d_model"],
+            num_heads=config["num_heads"],
+            n_attn_layers=config["n_attn_layers"],
+            fusion_strategy=config["fusion_strategy"],
+            head_hidden_dim=config["head_hidden_dim"],
+            dropout=config["dropout"],
+            mpnn_ensemble_paths=mpnn_paths,
+            mpnn_device=device.type,
+        )
+        model.load_state_dict(checkpoint["model_state_dict"])
+        model = model.to(device)
+        model.eval()
+        
+        # 加载测试数据
+        test_df = pd.read_parquet(test_path)
+        test_dataset = LNPDataset(test_df)
+        test_loader = DataLoader(
+            test_dataset, batch_size=batch_size, shuffle=False, collate_fn=collate_fn
+        )
+        
+        # 收集当前 fold 的预测
+        fold_preds = {k: [] for k in all_preds.keys()}
+        fold_targets = {k: [] for k in all_targets.keys()}
+        
+        with torch.no_grad():
+            pbar = tqdm(test_loader, desc=f"Fold {fold_idx} [Test]", leave=False)
+            for batch in pbar:
+                smiles = batch["smiles"]
+                tabular = {k: v.to(device) for k, v in batch["tabular"].items()}
+                targets = batch["targets"]
+                masks = batch["mask"]
+                
+                outputs = model(smiles, tabular)
+                
+                # Size
+                if "size" in masks and masks["size"].any():
+                    mask = masks["size"]
+                    fold_preds["size"].extend(
+                        outputs["size"].squeeze(-1)[mask].cpu().numpy().tolist()
+                    )
+                    fold_targets["size"].extend(
+                        targets["size"][mask].cpu().numpy().tolist()
+                    )
+                
+                # Delivery
+                if "delivery" in masks and masks["delivery"].any():
+                    mask = masks["delivery"]
+                    fold_preds["delivery"].extend(
+                        outputs["delivery"].squeeze(-1)[mask].cpu().numpy().tolist()
+                    )
+                    fold_targets["delivery"].extend(
+                        targets["delivery"][mask].cpu().numpy().tolist()
+                    )
+                
+                # PDI (classification)
+                if "pdi" in masks and masks["pdi"].any():
+                    mask = masks["pdi"]
+                    pdi_preds = outputs["pdi"][mask].argmax(dim=-1).cpu().numpy()
+                    pdi_targets = targets["pdi"][mask].cpu().numpy()
+                    fold_preds["pdi"].extend(pdi_preds.tolist())
+                    fold_targets["pdi"].extend(pdi_targets.tolist())
+                
+                # EE (classification)
+                if "ee" in masks and masks["ee"].any():
+                    mask = masks["ee"]
+                    ee_preds = outputs["ee"][mask].argmax(dim=-1).cpu().numpy()
+                    ee_targets = targets["ee"][mask].cpu().numpy()
+                    fold_preds["ee"].extend(ee_preds.tolist())
+                    fold_targets["ee"].extend(ee_targets.tolist())
+                
+                # Toxic (classification)
+                if "toxic" in masks and masks["toxic"].any():
+                    mask = masks["toxic"]
+                    toxic_preds = outputs["toxic"][mask].argmax(dim=-1).cpu().numpy()
+                    toxic_targets = targets["toxic"][mask].cpu().numpy().astype(int)
+                    fold_preds["toxic"].extend(toxic_preds.tolist())
+                    fold_targets["toxic"].extend(toxic_targets.tolist())
+                
+                # Biodist (distribution)
+                if "biodist" in masks and masks["biodist"].any():
+                    mask = masks["biodist"]
+                    biodist_preds = outputs["biodist"][mask].cpu().numpy()
+                    biodist_targets = targets["biodist"][mask].cpu().numpy()
+                    fold_preds["biodist"].extend(biodist_preds.tolist())
+                    fold_targets["biodist"].extend(biodist_targets.tolist())
+        
+        # 计算当前 fold 的指标
+        fold_metrics = {"fold_idx": fold_idx, "n_samples": len(test_df)}
+        
+        # 回归任务指标
+        for task in ["size", "delivery"]:
+            if fold_preds[task]:
+                p = np.array(fold_preds[task])
+                t = np.array(fold_targets[task])
+                fold_metrics[task] = {
+                    "n": len(p),
+                    "rmse": float(np.sqrt(mean_squared_error(t, p))),
+                    "mae": float(mean_absolute_error(t, p)),
+                    "r2": float(r2_score(t, p)),
+                }
+        
+        # 分类任务指标
+        for task in ["pdi", "ee", "toxic"]:
+            if fold_preds[task]:
+                p = np.array(fold_preds[task])
+                t = np.array(fold_targets[task])
+                fold_metrics[task] = {
+                    "n": len(p),
+                    "accuracy": float(accuracy_score(t, p)),
+                    "precision": float(precision_score(t, p, average="macro", zero_division=0)),
+                    "recall": float(recall_score(t, p, average="macro", zero_division=0)),
+                    "f1": float(f1_score(t, p, average="macro", zero_division=0)),
+                }
+        
+        # 分布任务指标
+        if fold_preds["biodist"]:
+            p = np.array(fold_preds["biodist"])
+            t = np.array(fold_targets["biodist"])
+            fold_metrics["biodist"] = {
+                "n": len(p),
+                "kl_divergence": kl_divergence(t, p),
+                "js_divergence": js_divergence(t, p),
+            }
+        
+        fold_results.append(fold_metrics)
+        
+        # 汇总到全局
+        for task in all_preds.keys():
+            all_preds[task].extend(fold_preds[task])
+            all_targets[task].extend(fold_targets[task])
+        
+        # 打印当前 fold 结果
+        log_parts = [f"Fold {fold_idx}: n={len(test_df)}"]
+        for task in ["delivery", "size"]:
+            if task in fold_metrics and isinstance(fold_metrics[task], dict):
+                log_parts.append(f"{task}_RMSE={fold_metrics[task]['rmse']:.4f}")
+                log_parts.append(f"{task}_R²={fold_metrics[task]['r2']:.4f}")
+        for task in ["pdi", "ee", "toxic"]:
+            if task in fold_metrics and isinstance(fold_metrics[task], dict):
+                log_parts.append(f"{task}_acc={fold_metrics[task]['accuracy']:.4f}")
+                log_parts.append(f"{task}_f1={fold_metrics[task]['f1']:.4f}")
+        if "biodist" in fold_metrics and isinstance(fold_metrics["biodist"], dict):
+            log_parts.append(f"biodist_KL={fold_metrics['biodist']['kl_divergence']:.4f}")
+            log_parts.append(f"biodist_JS={fold_metrics['biodist']['js_divergence']:.4f}")
+        logger.info(", ".join(log_parts))
+    
+    # 计算跨 fold 汇总统计
+    summary_stats = {}
+    for task in ["size", "delivery"]:
+        rmses = [r[task]["rmse"] for r in fold_results if task in r and isinstance(r[task], dict)]
+        r2s = [r[task]["r2"] for r in fold_results if task in r and isinstance(r[task], dict)]
+        if rmses:
+            summary_stats[task] = {
+                "rmse_mean": float(np.mean(rmses)),
+                "rmse_std": float(np.std(rmses)),
+                "r2_mean": float(np.mean(r2s)),
+                "r2_std": float(np.std(r2s)),
+            }
+    
+    for task in ["pdi", "ee", "toxic"]:
+        accs = [r[task]["accuracy"] for r in fold_results if task in r and isinstance(r[task], dict)]
+        f1s = [r[task]["f1"] for r in fold_results if task in r and isinstance(r[task], dict)]
+        if accs:
+            summary_stats[task] = {
+                "accuracy_mean": float(np.mean(accs)),
+                "accuracy_std": float(np.std(accs)),
+                "f1_mean": float(np.mean(f1s)),
+                "f1_std": float(np.std(f1s)),
+            }
+    
+    # 分布任务汇总
+    kls = [r["biodist"]["kl_divergence"] for r in fold_results if "biodist" in r and isinstance(r["biodist"], dict)]
+    jss = [r["biodist"]["js_divergence"] for r in fold_results if "biodist" in r and isinstance(r["biodist"], dict)]
+    if kls:
+        summary_stats["biodist"] = {
+            "kl_mean": float(np.mean(kls)),
+            "kl_std": float(np.std(kls)),
+            "js_mean": float(np.mean(jss)),
+            "js_std": float(np.std(jss)),
+        }
+    
+    # 计算整体 pooled 指标
+    overall = {}
+    for task in ["size", "delivery"]:
+        if all_preds[task]:
+            p = np.array(all_preds[task])
+            t = np.array(all_targets[task])
+            overall[task] = {
+                "n_samples": len(p),
+                "mse": float(mean_squared_error(t, p)),
+                "rmse": float(np.sqrt(mean_squared_error(t, p))),
+                "mae": float(mean_absolute_error(t, p)),
+                "r2": float(r2_score(t, p)),
+            }
+    
+    for task in ["pdi", "ee", "toxic"]:
+        if all_preds[task]:
+            p = np.array(all_preds[task])
+            t = np.array(all_targets[task])
+            overall[task] = {
+                "n_samples": len(p),
+                "accuracy": float(accuracy_score(t, p)),
+                "precision": float(precision_score(t, p, average="macro", zero_division=0)),
+                "recall": float(recall_score(t, p, average="macro", zero_division=0)),
+                "f1": float(f1_score(t, p, average="macro", zero_division=0)),
+            }
+    
+    # 分布任务
+    if all_preds["biodist"]:
+        p = np.array(all_preds["biodist"])
+        t = np.array(all_targets["biodist"])
+        overall["biodist"] = {
+            "n_samples": len(p),
+            "kl_divergence": kl_divergence(t, p),
+            "js_divergence": js_divergence(t, p),
+        }
+    
+    # 打印汇总结果
+    logger.info("\n" + "=" * 60)
+    logger.info("CV TEST EVALUATION RESULTS")
+    logger.info("=" * 60)
+    
+    logger.info(f"\n[Summary Statistics (across {len(fold_results)} folds)]")
+    for task, stats in summary_stats.items():
+        if "rmse_mean" in stats:
+            logger.info(f"  {task}: RMSE={stats['rmse_mean']:.4f}±{stats['rmse_std']:.4f}, R²={stats['r2_mean']:.4f}±{stats['r2_std']:.4f}")
+        elif "accuracy_mean" in stats:
+            logger.info(f"  {task}: Accuracy={stats['accuracy_mean']:.4f}±{stats['accuracy_std']:.4f}, F1={stats['f1_mean']:.4f}±{stats['f1_std']:.4f}")
+        elif "kl_mean" in stats:
+            logger.info(f"  {task}: KL={stats['kl_mean']:.4f}±{stats['kl_std']:.4f}, JS={stats['js_mean']:.4f}±{stats['js_std']:.4f}")
+    
+    logger.info(f"\n[Overall (all samples pooled)]")
+    for task, metrics in overall.items():
+        if "rmse" in metrics:
+            logger.info(f"  {task} (n={metrics['n_samples']}): RMSE={metrics['rmse']:.4f}, MAE={metrics['mae']:.4f}, R²={metrics['r2']:.4f}")
+        elif "accuracy" in metrics:
+            logger.info(f"  {task} (n={metrics['n_samples']}): Accuracy={metrics['accuracy']:.4f}, Precision={metrics['precision']:.4f}, Recall={metrics['recall']:.4f}, F1={metrics['f1']:.4f}")
+        elif "kl_divergence" in metrics:
+            logger.info(f"  {task} (n={metrics['n_samples']}): KL={metrics['kl_divergence']:.4f}, JS={metrics['js_divergence']:.4f}")
+    
+    # 保存结果
+    results = {
+        "fold_results": fold_results,
+        "summary_stats": summary_stats,
+        "overall": overall,
+    }
+    
+    output_path.parent.mkdir(parents=True, exist_ok=True)
+    with open(output_path, "w") as f:
+        json.dump(results, f, indent=2)
+    logger.success(f"\nSaved test results to {output_path}")
+
+
+if __name__ == "__main__":
+    app()
+

models/finetune_cv/config.json

@@ -0,0 +1,16 @@
+{
+  "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": null,
+  "freeze_backbone": false
+}

models/finetune_cv/cv_results.json

@@ -0,0 +1,54 @@
+{
+  "fold_results": [
+    {
+      "fold_idx": 0,
+      "best_val_loss": 5.7676777839660645,
+      "epochs_trained": 24,
+      "final_train_loss": 1.4942118644714355
+    },
+    {
+      "fold_idx": 1,
+      "best_val_loss": 8.418675899505615,
+      "epochs_trained": 20,
+      "final_train_loss": 1.4902493238449097
+    },
+    {
+      "fold_idx": 2,
+      "best_val_loss": 3.5122547830854143,
+      "epochs_trained": 25,
+      "final_train_loss": 1.7609570423762004
+    },
+    {
+      "fold_idx": 3,
+      "best_val_loss": 3.165306806564331,
+      "epochs_trained": 21,
+      "final_train_loss": 2.0073827385902403
+    },
+    {
+      "fold_idx": 4,
+      "best_val_loss": 2.996154228846232,
+      "epochs_trained": 18,
+      "final_train_loss": 1.9732873006300493
+    }
+  ],
+  "summary": {
+    "val_loss_mean": 4.772013900393532,
+    "val_loss_std": 2.0790222989111475
+  },
+  "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": null,
+    "freeze_backbone": false
+  }
+}

models/finetune_cv/fold_0/history.json

@@ -0,0 +1,510 @@
+{
+  "train": [
+    {
+      "loss": 17.65872812271118,
+      "loss_size": 12.601411867141724,
+      "loss_pdi": 1.3666706204414367,
+      "loss_ee": 1.0830313920974732,
+      "loss_delivery": 0.5962779104709626,
+      "loss_biodist": 1.3918164849281311,
+      "loss_toxic": 0.6195200622081757
+    },
+    {
+      "loss": 5.925264883041382,
+      "loss_size": 1.8580878481268883,
+      "loss_pdi": 1.1011681258678436,
+      "loss_ee": 0.971046245098114,
+      "loss_delivery": 0.5075224950909615,
+      "loss_biodist": 1.1051940202713013,
+      "loss_toxic": 0.38224617540836336
+    },
+    {
+      "loss": 3.4781792640686033,
+      "loss_size": 0.23610344529151917,
+      "loss_pdi": 0.8137399554252625,
+      "loss_ee": 0.9135127127170563,
+      "loss_delivery": 0.4596045270562172,
+      "loss_biodist": 0.8695587992668152,
+      "loss_toxic": 0.18565986081957817
+    },
+    {
+      "loss": 2.9488561868667604,
+      "loss_size": 0.23130029290914536,
+      "loss_pdi": 0.644479614496231,
+      "loss_ee": 0.8721524059772492,
+      "loss_delivery": 0.4146773874759674,
+      "loss_biodist": 0.646893310546875,
+      "loss_toxic": 0.13935319259762763
+    },
+    {
+      "loss": 2.6432241678237913,
+      "loss_size": 0.16843259893357754,
+      "loss_pdi": 0.5857123643159866,
+      "loss_ee": 0.8315786123275757,
+      "loss_delivery": 0.4049036353826523,
+      "loss_biodist": 0.5410242855548859,
+      "loss_toxic": 0.11157271154224872
+    },
+    {
+      "loss": 2.461507487297058,
+      "loss_size": 0.18602822050452233,
+      "loss_pdi": 0.5872043997049332,
+      "loss_ee": 0.8179578661918641,
+      "loss_delivery": 0.32779163047671317,
+      "loss_biodist": 0.45097417533397677,
+      "loss_toxic": 0.09155115596950054
+    },
+    {
+      "loss": 2.3792370796203612,
+      "loss_size": 0.2090120367705822,
+      "loss_pdi": 0.5358257800340652,
+      "loss_ee": 0.8088949501514435,
+      "loss_delivery": 0.3434994474053383,
+      "loss_biodist": 0.40993946194648745,
+      "loss_toxic": 0.07206540685147048
+    },
+    {
+      "loss": 2.207099366188049,
+      "loss_size": 0.1589151345193386,
+      "loss_pdi": 0.5283154606819153,
+      "loss_ee": 0.7723551869392395,
+      "loss_delivery": 0.35645291954278946,
+      "loss_biodist": 0.3404483631253242,
+      "loss_toxic": 0.05061229532584548
+    },
+    {
+      "loss": 2.1428971529006957,
+      "loss_size": 0.19335013553500174,
+      "loss_pdi": 0.5021985083818435,
+      "loss_ee": 0.7642539083957672,
+      "loss_delivery": 0.31821031123399734,
+      "loss_biodist": 0.32588216066360476,
+      "loss_toxic": 0.03900211993604898
+    },
+    {
+      "loss": 1.9874909400939942,
+      "loss_size": 0.1736245721578598,
+      "loss_pdi": 0.46206980347633364,
+      "loss_ee": 0.7373365700244904,
+      "loss_delivery": 0.29703493416309357,
+      "loss_biodist": 0.2863417714834213,
+      "loss_toxic": 0.031083252932876348
+    },
+    {
+      "loss": 1.9297520160675048,
+      "loss_size": 0.1635374441742897,
+      "loss_pdi": 0.4737923800945282,
+      "loss_ee": 0.7171129584312439,
+      "loss_delivery": 0.28808903992176055,
+      "loss_biodist": 0.25874830335378646,
+      "loss_toxic": 0.028471904620528222
+    },
+    {
+      "loss": 1.8647576332092286,
+      "loss_size": 0.14790172204375268,
+      "loss_pdi": 0.4427785277366638,
+      "loss_ee": 0.7089932143688202,
+      "loss_delivery": 0.30143058970570563,
+      "loss_biodist": 0.24234647750854493,
+      "loss_toxic": 0.021307120053097605
+    },
+    {
+      "loss": 1.7996623039245605,
+      "loss_size": 0.1429538145661354,
+      "loss_pdi": 0.45114057660102846,
+      "loss_ee": 0.681770408153534,
+      "loss_delivery": 0.2735618159174919,
+      "loss_biodist": 0.2338838443160057,
+      "loss_toxic": 0.01635184111073613
+    },
+    {
+      "loss": 1.7303769707679748,
+      "loss_size": 0.13725369721651076,
+      "loss_pdi": 0.43492600619792937,
+      "loss_ee": 0.6648448914289474,
+      "loss_delivery": 0.2714417055249214,
+      "loss_biodist": 0.20898159295320512,
+      "loss_toxic": 0.012929048202931882
+    },
+    {
+      "loss": 1.702065145969391,
+      "loss_size": 0.1783118523657322,
+      "loss_pdi": 0.4118753671646118,
+      "loss_ee": 0.640222480893135,
+      "loss_delivery": 0.2610591858625412,
+      "loss_biodist": 0.20058825612068176,
+      "loss_toxic": 0.01000797227025032
+    },
+    {
+      "loss": 1.6243244886398316,
+      "loss_size": 0.1371393844485283,
+      "loss_pdi": 0.3978125751018524,
+      "loss_ee": 0.6315451622009277,
+      "loss_delivery": 0.2618463449180126,
+      "loss_biodist": 0.18574777096509934,
+      "loss_toxic": 0.010233237966895103
+    },
+    {
+      "loss": 1.645119547843933,
+      "loss_size": 0.13622624576091766,
+      "loss_pdi": 0.4013118803501129,
+      "loss_ee": 0.639850401878357,
+      "loss_delivery": 0.2615354858338833,
+      "loss_biodist": 0.19717498123645782,
+      "loss_toxic": 0.009020529384724797
+    },
+    {
+      "loss": 1.5792422771453858,
+      "loss_size": 0.12063037976622581,
+      "loss_pdi": 0.40477685928344725,
+      "loss_ee": 0.6168571084737777,
+      "loss_delivery": 0.23877703920006751,
+      "loss_biodist": 0.1887524366378784,
+      "loss_toxic": 0.009448455832898616
+    },
+    {
+      "loss": 1.5701380014419555,
+      "loss_size": 0.12370488420128822,
+      "loss_pdi": 0.3944096490740776,
+      "loss_ee": 0.6204680263996124,
+      "loss_delivery": 0.2499392546713352,
+      "loss_biodist": 0.1741167649626732,
+      "loss_toxic": 0.00749938020016998
+    },
+    {
+      "loss": 1.5445807576179504,
+      "loss_size": 0.12085893377661705,
+      "loss_pdi": 0.4022176057100296,
+      "loss_ee": 0.6029386401176453,
+      "loss_delivery": 0.2460342638194561,
+      "loss_biodist": 0.16601160615682603,
+      "loss_toxic": 0.006519717467017472
+    },
+    {
+      "loss": 1.4764926195144654,
+      "loss_size": 0.11393929794430732,
+      "loss_pdi": 0.3614879995584488,
+      "loss_ee": 0.5874974340200424,
+      "loss_delivery": 0.2382828861474991,
+      "loss_biodist": 0.168075630068779,
+      "loss_toxic": 0.00720936032012105
+    },
+    {
+      "loss": 1.4663256525993347,
+      "loss_size": 0.10480817258358002,
+      "loss_pdi": 0.3699364930391312,
+      "loss_ee": 0.591068571805954,
+      "loss_delivery": 0.23481545299291612,
+      "loss_biodist": 0.1582734301686287,
+      "loss_toxic": 0.007423530006781221
+    },
+    {
+      "loss": 1.4797919273376465,
+      "loss_size": 0.11906521767377853,
+      "loss_pdi": 0.3831163257360458,
+      "loss_ee": 0.5810098886489868,
+      "loss_delivery": 0.22465722858905793,
+      "loss_biodist": 0.16469249799847602,
+      "loss_toxic": 0.007250743336044252
+    },
+    {
+      "loss": 1.4942118644714355,
+      "loss_size": 0.11249525547027588,
+      "loss_pdi": 0.3718418627977371,
+      "loss_ee": 0.5973137259483338,
+      "loss_delivery": 0.23963096588850022,
+      "loss_biodist": 0.16598810032010078,
+      "loss_toxic": 0.006941930414177478
+    }
+  ],
+  "val": [
+    {
+      "loss": 13.683866500854492,
+      "loss_size": 5.657964706420898,
+      "loss_pdi": 1.1590962409973145,
+      "loss_ee": 1.0155898332595825,
+      "loss_delivery": 4.1429033279418945,
+      "loss_biodist": 1.128843069076538,
+      "loss_toxic": 0.579468846321106,
+      "acc_pdi": 0.7407407407407407,
+      "acc_ee": 0.6296296296296297,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 7.161172866821289,
+      "loss_size": 0.1799931526184082,
+      "loss_pdi": 0.8303115963935852,
+      "loss_ee": 0.942605197429657,
+      "loss_delivery": 3.986294984817505,
+      "loss_biodist": 1.022797703742981,
+      "loss_toxic": 0.19917015731334686,
+      "acc_pdi": 0.8888888888888888,
+      "acc_ee": 0.5925925925925926,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 6.554836273193359,
+      "loss_size": 0.04609166830778122,
+      "loss_pdi": 0.4924769997596741,
+      "loss_ee": 0.965587317943573,
+      "loss_delivery": 3.978637933731079,
+      "loss_biodist": 1.0135102272033691,
+      "loss_toxic": 0.05853228643536568,
+      "acc_pdi": 0.8888888888888888,
+      "acc_ee": 0.5925925925925926,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 6.843129634857178,
+      "loss_size": 0.07650057226419449,
+      "loss_pdi": 0.43551138043403625,
+      "loss_ee": 0.9353340864181519,
+      "loss_delivery": 4.557775974273682,
+      "loss_biodist": 0.7909315228462219,
+      "loss_toxic": 0.047075945883989334,
+      "acc_pdi": 0.8888888888888888,
+      "acc_ee": 0.5925925925925926,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 6.711758613586426,
+      "loss_size": 0.04316325858235359,
+      "loss_pdi": 0.41873815655708313,
+      "loss_ee": 1.0096691846847534,
+      "loss_delivery": 4.517927169799805,
+      "loss_biodist": 0.6788683533668518,
+      "loss_toxic": 0.04339226707816124,
+      "acc_pdi": 0.8888888888888888,
+      "acc_ee": 0.5925925925925926,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 6.905030250549316,
+      "loss_size": 0.045318666845560074,
+      "loss_pdi": 0.38593801856040955,
+      "loss_ee": 1.0019593238830566,
+      "loss_delivery": 4.807835578918457,
+      "loss_biodist": 0.6247215867042542,
+      "loss_toxic": 0.039257097989320755,
+      "acc_pdi": 0.8888888888888888,
+      "acc_ee": 0.5185185185185185,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 6.417820930480957,
+      "loss_size": 0.05034356936812401,
+      "loss_pdi": 0.4149726331233978,
+      "loss_ee": 0.9869357943534851,
+      "loss_delivery": 4.405001640319824,
+      "loss_biodist": 0.533240556716919,
+      "loss_toxic": 0.02732720412313938,
+      "acc_pdi": 0.8888888888888888,
+      "acc_ee": 0.5925925925925926,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 6.608631610870361,
+      "loss_size": 0.05222579464316368,
+      "loss_pdi": 0.4375711679458618,
+      "loss_ee": 1.0041171312332153,
+      "loss_delivery": 4.578192234039307,
+      "loss_biodist": 0.5125234723091125,
+      "loss_toxic": 0.02400212176144123,
+      "acc_pdi": 0.8888888888888888,
+      "acc_ee": 0.5185185185185185,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 5.7676777839660645,
+      "loss_size": 0.09589201211929321,
+      "loss_pdi": 0.3261733949184418,
+      "loss_ee": 0.9482788443565369,
+      "loss_delivery": 3.856112003326416,
+      "loss_biodist": 0.5298716425895691,
+      "loss_toxic": 0.011350298300385475,
+      "acc_pdi": 0.8888888888888888,
+      "acc_ee": 0.5555555555555556,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 6.920990943908691,
+      "loss_size": 0.05388057231903076,
+      "loss_pdi": 0.39705148339271545,
+      "loss_ee": 0.990842878818512,
+      "loss_delivery": 5.025243282318115,
+      "loss_biodist": 0.4346938133239746,
+      "loss_toxic": 0.019278930500149727,
+      "acc_pdi": 0.8888888888888888,
+      "acc_ee": 0.5555555555555556,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 5.798760890960693,
+      "loss_size": 0.09857960045337677,
+      "loss_pdi": 0.33329641819000244,
+      "loss_ee": 0.9614524245262146,
+      "loss_delivery": 4.000489711761475,
+      "loss_biodist": 0.39874210953712463,
+      "loss_toxic": 0.006200834643095732,
+      "acc_pdi": 0.8888888888888888,
+      "acc_ee": 0.5555555555555556,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 6.575327396392822,
+      "loss_size": 0.054699063301086426,
+      "loss_pdi": 0.33702051639556885,
+      "loss_ee": 0.9436452388763428,
+      "loss_delivery": 4.817119121551514,
+      "loss_biodist": 0.41582298278808594,
+      "loss_toxic": 0.007020703982561827,
+      "acc_pdi": 0.8888888888888888,
+      "acc_ee": 0.5555555555555556,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 5.989306449890137,
+      "loss_size": 0.09009546041488647,
+      "loss_pdi": 0.3044246733188629,
+      "loss_ee": 1.0130207538604736,
+      "loss_delivery": 4.140576362609863,
+      "loss_biodist": 0.4378862977027893,
+      "loss_toxic": 0.0033026484306901693,
+      "acc_pdi": 0.8888888888888888,
+      "acc_ee": 0.5555555555555556,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 6.383339881896973,
+      "loss_size": 0.1530081033706665,
+      "loss_pdi": 0.29700207710266113,
+      "loss_ee": 0.9943283796310425,
+      "loss_delivery": 4.564785480499268,
+      "loss_biodist": 0.37085360288619995,
+      "loss_toxic": 0.003362649120390415,
+      "acc_pdi": 0.8888888888888888,
+      "acc_ee": 0.5185185185185185,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 6.233416557312012,
+      "loss_size": 0.1473817676305771,
+      "loss_pdi": 0.2754640281200409,
+      "loss_ee": 0.9803684949874878,
+      "loss_delivery": 4.443488597869873,
+      "loss_biodist": 0.38424524664878845,
+      "loss_toxic": 0.0024682653602212667,
+      "acc_pdi": 0.8888888888888888,
+      "acc_ee": 0.5925925925925926,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 6.094257354736328,
+      "loss_size": 0.10127364844083786,
+      "loss_pdi": 0.2960923910140991,
+      "loss_ee": 1.0121080875396729,
+      "loss_delivery": 4.2689008712768555,
+      "loss_biodist": 0.4132467210292816,
+      "loss_toxic": 0.0026356647722423077,
+      "acc_pdi": 0.8888888888888888,
+      "acc_ee": 0.5555555555555556,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 6.0315470695495605,
+      "loss_size": 0.13236114382743835,
+      "loss_pdi": 0.29554903507232666,
+      "loss_ee": 0.9912998080253601,
+      "loss_delivery": 4.2240777015686035,
+      "loss_biodist": 0.3861069977283478,
+      "loss_toxic": 0.002152827335521579,
+      "acc_pdi": 0.8888888888888888,
+      "acc_ee": 0.5925925925925926,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 6.13291597366333,
+      "loss_size": 0.10603927820920944,
+      "loss_pdi": 0.30880627036094666,
+      "loss_ee": 1.0417256355285645,
+      "loss_delivery": 4.337818622589111,
+      "loss_biodist": 0.33598482608795166,
+      "loss_toxic": 0.002541647758334875,
+      "acc_pdi": 0.8888888888888888,
+      "acc_ee": 0.5555555555555556,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 5.918347358703613,
+      "loss_size": 0.11423231661319733,
+      "loss_pdi": 0.2779754102230072,
+      "loss_ee": 1.023812174797058,
+      "loss_delivery": 4.137387275695801,
+      "loss_biodist": 0.36283549666404724,
+      "loss_toxic": 0.002104171784594655,
+      "acc_pdi": 0.8888888888888888,
+      "acc_ee": 0.5185185185185185,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 6.354115962982178,
+      "loss_size": 0.1212025135755539,
+      "loss_pdi": 0.2848753333091736,
+      "loss_ee": 1.031553030014038,
+      "loss_delivery": 4.554471969604492,
+      "loss_biodist": 0.3598195016384125,
+      "loss_toxic": 0.0021939175203442574,
+      "acc_pdi": 0.8518518518518519,
+      "acc_ee": 0.5555555555555556,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 5.881389141082764,
+      "loss_size": 0.1030399352312088,
+      "loss_pdi": 0.2791188657283783,
+      "loss_ee": 1.0205037593841553,
+      "loss_delivery": 4.111578464508057,
+      "loss_biodist": 0.3657107949256897,
+      "loss_toxic": 0.0014369667042046785,
+      "acc_pdi": 0.8518518518518519,
+      "acc_ee": 0.5555555555555556,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 6.1028852462768555,
+      "loss_size": 0.10241233557462692,
+      "loss_pdi": 0.300007700920105,
+      "loss_ee": 1.0756882429122925,
+      "loss_delivery": 4.258440971374512,
+      "loss_biodist": 0.3646480441093445,
+      "loss_toxic": 0.0016881312476471066,
+      "acc_pdi": 0.8888888888888888,
+      "acc_ee": 0.5185185185185185,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 6.128824234008789,
+      "loss_size": 0.1437627077102661,
+      "loss_pdi": 0.29325851798057556,
+      "loss_ee": 1.0818182229995728,
+      "loss_delivery": 4.236568450927734,
+      "loss_biodist": 0.3719424605369568,
+      "loss_toxic": 0.001474093529395759,
+      "acc_pdi": 0.8888888888888888,
+      "acc_ee": 0.5185185185185185,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 6.055476188659668,
+      "loss_size": 0.13312266767024994,
+      "loss_pdi": 0.28571468591690063,
+      "loss_ee": 1.066524624824524,
+      "loss_delivery": 4.214193820953369,
+      "loss_biodist": 0.35442692041397095,
+      "loss_toxic": 0.0014939504908397794,
+      "acc_pdi": 0.8518518518518519,
+      "acc_ee": 0.5185185185185185,
+      "acc_toxic": 1.0
+    }
+  ]
+}

models/finetune_cv/fold_1/history.json

@@ -0,0 +1,426 @@
+{
+  "train": [
+    {
+      "loss": 21.963233947753906,
+      "loss_size": 16.82633171081543,
+      "loss_pdi": 1.2230936765670777,
+      "loss_ee": 1.0703922033309936,
+      "loss_delivery": 1.0690569162368775,
+      "loss_biodist": 1.1534382343292235,
+      "loss_toxic": 0.6209211587905884
+    },
+    {
+      "loss": 13.145495796203614,
+      "loss_size": 8.676862716674805,
+      "loss_pdi": 1.0655134558677672,
+      "loss_ee": 0.8999906063079834,
+      "loss_delivery": 0.8303895950317383,
+      "loss_biodist": 1.122723388671875,
+      "loss_toxic": 0.5500160694122315
+    },
+    {
+      "loss": 7.351448345184326,
+      "loss_size": 3.415665292739868,
+      "loss_pdi": 0.8565655469894409,
+      "loss_ee": 0.7837236523628235,
+      "loss_delivery": 0.8804788589477539,
+      "loss_biodist": 1.011645209789276,
+      "loss_toxic": 0.40336963534355164
+    },
+    {
+      "loss": 4.39948205947876,
+      "loss_size": 0.9713698267936707,
+      "loss_pdi": 0.6989291191101075,
+      "loss_ee": 0.6805540442466735,
+      "loss_delivery": 0.7624839186668396,
+      "loss_biodist": 0.9798830866813659,
+      "loss_toxic": 0.3062621414661407
+    },
+    {
+      "loss": 3.375754451751709,
+      "loss_size": 0.24608666747808455,
+      "loss_pdi": 0.5557448148727417,
+      "loss_ee": 0.6684133768081665,
+      "loss_delivery": 0.7611681580543518,
+      "loss_biodist": 0.919653308391571,
+      "loss_toxic": 0.22468801140785216
+    },
+    {
+      "loss": 2.9307605743408205,
+      "loss_size": 0.1106911577284336,
+      "loss_pdi": 0.5004462003707886,
+      "loss_ee": 0.6227471172809601,
+      "loss_delivery": 0.6758030593395233,
+      "loss_biodist": 0.8190896153450012,
+      "loss_toxic": 0.20198351740837098
+    },
+    {
+      "loss": 2.731675052642822,
+      "loss_size": 0.13740637749433518,
+      "loss_pdi": 0.4836215674877167,
+      "loss_ee": 0.5896897256374359,
+      "loss_delivery": 0.5866121172904968,
+      "loss_biodist": 0.7556124567985535,
+      "loss_toxic": 0.17873288169503213
+    },
+    {
+      "loss": 2.4887039184570314,
+      "loss_size": 0.12009606957435608,
+      "loss_pdi": 0.4361336886882782,
+      "loss_ee": 0.597134268283844,
+      "loss_delivery": 0.5648026138544082,
+      "loss_biodist": 0.6326960444450378,
+      "loss_toxic": 0.13784122765064238
+    },
+    {
+      "loss": 2.1680586099624635,
+      "loss_size": 0.12401954531669616,
+      "loss_pdi": 0.40216060280799865,
+      "loss_ee": 0.5528951227664948,
+      "loss_delivery": 0.42899617552757263,
+      "loss_biodist": 0.5442585527896882,
+      "loss_toxic": 0.1157285787165165
+    },
+    {
+      "loss": 2.1059993267059327,
+      "loss_size": 0.13299092650413513,
+      "loss_pdi": 0.38143277168273926,
+      "loss_ee": 0.5274551689624787,
+      "loss_delivery": 0.47739412933588027,
+      "loss_biodist": 0.4953398108482361,
+      "loss_toxic": 0.0913865402340889
+    },
+    {
+      "loss": 1.9570286750793457,
+      "loss_size": 0.1426382303237915,
+      "loss_pdi": 0.38325140476226804,
+      "loss_ee": 0.49524895548820497,
+      "loss_delivery": 0.42715947031974794,
+      "loss_biodist": 0.4287752747535706,
+      "loss_toxic": 0.07995530962944031
+    },
+    {
+      "loss": 1.8469573497772216,
+      "loss_size": 0.14165955781936646,
+      "loss_pdi": 0.36685559153556824,
+      "loss_ee": 0.4988661766052246,
+      "loss_delivery": 0.36661114990711213,
+      "loss_biodist": 0.39747334718704225,
+      "loss_toxic": 0.07549156174063683
+    },
+    {
+      "loss": 1.6980855226516725,
+      "loss_size": 0.11332993358373641,
+      "loss_pdi": 0.350938493013382,
+      "loss_ee": 0.47553136944770813,
+      "loss_delivery": 0.30049399137496946,
+      "loss_biodist": 0.3953311860561371,
+      "loss_toxic": 0.062460555136203764
+    },
+    {
+      "loss": 1.743706512451172,
+      "loss_size": 0.12467859983444214,
+      "loss_pdi": 0.3706244468688965,
+      "loss_ee": 0.4802402436733246,
+      "loss_delivery": 0.36484516113996507,
+      "loss_biodist": 0.3557030588388443,
+      "loss_toxic": 0.04761496149003506
+    },
+    {
+      "loss": 1.7470735549926757,
+      "loss_size": 0.10215002745389938,
+      "loss_pdi": 0.3553147315979004,
+      "loss_ee": 0.4548905730247498,
+      "loss_delivery": 0.4480485826730728,
+      "loss_biodist": 0.3265932142734528,
+      "loss_toxic": 0.06007647253572941
+    },
+    {
+      "loss": 1.7687433004379272,
+      "loss_size": 0.10528398901224137,
+      "loss_pdi": 0.35497177839279176,
+      "loss_ee": 0.4946293234825134,
+      "loss_delivery": 0.44853600263595583,
+      "loss_biodist": 0.3113987982273102,
+      "loss_toxic": 0.053923492506146434
+    },
+    {
+      "loss": 1.573294997215271,
+      "loss_size": 0.11145550012588501,
+      "loss_pdi": 0.33941014409065245,
+      "loss_ee": 0.42823529839515684,
+      "loss_delivery": 0.34292849004268644,
+      "loss_biodist": 0.3095307767391205,
+      "loss_toxic": 0.04173475466668606
+    },
+    {
+      "loss": 1.482050108909607,
+      "loss_size": 0.13211917281150817,
+      "loss_pdi": 0.31831381320953367,
+      "loss_ee": 0.4258797198534012,
+      "loss_delivery": 0.26612227857112886,
+      "loss_biodist": 0.30344046354293824,
+      "loss_toxic": 0.03617466017603874
+    },
+    {
+      "loss": 1.5079625368118286,
+      "loss_size": 0.1129397764801979,
+      "loss_pdi": 0.3118207275867462,
+      "loss_ee": 0.4255594819784164,
+      "loss_delivery": 0.30544502288103104,
+      "loss_biodist": 0.31328115463256834,
+      "loss_toxic": 0.03891638442873955
+    },
+    {
+      "loss": 1.4902493238449097,
+      "loss_size": 0.09879767149686813,
+      "loss_pdi": 0.333440762758255,
+      "loss_ee": 0.430321592092514,
+      "loss_delivery": 0.3070627197623253,
+      "loss_biodist": 0.28984564244747163,
+      "loss_toxic": 0.030780918896198273
+    }
+  ],
+  "val": [
+    {
+      "loss": 24.328961690266926,
+      "loss_size": 15.672358830769857,
+      "loss_pdi": 1.268057902654012,
+      "loss_ee": 1.0569811463356018,
+      "loss_delivery": 4.617272272706032,
+      "loss_biodist": 1.0806464751561482,
+      "loss_toxic": 0.633646289507548,
+      "acc_pdi": 0.6105263157894737,
+      "acc_ee": 0.5894736842105263,
+      "acc_toxic": 0.8939393939393939
+    },
+    {
+      "loss": 17.03301429748535,
+      "loss_size": 8.649629751841227,
+      "loss_pdi": 1.165820797284444,
+      "loss_ee": 0.9437925020853678,
+      "loss_delivery": 4.629274984200795,
+      "loss_biodist": 1.0683060089747112,
+      "loss_toxic": 0.5761909882227579,
+      "acc_pdi": 0.6105263157894737,
+      "acc_ee": 0.6736842105263158,
+      "acc_toxic": 0.8939393939393939
+    },
+    {
+      "loss": 11.635572751363119,
+      "loss_size": 3.504341204961141,
+      "loss_pdi": 1.0855141083399455,
+      "loss_ee": 0.8674407601356506,
+      "loss_delivery": 4.705501407384872,
+      "loss_biodist": 1.0376905004183452,
+      "loss_toxic": 0.43508487939834595,
+      "acc_pdi": 0.6105263157894737,
+      "acc_ee": 0.6736842105263158,
+      "acc_toxic": 0.8939393939393939
+    },
+    {
+      "loss": 9.058362166086832,
+      "loss_size": 1.0461570421854656,
+      "loss_pdi": 1.070031762123108,
+      "loss_ee": 0.8463932275772095,
+      "loss_delivery": 4.791346887747447,
+      "loss_biodist": 0.9781110286712646,
+      "loss_toxic": 0.32632239659627277,
+      "acc_pdi": 0.6105263157894737,
+      "acc_ee": 0.6736842105263158,
+      "acc_toxic": 0.8939393939393939
+    },
+    {
+      "loss": 8.418675899505615,
+      "loss_size": 0.3764288102587064,
+      "loss_pdi": 1.0916812817255657,
+      "loss_ee": 0.8714254101117452,
+      "loss_delivery": 4.8696667949358625,
+      "loss_biodist": 0.9307892719904581,
+      "loss_toxic": 0.278684730331103,
+      "acc_pdi": 0.6105263157894737,
+      "acc_ee": 0.6736842105263158,
+      "acc_toxic": 0.8939393939393939
+    },
+    {
+      "loss": 8.51748021443685,
+      "loss_size": 0.33909208327531815,
+      "loss_pdi": 1.103804111480713,
+      "loss_ee": 0.8707688599824905,
+      "loss_delivery": 5.0624091029167175,
+      "loss_biodist": 0.8743396997451782,
+      "loss_toxic": 0.2670666699608167,
+      "acc_pdi": 0.6105263157894737,
+      "acc_ee": 0.6736842105263158,
+      "acc_toxic": 0.8939393939393939
+    },
+    {
+      "loss": 8.701509237289429,
+      "loss_size": 0.38883806640903157,
+      "loss_pdi": 1.0901564558347066,
+      "loss_ee": 0.8219001442193985,
+      "loss_delivery": 5.329233412941297,
+      "loss_biodist": 0.808117667833964,
+      "loss_toxic": 0.2632630293567975,
+      "acc_pdi": 0.6105263157894737,
+      "acc_ee": 0.6736842105263158,
+      "acc_toxic": 0.8939393939393939
+    },
+    {
+      "loss": 8.602253516515097,
+      "loss_size": 0.399209912866354,
+      "loss_pdi": 1.035650501648585,
+      "loss_ee": 0.8119546920061111,
+      "loss_delivery": 5.297288862367471,
+      "loss_biodist": 0.8003136416276296,
+      "loss_toxic": 0.2578362462421258,
+      "acc_pdi": 0.6105263157894737,
+      "acc_ee": 0.6736842105263158,
+      "acc_toxic": 0.8939393939393939
+    },
+    {
+      "loss": 8.610430796941122,
+      "loss_size": 0.3884888291358948,
+      "loss_pdi": 0.9680223266283671,
+      "loss_ee": 0.8063104202349981,
+      "loss_delivery": 5.504999443888664,
+      "loss_biodist": 0.7153328458468119,
+      "loss_toxic": 0.22727691816786924,
+      "acc_pdi": 0.6105263157894737,
+      "acc_ee": 0.6736842105263158,
+      "acc_toxic": 0.8939393939393939
+    },
+    {
+      "loss": 8.894750118255615,
+      "loss_size": 0.38015256201227504,
+      "loss_pdi": 0.9849910040696462,
+      "loss_ee": 0.8192636320988337,
+      "loss_delivery": 5.8433875640233355,
+      "loss_biodist": 0.6525928874810537,
+      "loss_toxic": 0.21436312049627304,
+      "acc_pdi": 0.6105263157894737,
+      "acc_ee": 0.6736842105263158,
+      "acc_toxic": 0.8939393939393939
+    },
+    {
+      "loss": 8.684672435124716,
+      "loss_size": 0.39142270882924396,
+      "loss_pdi": 0.9926454623540243,
+      "loss_ee": 0.8487897912661234,
+      "loss_delivery": 5.675399616360664,
+      "loss_biodist": 0.5763055086135864,
+      "loss_toxic": 0.2001086367915074,
+      "acc_pdi": 0.6105263157894737,
+      "acc_ee": 0.6736842105263158,
+      "acc_toxic": 0.8939393939393939
+    },
+    {
+      "loss": 8.468807538350424,
+      "loss_size": 0.37035099665323895,
+      "loss_pdi": 0.9933059811592102,
+      "loss_ee": 0.8365495651960373,
+      "loss_delivery": 5.39086152613163,
+      "loss_biodist": 0.6555034021536509,
+      "loss_toxic": 0.22223659542699656,
+      "acc_pdi": 0.6105263157894737,
+      "acc_ee": 0.6736842105263158,
+      "acc_toxic": 0.8939393939393939
+    },
+    {
+      "loss": 8.48760732014974,
+      "loss_size": 0.3547621878484885,
+      "loss_pdi": 1.008083571990331,
+      "loss_ee": 0.8507340376575788,
+      "loss_delivery": 5.329072058200836,
+      "loss_biodist": 0.7051869928836823,
+      "loss_toxic": 0.23976873668531576,
+      "acc_pdi": 0.6105263157894737,
+      "acc_ee": 0.6736842105263158,
+      "acc_toxic": 0.8939393939393939
+    },
+    {
+      "loss": 8.534255782763163,
+      "loss_size": 0.35214799270033836,
+      "loss_pdi": 1.0083338419596355,
+      "loss_ee": 0.8703259030977885,
+      "loss_delivery": 5.4809657235940294,
+      "loss_biodist": 0.6066243648529053,
+      "loss_toxic": 0.21585797673712173,
+      "acc_pdi": 0.6105263157894737,
+      "acc_ee": 0.6842105263157895,
+      "acc_toxic": 0.8939393939393939
+    },
+    {
+      "loss": 8.59092911084493,
+      "loss_size": 0.3520332872867584,
+      "loss_pdi": 0.9944024880727133,
+      "loss_ee": 0.8839219162861506,
+      "loss_delivery": 5.593439628680547,
+      "loss_biodist": 0.562449519832929,
+      "loss_toxic": 0.20468231476843357,
+      "acc_pdi": 0.6105263157894737,
+      "acc_ee": 0.6736842105263158,
+      "acc_toxic": 0.8939393939393939
+    },
+    {
+      "loss": 8.581690510114035,
+      "loss_size": 0.3468632685641448,
+      "loss_pdi": 1.0153752664724986,
+      "loss_ee": 0.884696863591671,
+      "loss_delivery": 5.548932209610939,
+      "loss_biodist": 0.576594889163971,
+      "loss_toxic": 0.20922777770708004,
+      "acc_pdi": 0.6105263157894737,
+      "acc_ee": 0.6736842105263158,
+      "acc_toxic": 0.8939393939393939
+    },
+    {
+      "loss": 8.60028068224589,
+      "loss_size": 0.34553587809205055,
+      "loss_pdi": 1.0314316948254902,
+      "loss_ee": 0.8696443388859431,
+      "loss_delivery": 5.513105024894078,
+      "loss_biodist": 0.6187789390484492,
+      "loss_toxic": 0.2217849005634586,
+      "acc_pdi": 0.6105263157894737,
+      "acc_ee": 0.6947368421052632,
+      "acc_toxic": 0.8939393939393939
+    },
+    {
+      "loss": 8.721842130025228,
+      "loss_size": 0.3432792164385319,
+      "loss_pdi": 1.044082870086034,
+      "loss_ee": 0.8888355021675428,
+      "loss_delivery": 5.590167284011841,
+      "loss_biodist": 0.6300752113262812,
+      "loss_toxic": 0.22540184513976178,
+      "acc_pdi": 0.6105263157894737,
+      "acc_ee": 0.6736842105263158,
+      "acc_toxic": 0.8939393939393939
+    },
+    {
+      "loss": 8.821967244148254,
+      "loss_size": 0.3423520748813947,
+      "loss_pdi": 1.0627215206623077,
+      "loss_ee": 0.9012102037668228,
+      "loss_delivery": 5.6443866689999895,
+      "loss_biodist": 0.6428664823373159,
+      "loss_toxic": 0.2284308553983768,
+      "acc_pdi": 0.6105263157894737,
+      "acc_ee": 0.6631578947368421,
+      "acc_toxic": 0.8939393939393939
+    },
+    {
+      "loss": 8.798149506251017,
+      "loss_size": 0.3439513569076856,
+      "loss_pdi": 1.074403668443362,
+      "loss_ee": 0.9037297517061234,
+      "loss_delivery": 5.62445667386055,
+      "loss_biodist": 0.6279164751370748,
+      "loss_toxic": 0.22369086369872093,
+      "acc_pdi": 0.6105263157894737,
+      "acc_ee": 0.6631578947368421,
+      "acc_toxic": 0.8939393939393939
+    }
+  ]
+}

models/finetune_cv/fold_2/history.json

@@ -0,0 +1,531 @@
+{
+  "train": [
+    {
+      "loss": 22.972569465637207,
+      "loss_size": 17.22947899500529,
+      "loss_pdi": 1.3510672052701314,
+      "loss_ee": 1.0506827433904011,
+      "loss_delivery": 1.5721548050642014,
+      "loss_biodist": 1.1304322481155396,
+      "loss_toxic": 0.6387530366579691
+    },
+    {
+      "loss": 12.718681335449219,
+      "loss_size": 7.335077285766602,
+      "loss_pdi": 1.198062241077423,
+      "loss_ee": 0.97108127673467,
+      "loss_delivery": 1.639556477467219,
+      "loss_biodist": 1.0746847093105316,
+      "loss_toxic": 0.50021959344546
+    },
+    {
+      "loss": 6.867454210917155,
+      "loss_size": 2.3153140544891357,
+      "loss_pdi": 1.0159071584542592,
+      "loss_ee": 0.853508859872818,
+      "loss_delivery": 1.2862873176733653,
+      "loss_biodist": 1.0416639745235443,
+      "loss_toxic": 0.35477257271607715
+    },
+    {
+      "loss": 4.856432318687439,
+      "loss_size": 0.5409951706727346,
+      "loss_pdi": 0.8652523259321848,
+      "loss_ee": 0.7771940131982168,
+      "loss_delivery": 1.413562481602033,
+      "loss_biodist": 0.9977987806002299,
+      "loss_toxic": 0.26162934054931003
+    },
+    {
+      "loss": 4.253215591112773,
+      "loss_size": 0.2641367167234421,
+      "loss_pdi": 0.739859402179718,
+      "loss_ee": 0.7256686190764109,
+      "loss_delivery": 1.4241955528656642,
+      "loss_biodist": 0.8935903211434683,
+      "loss_toxic": 0.2057649294535319
+    },
+    {
+      "loss": 3.8961705764134726,
+      "loss_size": 0.2962125514944394,
+      "loss_pdi": 0.682400623957316,
+      "loss_ee": 0.6820215880870819,
+      "loss_delivery": 1.2787245536843936,
+      "loss_biodist": 0.7812575101852417,
+      "loss_toxic": 0.1755537080268065
+    },
+    {
+      "loss": 3.4790991942087808,
+      "loss_size": 0.3047281603018443,
+      "loss_pdi": 0.6409291823705038,
+      "loss_ee": 0.6178905169169108,
+      "loss_delivery": 1.1121559316913288,
+      "loss_biodist": 0.6434484819571177,
+      "loss_toxic": 0.15994682783881822
+    },
+    {
+      "loss": 3.2075613339742026,
+      "loss_size": 0.3421506683031718,
+      "loss_pdi": 0.5879766543706259,
+      "loss_ee": 0.5811398377021154,
+      "loss_delivery": 1.0462109719713528,
+      "loss_biodist": 0.520307645201683,
+      "loss_toxic": 0.12977550799647966
+    },
+    {
+      "loss": 2.861353278160095,
+      "loss_size": 0.2742840300003688,
+      "loss_pdi": 0.5437282969554266,
+      "loss_ee": 0.5531725088755289,
+      "loss_delivery": 0.9213679246604443,
+      "loss_biodist": 0.4499489863713582,
+      "loss_toxic": 0.11885150956610839
+    },
+    {
+      "loss": 2.6909215847651162,
+      "loss_size": 0.23881135260065398,
+      "loss_pdi": 0.5229279547929764,
+      "loss_ee": 0.5285524874925613,
+      "loss_delivery": 0.8911051253477732,
+      "loss_biodist": 0.4015616128842036,
+      "loss_toxic": 0.10796305599311988
+    },
+    {
+      "loss": 2.5927247206370034,
+      "loss_size": 0.27356760079662007,
+      "loss_pdi": 0.5166990955670675,
+      "loss_ee": 0.5059170673290888,
+      "loss_delivery": 0.8377179056406021,
+      "loss_biodist": 0.3519642899433772,
+      "loss_toxic": 0.10685871541500092
+    },
+    {
+      "loss": 2.3971973856290183,
+      "loss_size": 0.2688147674004237,
+      "loss_pdi": 0.4851151605447133,
+      "loss_ee": 0.47870688637097675,
+      "loss_delivery": 0.7584750155607859,
+      "loss_biodist": 0.3166690344611804,
+      "loss_toxic": 0.08941652067005634
+    },
+    {
+      "loss": 2.2271180947621665,
+      "loss_size": 0.2559296215573947,
+      "loss_pdi": 0.467803418636322,
+      "loss_ee": 0.4819647620121638,
+      "loss_delivery": 0.6487737223505974,
+      "loss_biodist": 0.2930952211221059,
+      "loss_toxic": 0.079551310899357
+    },
+    {
+      "loss": 2.1467134952545166,
+      "loss_size": 0.2658323546250661,
+      "loss_pdi": 0.47287177046140033,
+      "loss_ee": 0.4580538024504979,
+      "loss_delivery": 0.6110207016269366,
+      "loss_biodist": 0.26590356479088467,
+      "loss_toxic": 0.07303123424450557
+    },
+    {
+      "loss": 2.0699684421221414,
+      "loss_size": 0.23655260602633157,
+      "loss_pdi": 0.46446068088213605,
+      "loss_ee": 0.43884341915448505,
+      "loss_delivery": 0.5945644030968348,
+      "loss_biodist": 0.26856863250335056,
+      "loss_toxic": 0.06697871504972379
+    },
+    {
+      "loss": 2.012367367744446,
+      "loss_size": 0.20358355715870857,
+      "loss_pdi": 0.44864421089490253,
+      "loss_ee": 0.4260970900456111,
+      "loss_delivery": 0.6111055202782154,
+      "loss_biodist": 0.24829111248254776,
+      "loss_toxic": 0.07464585608492295
+    },
+    {
+      "loss": 1.9354575673739116,
+      "loss_size": 0.19155597686767578,
+      "loss_pdi": 0.43001438677310944,
+      "loss_ee": 0.4029633104801178,
+      "loss_delivery": 0.5866967861851057,
+      "loss_biodist": 0.26284457246462506,
+      "loss_toxic": 0.06138256782044967
+    },
+    {
+      "loss": 1.9248821139335632,
+      "loss_size": 0.19836385796467462,
+      "loss_pdi": 0.43165912727514905,
+      "loss_ee": 0.4223821411530177,
+      "loss_delivery": 0.5774712382505337,
+      "loss_biodist": 0.23008103668689728,
+      "loss_toxic": 0.06492467441906531
+    },
+    {
+      "loss": 1.7986130317052205,
+      "loss_size": 0.1977602814634641,
+      "loss_pdi": 0.4213625093301137,
+      "loss_ee": 0.3969506522019704,
+      "loss_delivery": 0.4972396679222584,
+      "loss_biodist": 0.22815552850564322,
+      "loss_toxic": 0.05714430411656698
+    },
+    {
+      "loss": 1.8008437156677246,
+      "loss_size": 0.20143492271502814,
+      "loss_pdi": 0.4257240394751231,
+      "loss_ee": 0.3939937750498454,
+      "loss_delivery": 0.4996156108876069,
+      "loss_biodist": 0.22945881386597952,
+      "loss_toxic": 0.05061656702309847
+    },
+    {
+      "loss": 1.8123606244723003,
+      "loss_size": 0.23175274084011713,
+      "loss_pdi": 0.41065867245197296,
+      "loss_ee": 0.38645289838314056,
+      "loss_delivery": 0.5105274474869171,
+      "loss_biodist": 0.22759289046128592,
+      "loss_toxic": 0.04537593169758717
+    },
+    {
+      "loss": 1.85766206185023,
+      "loss_size": 0.2237167110045751,
+      "loss_pdi": 0.4198872745037079,
+      "loss_ee": 0.39036936064561206,
+      "loss_delivery": 0.5356200908621153,
+      "loss_biodist": 0.2327907457947731,
+      "loss_toxic": 0.055277835965777435
+    },
+    {
+      "loss": 1.7299150824546814,
+      "loss_size": 0.16866947089632353,
+      "loss_pdi": 0.40182044357061386,
+      "loss_ee": 0.37123599648475647,
+      "loss_delivery": 0.5183743331581354,
+      "loss_biodist": 0.22459317495425543,
+      "loss_toxic": 0.04522162117063999
+    },
+    {
+      "loss": 1.8115381598472595,
+      "loss_size": 0.21021889025966325,
+      "loss_pdi": 0.3938516428073247,
+      "loss_ee": 0.3856282929579417,
+      "loss_delivery": 0.5463737193495035,
+      "loss_biodist": 0.22467835744222006,
+      "loss_toxic": 0.05078731415172418
+    },
+    {
+      "loss": 1.7609570423762004,
+      "loss_size": 0.20672637100021043,
+      "loss_pdi": 0.3868243644634883,
+      "loss_ee": 0.3775654385487239,
+      "loss_delivery": 0.5160359914104143,
+      "loss_biodist": 0.22730938345193863,
+      "loss_toxic": 0.04649555139864484
+    }
+  ],
+  "val": [
+    {
+      "loss": 19.896042142595565,
+      "loss_size": 14.947636876787458,
+      "loss_pdi": 1.3514722074781145,
+      "loss_ee": 1.0372784308024816,
+      "loss_delivery": 0.5157596128327506,
+      "loss_biodist": 1.3665738276072912,
+      "loss_toxic": 0.6773212381771633,
+      "acc_pdi": 0.22564102564102564,
+      "acc_ee": 0.4512820512820513,
+      "acc_toxic": 0.7073170731707317
+    },
+    {
+      "loss": 10.277108192443848,
+      "loss_size": 5.728530270712716,
+      "loss_pdi": 1.2047701733452933,
+      "loss_ee": 1.013599353177207,
+      "loss_delivery": 0.5329383058207375,
+      "loss_biodist": 1.3288453817367554,
+      "loss_toxic": 0.4684244394302368,
+      "acc_pdi": 0.358974358974359,
+      "acc_ee": 0.4461538461538462,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 4.947442190987723,
+      "loss_size": 0.9055690084184919,
+      "loss_pdi": 0.9325166344642639,
+      "loss_ee": 1.071527932371412,
+      "loss_delivery": 0.5525430504764829,
+      "loss_biodist": 1.2514750446592058,
+      "loss_toxic": 0.23381062703473227,
+      "acc_pdi": 0.7076923076923077,
+      "acc_ee": 0.4205128205128205,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 4.072668756757464,
+      "loss_size": 0.18621658267719404,
+      "loss_pdi": 0.7640595691544669,
+      "loss_ee": 1.2155327456338065,
+      "loss_delivery": 0.5523517067943301,
+      "loss_biodist": 1.214196733066014,
+      "loss_toxic": 0.14031140506267548,
+      "acc_pdi": 0.7076923076923077,
+      "acc_ee": 0.4205128205128205,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 4.053883859089443,
+      "loss_size": 0.24764748875583922,
+      "loss_pdi": 0.6859285916600909,
+      "loss_ee": 1.3019903557641166,
+      "loss_delivery": 0.5384655041354043,
+      "loss_biodist": 1.191110406603132,
+      "loss_toxic": 0.08874151536396571,
+      "acc_pdi": 0.7076923076923077,
+      "acc_ee": 0.4205128205128205,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.9850601128169467,
+      "loss_size": 0.18243093735405377,
+      "loss_pdi": 0.6606386282614299,
+      "loss_ee": 1.2955879313605172,
+      "loss_delivery": 0.6222422846726009,
+      "loss_biodist": 1.1586603011403764,
+      "loss_toxic": 0.06550001353025436,
+      "acc_pdi": 0.7076923076923077,
+      "acc_ee": 0.4205128205128205,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.6691461631229947,
+      "loss_size": 0.18682856378810747,
+      "loss_pdi": 0.6505168399640492,
+      "loss_ee": 1.2279302733285087,
+      "loss_delivery": 0.5776888344969068,
+      "loss_biodist": 0.963392470564161,
+      "loss_toxic": 0.06278917672378677,
+      "acc_pdi": 0.7076923076923077,
+      "acc_ee": 0.4205128205128205,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.6420267650059293,
+      "loss_size": 0.1952320017984935,
+      "loss_pdi": 0.6510439600263324,
+      "loss_ee": 1.1954282522201538,
+      "loss_delivery": 0.7644538623946053,
+      "loss_biodist": 0.7812093198299408,
+      "loss_toxic": 0.05465935756053243,
+      "acc_pdi": 0.7076923076923077,
+      "acc_ee": 0.35384615384615387,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.707563672746931,
+      "loss_size": 0.2168926394411496,
+      "loss_pdi": 0.6468359615121569,
+      "loss_ee": 1.2361225570951189,
+      "loss_delivery": 0.8388645563806806,
+      "loss_biodist": 0.7232611009052822,
+      "loss_toxic": 0.04558686592749187,
+      "acc_pdi": 0.7076923076923077,
+      "acc_ee": 0.37435897435897436,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.5122547830854143,
+      "loss_size": 0.2614529473440988,
+      "loss_pdi": 0.6344352598701205,
+      "loss_ee": 1.2337199449539185,
+      "loss_delivery": 0.7557642417294639,
+      "loss_biodist": 0.5974612619195666,
+      "loss_toxic": 0.029421218537858555,
+      "acc_pdi": 0.7076923076923077,
+      "acc_ee": 0.38461538461538464,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 4.085699932915824,
+      "loss_size": 0.2200212436062949,
+      "loss_pdi": 0.6225023801837649,
+      "loss_ee": 1.1934180770601546,
+      "loss_delivery": 1.4556497505732946,
+      "loss_biodist": 0.5685178296906608,
+      "loss_toxic": 0.02559054403432778,
+      "acc_pdi": 0.7076923076923077,
+      "acc_ee": 0.4205128205128205,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.8555148329053606,
+      "loss_size": 0.25200862703578814,
+      "loss_pdi": 0.6161004496472222,
+      "loss_ee": 1.2155306509562902,
+      "loss_delivery": 1.1912458751882826,
+      "loss_biodist": 0.5578728743961879,
+      "loss_toxic": 0.022756420208939483,
+      "acc_pdi": 0.7128205128205128,
+      "acc_ee": 0.4153846153846154,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 4.016897848674229,
+      "loss_size": 0.24823468178510666,
+      "loss_pdi": 0.6098369508981705,
+      "loss_ee": 1.2048260143824987,
+      "loss_delivery": 1.3509162408964974,
+      "loss_biodist": 0.5822246244975499,
+      "loss_toxic": 0.020859350051198686,
+      "acc_pdi": 0.717948717948718,
+      "acc_ee": 0.4153846153846154,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.9899745328085765,
+      "loss_size": 0.27859273659331457,
+      "loss_pdi": 0.6051683936800275,
+      "loss_ee": 1.1875721216201782,
+      "loss_delivery": 1.369072552238192,
+      "loss_biodist": 0.5321473862443652,
+      "loss_toxic": 0.017421354805784568,
+      "acc_pdi": 0.717948717948718,
+      "acc_ee": 0.40512820512820513,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 4.112551552908761,
+      "loss_size": 0.23502862347023828,
+      "loss_pdi": 0.6127274334430695,
+      "loss_ee": 1.2102909428732735,
+      "loss_delivery": 1.5001615626471383,
+      "loss_biodist": 0.5411617543016162,
+      "loss_toxic": 0.013181165459432773,
+      "acc_pdi": 0.7025641025641025,
+      "acc_ee": 0.4205128205128205,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 4.031796966280256,
+      "loss_size": 0.27996559121779035,
+      "loss_pdi": 0.6218498295971325,
+      "loss_ee": 1.264663577079773,
+      "loss_delivery": 1.225053642477308,
+      "loss_biodist": 0.6304309921605247,
+      "loss_toxic": 0.009833223053387232,
+      "acc_pdi": 0.7128205128205128,
+      "acc_ee": 0.4153846153846154,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 4.108100175857544,
+      "loss_size": 0.2646343271647181,
+      "loss_pdi": 0.6244613996573857,
+      "loss_ee": 1.2785721336092268,
+      "loss_delivery": 1.2817653375012534,
+      "loss_biodist": 0.6491539776325226,
+      "loss_toxic": 0.009512946475297213,
+      "acc_pdi": 0.7025641025641025,
+      "acc_ee": 0.4,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 4.237571648188999,
+      "loss_size": 0.24608339369297028,
+      "loss_pdi": 0.6387959729347911,
+      "loss_ee": 1.263997495174408,
+      "loss_delivery": 1.3677963316440582,
+      "loss_biodist": 0.7113959235804421,
+      "loss_toxic": 0.00950257752888969,
+      "acc_pdi": 0.7025641025641025,
+      "acc_ee": 0.40512820512820513,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 4.677373443331037,
+      "loss_size": 0.25485377439430784,
+      "loss_pdi": 0.6646602579525539,
+      "loss_ee": 1.2957249539239066,
+      "loss_delivery": 1.609152581010546,
+      "loss_biodist": 0.8443670613425118,
+      "loss_toxic": 0.008614770535911833,
+      "acc_pdi": 0.7025641025641025,
+      "acc_ee": 0.4,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 4.661478791918073,
+      "loss_size": 0.22112409876925604,
+      "loss_pdi": 0.6840873499001775,
+      "loss_ee": 1.345719371523176,
+      "loss_delivery": 1.5750049437795366,
+      "loss_biodist": 0.8278610365731376,
+      "loss_toxic": 0.0076820029810603175,
+      "acc_pdi": 0.6974358974358974,
+      "acc_ee": 0.39487179487179486,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 4.7805344717843195,
+      "loss_size": 0.23743291412081038,
+      "loss_pdi": 0.6911627639617238,
+      "loss_ee": 1.3796877009528024,
+      "loss_delivery": 1.6191245743206568,
+      "loss_biodist": 0.8459444258894239,
+      "loss_toxic": 0.00718201809961881,
+      "acc_pdi": 0.7076923076923077,
+      "acc_ee": 0.4,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 5.050536905016218,
+      "loss_size": 0.2340430138366563,
+      "loss_pdi": 0.7009050281984466,
+      "loss_ee": 1.3757387569972448,
+      "loss_delivery": 1.760018629687173,
+      "loss_biodist": 0.9723425933292934,
+      "loss_toxic": 0.007489000846232686,
+      "acc_pdi": 0.7128205128205128,
+      "acc_ee": 0.39487179487179486,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 5.172625916344779,
+      "loss_size": 0.23549717558281763,
+      "loss_pdi": 0.6980976568801063,
+      "loss_ee": 1.357451047216143,
+      "loss_delivery": 1.893591480595725,
+      "loss_biodist": 0.9802024279321943,
+      "loss_toxic": 0.0077861944612647805,
+      "acc_pdi": 0.717948717948718,
+      "acc_ee": 0.38974358974358975,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 5.048826490129743,
+      "loss_size": 0.2420537450483867,
+      "loss_pdi": 0.7013733184763363,
+      "loss_ee": 1.353548560823713,
+      "loss_delivery": 1.7931698901312692,
+      "loss_biodist": 0.9509889696325574,
+      "loss_toxic": 0.007692053714501006,
+      "acc_pdi": 0.7128205128205128,
+      "acc_ee": 0.38974358974358975,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 4.951304980686733,
+      "loss_size": 0.24394649054322923,
+      "loss_pdi": 0.7197230202811105,
+      "loss_ee": 1.3789095027106149,
+      "loss_delivery": 1.6561105762209212,
+      "loss_biodist": 0.9460095167160034,
+      "loss_toxic": 0.006605847106714334,
+      "acc_pdi": 0.7076923076923077,
+      "acc_ee": 0.38974358974358975,
+      "acc_toxic": 1.0
+    }
+  ]
+}

models/finetune_cv/fold_3/history.json

@@ -0,0 +1,447 @@
+{
+  "train": [
+    {
+      "loss": 19.317236709594727,
+      "loss_size": 14.108779287338256,
+      "loss_pdi": 1.2223044037818909,
+      "loss_ee": 1.1201724767684937,
+      "loss_delivery": 1.0989456713199615,
+      "loss_biodist": 1.2243955612182618,
+      "loss_toxic": 0.5426396489143371
+    },
+    {
+      "loss": 7.058736562728882,
+      "loss_size": 2.407193088531494,
+      "loss_pdi": 1.0396445631980895,
+      "loss_ee": 1.043432891368866,
+      "loss_delivery": 1.183875671029091,
+      "loss_biodist": 1.0435532987117768,
+      "loss_toxic": 0.3410369783639908
+    },
+    {
+      "loss": 4.30960967540741,
+      "loss_size": 0.3954987198114395,
+      "loss_pdi": 0.8135693371295929,
+      "loss_ee": 0.9796469569206238,
+      "loss_delivery": 1.0894740536808967,
+      "loss_biodist": 0.8407172799110413,
+      "loss_toxic": 0.1907033085823059
+    },
+    {
+      "loss": 3.814995551109314,
+      "loss_size": 0.3033123552799225,
+      "loss_pdi": 0.7126355469226837,
+      "loss_ee": 0.9527663111686706,
+      "loss_delivery": 1.0963637247681617,
+      "loss_biodist": 0.613499540090561,
+      "loss_toxic": 0.13641793020069598
+    },
+    {
+      "loss": 3.4925455331802366,
+      "loss_size": 0.31617238447070123,
+      "loss_pdi": 0.6569374144077301,
+      "loss_ee": 0.9212668299674988,
+      "loss_delivery": 1.0250366538763047,
+      "loss_biodist": 0.4527473896741867,
+      "loss_toxic": 0.12038486637175083
+    },
+    {
+      "loss": 3.255272912979126,
+      "loss_size": 0.27022836059331895,
+      "loss_pdi": 0.6289663434028625,
+      "loss_ee": 0.9047561466693879,
+      "loss_delivery": 0.9742588266730309,
+      "loss_biodist": 0.3838836058974266,
+      "loss_toxic": 0.09317965283989907
+    },
+    {
+      "loss": 3.281973719596863,
+      "loss_size": 0.26578598394989966,
+      "loss_pdi": 0.5881433010101318,
+      "loss_ee": 0.8700660288333892,
+      "loss_delivery": 1.1519657298922539,
+      "loss_biodist": 0.3240764126181602,
+      "loss_toxic": 0.08193630240857601
+    },
+    {
+      "loss": 2.7810576915740968,
+      "loss_size": 0.2505718767642975,
+      "loss_pdi": 0.5593925356864929,
+      "loss_ee": 0.8196510195732116,
+      "loss_delivery": 0.7932070523500443,
+      "loss_biodist": 0.27653754949569703,
+      "loss_toxic": 0.08169759791344404
+    },
+    {
+      "loss": 2.644732141494751,
+      "loss_size": 0.27979295402765275,
+      "loss_pdi": 0.5457461476325989,
+      "loss_ee": 0.7845215618610382,
+      "loss_delivery": 0.722122372686863,
+      "loss_biodist": 0.2437703028321266,
+      "loss_toxic": 0.06877880096435547
+    },
+    {
+      "loss": 2.5743841886520387,
+      "loss_size": 0.21236803606152535,
+      "loss_pdi": 0.5281321376562118,
+      "loss_ee": 0.7772053182125092,
+      "loss_delivery": 0.7842913195490837,
+      "loss_biodist": 0.20931598618626596,
+      "loss_toxic": 0.0630713876336813
+    },
+    {
+      "loss": 2.493379771709442,
+      "loss_size": 0.2545281477272511,
+      "loss_pdi": 0.514763566851616,
+      "loss_ee": 0.7416582465171814,
+      "loss_delivery": 0.7315813854336739,
+      "loss_biodist": 0.18844463676214218,
+      "loss_toxic": 0.062403830140829085
+    },
+    {
+      "loss": 2.3714203119277952,
+      "loss_size": 0.21288565024733544,
+      "loss_pdi": 0.5149440914392471,
+      "loss_ee": 0.7432775914669036,
+      "loss_delivery": 0.6615208894014358,
+      "loss_biodist": 0.17799324095249175,
+      "loss_toxic": 0.06079882858321071
+    },
+    {
+      "loss": 2.3138927936553957,
+      "loss_size": 0.22406778559088708,
+      "loss_pdi": 0.5060430943965912,
+      "loss_ee": 0.7270951688289642,
+      "loss_delivery": 0.6268678307533264,
+      "loss_biodist": 0.17946239709854125,
+      "loss_toxic": 0.050356499617919326
+    },
+    {
+      "loss": 2.2404407501220702,
+      "loss_size": 0.23460092321038245,
+      "loss_pdi": 0.4892877459526062,
+      "loss_ee": 0.6908941030502319,
+      "loss_delivery": 0.6124202072620392,
+      "loss_biodist": 0.16842604279518128,
+      "loss_toxic": 0.044811736792325974
+    },
+    {
+      "loss": 2.2448294520378114,
+      "loss_size": 0.21119624376296997,
+      "loss_pdi": 0.479864901304245,
+      "loss_ee": 0.6906192302703857,
+      "loss_delivery": 0.6555144399404526,
+      "loss_biodist": 0.16310803219676018,
+      "loss_toxic": 0.044526621932163835
+    },
+    {
+      "loss": 2.1580574989318846,
+      "loss_size": 0.18697498068213464,
+      "loss_pdi": 0.48660930395126345,
+      "loss_ee": 0.6810935467481614,
+      "loss_delivery": 0.6051739566028118,
+      "loss_biodist": 0.15406969040632248,
+      "loss_toxic": 0.04413598729297519
+    },
+    {
+      "loss": 2.114891529083252,
+      "loss_size": 0.17799586579203605,
+      "loss_pdi": 0.4589719235897064,
+      "loss_ee": 0.6686563313007354,
+      "loss_delivery": 0.6179293170571327,
+      "loss_biodist": 0.1526280902326107,
+      "loss_toxic": 0.03870999766513705
+    },
+    {
+      "loss": 2.1680126667022703,
+      "loss_size": 0.18272313922643663,
+      "loss_pdi": 0.47693236321210863,
+      "loss_ee": 0.6723115026950837,
+      "loss_delivery": 0.6574018053710461,
+      "loss_biodist": 0.14306045994162558,
+      "loss_toxic": 0.03558342705946416
+    },
+    {
+      "loss": 2.0243090748786927,
+      "loss_size": 0.19451010078191758,
+      "loss_pdi": 0.46296934187412264,
+      "loss_ee": 0.6654580652713775,
+      "loss_delivery": 0.5195972554385662,
+      "loss_biodist": 0.1416195034980774,
+      "loss_toxic": 0.04015482016839087
+    },
+    {
+      "loss": 1.980038857460022,
+      "loss_size": 0.19992023780941964,
+      "loss_pdi": 0.4373833805322647,
+      "loss_ee": 0.6562270969152451,
+      "loss_delivery": 0.5170416861772538,
+      "loss_biodist": 0.13248837292194365,
+      "loss_toxic": 0.03697808152064681
+    },
+    {
+      "loss": 2.0073827385902403,
+      "loss_size": 0.17545675858855247,
+      "loss_pdi": 0.43559625148773196,
+      "loss_ee": 0.6394164443016053,
+      "loss_delivery": 0.5809337809681893,
+      "loss_biodist": 0.13504885137081146,
+      "loss_toxic": 0.040930699557065964
+    }
+  ],
+  "val": [
+    {
+      "loss": 10.945204257965088,
+      "loss_size": 6.681218147277832,
+      "loss_pdi": 1.0216107964515686,
+      "loss_ee": 1.0486068725585938,
+      "loss_delivery": 0.4687899202108383,
+      "loss_biodist": 1.215298354625702,
+      "loss_toxic": 0.5096809715032578,
+      "acc_pdi": 0.8823529411764706,
+      "acc_ee": 0.6470588235294118,
+      "acc_toxic": 0.851063829787234
+    },
+    {
+      "loss": 4.0456085205078125,
+      "loss_size": 0.3470493406057358,
+      "loss_pdi": 0.7843169867992401,
+      "loss_ee": 0.8336820006370544,
+      "loss_delivery": 0.42519159615039825,
+      "loss_biodist": 1.1528617143630981,
+      "loss_toxic": 0.5025068372488022,
+      "acc_pdi": 0.8823529411764706,
+      "acc_ee": 0.8431372549019608,
+      "acc_toxic": 0.851063829787234
+    },
+    {
+      "loss": 3.553251266479492,
+      "loss_size": 0.07565776817500591,
+      "loss_pdi": 0.5630811750888824,
+      "loss_ee": 0.6947644650936127,
+      "loss_delivery": 0.4020952582359314,
+      "loss_biodist": 1.1898333430290222,
+      "loss_toxic": 0.627819336950779,
+      "acc_pdi": 0.8823529411764706,
+      "acc_ee": 0.8431372549019608,
+      "acc_toxic": 0.851063829787234
+    },
+    {
+      "loss": 3.2876256704330444,
+      "loss_size": 0.056399866938591,
+      "loss_pdi": 0.579200953245163,
+      "loss_ee": 0.5947848558425903,
+      "loss_delivery": 0.4561047703027725,
+      "loss_biodist": 1.0357274413108826,
+      "loss_toxic": 0.5654077678918839,
+      "acc_pdi": 0.8823529411764706,
+      "acc_ee": 0.8431372549019608,
+      "acc_toxic": 0.851063829787234
+    },
+    {
+      "loss": 3.2625988721847534,
+      "loss_size": 0.11296019703149796,
+      "loss_pdi": 0.5352367609739304,
+      "loss_ee": 0.6021667718887329,
+      "loss_delivery": 0.5046610683202744,
+      "loss_biodist": 1.0080225467681885,
+      "loss_toxic": 0.4995514266192913,
+      "acc_pdi": 0.8823529411764706,
+      "acc_ee": 0.8431372549019608,
+      "acc_toxic": 0.851063829787234
+    },
+    {
+      "loss": 3.165306806564331,
+      "loss_size": 0.09038551151752472,
+      "loss_pdi": 0.5058617442846298,
+      "loss_ee": 0.6476156711578369,
+      "loss_delivery": 0.43027013540267944,
+      "loss_biodist": 1.0445645153522491,
+      "loss_toxic": 0.4466092698276043,
+      "acc_pdi": 0.8823529411764706,
+      "acc_ee": 0.8431372549019608,
+      "acc_toxic": 0.851063829787234
+    },
+    {
+      "loss": 3.2408690452575684,
+      "loss_size": 0.22243183851242065,
+      "loss_pdi": 0.4985402673482895,
+      "loss_ee": 0.571824312210083,
+      "loss_delivery": 0.43825456500053406,
+      "loss_biodist": 0.9937507510185242,
+      "loss_toxic": 0.5160673335194588,
+      "acc_pdi": 0.8823529411764706,
+      "acc_ee": 0.8431372549019608,
+      "acc_toxic": 0.851063829787234
+    },
+    {
+      "loss": 3.5194804668426514,
+      "loss_size": 0.36968255043029785,
+      "loss_pdi": 0.4991031885147095,
+      "loss_ee": 0.5797468274831772,
+      "loss_delivery": 0.5644859671592712,
+      "loss_biodist": 0.9723091125488281,
+      "loss_toxic": 0.5341527052223682,
+      "acc_pdi": 0.8823529411764706,
+      "acc_ee": 0.8431372549019608,
+      "acc_toxic": 0.851063829787234
+    },
+    {
+      "loss": 3.539685606956482,
+      "loss_size": 0.38313066959381104,
+      "loss_pdi": 0.528433233499527,
+      "loss_ee": 0.5810057669878006,
+      "loss_delivery": 0.44039086997509,
+      "loss_biodist": 1.0017918348312378,
+      "loss_toxic": 0.6049331650137901,
+      "acc_pdi": 0.8823529411764706,
+      "acc_ee": 0.8431372549019608,
+      "acc_toxic": 0.851063829787234
+    },
+    {
+      "loss": 3.9403724670410156,
+      "loss_size": 0.6225972771644592,
+      "loss_pdi": 0.5688649713993073,
+      "loss_ee": 0.6205386221408844,
+      "loss_delivery": 0.6095166206359863,
+      "loss_biodist": 0.9419751763343811,
+      "loss_toxic": 0.576879795640707,
+      "acc_pdi": 0.803921568627451,
+      "acc_ee": 0.8431372549019608,
+      "acc_toxic": 0.851063829787234
+    },
+    {
+      "loss": 3.8653494119644165,
+      "loss_size": 0.6294703483581543,
+      "loss_pdi": 0.5615053772926331,
+      "loss_ee": 0.6072992980480194,
+      "loss_delivery": 0.47824281454086304,
+      "loss_biodist": 0.964938759803772,
+      "loss_toxic": 0.6238927394151688,
+      "acc_pdi": 0.8431372549019608,
+      "acc_ee": 0.8431372549019608,
+      "acc_toxic": 0.851063829787234
+    },
+    {
+      "loss": 4.109289169311523,
+      "loss_size": 0.753549188375473,
+      "loss_pdi": 0.6232334971427917,
+      "loss_ee": 0.6752453744411469,
+      "loss_delivery": 0.4541686922311783,
+      "loss_biodist": 1.0001222491264343,
+      "loss_toxic": 0.6029700562357903,
+      "acc_pdi": 0.7450980392156863,
+      "acc_ee": 0.7254901960784313,
+      "acc_toxic": 0.851063829787234
+    },
+    {
+      "loss": 4.084217548370361,
+      "loss_size": 0.6689053475856781,
+      "loss_pdi": 0.5947604179382324,
+      "loss_ee": 0.6819752305746078,
+      "loss_delivery": 0.5174736380577087,
+      "loss_biodist": 0.9870622158050537,
+      "loss_toxic": 0.6340407878160477,
+      "acc_pdi": 0.7843137254901961,
+      "acc_ee": 0.7647058823529411,
+      "acc_toxic": 0.851063829787234
+    },
+    {
+      "loss": 3.762814164161682,
+      "loss_size": 0.5682831406593323,
+      "loss_pdi": 0.5777421444654465,
+      "loss_ee": 0.7156199663877487,
+      "loss_delivery": 0.44971026480197906,
+      "loss_biodist": 0.9156049191951752,
+      "loss_toxic": 0.5358536541461945,
+      "acc_pdi": 0.7450980392156863,
+      "acc_ee": 0.7058823529411765,
+      "acc_toxic": 0.851063829787234
+    },
+    {
+      "loss": 4.147223711013794,
+      "loss_size": 0.6644828915596008,
+      "loss_pdi": 0.5911359935998917,
+      "loss_ee": 0.713784396648407,
+      "loss_delivery": 0.500703439116478,
+      "loss_biodist": 1.0310384333133698,
+      "loss_toxic": 0.6460786163806915,
+      "acc_pdi": 0.7647058823529411,
+      "acc_ee": 0.7058823529411765,
+      "acc_toxic": 0.851063829787234
+    },
+    {
+      "loss": 4.157853841781616,
+      "loss_size": 0.7414849102497101,
+      "loss_pdi": 0.630668580532074,
+      "loss_ee": 0.6938402056694031,
+      "loss_delivery": 0.50765261054039,
+      "loss_biodist": 0.9891600012779236,
+      "loss_toxic": 0.5950475558638573,
+      "acc_pdi": 0.7450980392156863,
+      "acc_ee": 0.7450980392156863,
+      "acc_toxic": 0.851063829787234
+    },
+    {
+      "loss": 4.2473719120025635,
+      "loss_size": 0.8058429956436157,
+      "loss_pdi": 0.5982940196990967,
+      "loss_ee": 0.7209844589233398,
+      "loss_delivery": 0.5253763496875763,
+      "loss_biodist": 0.973820835351944,
+      "loss_toxic": 0.6230533868074417,
+      "acc_pdi": 0.803921568627451,
+      "acc_ee": 0.6862745098039216,
+      "acc_toxic": 0.851063829787234
+    },
+    {
+      "loss": 3.90485680103302,
+      "loss_size": 0.5400884747505188,
+      "loss_pdi": 0.5671974420547485,
+      "loss_ee": 0.7085212767124176,
+      "loss_delivery": 0.5078988373279572,
+      "loss_biodist": 0.9940473735332489,
+      "loss_toxic": 0.5871035009622574,
+      "acc_pdi": 0.803921568627451,
+      "acc_ee": 0.7450980392156863,
+      "acc_toxic": 0.851063829787234
+    },
+    {
+      "loss": 4.193094968795776,
+      "loss_size": 0.764777421951294,
+      "loss_pdi": 0.5734306275844574,
+      "loss_ee": 0.7070393562316895,
+      "loss_delivery": 0.5335722267627716,
+      "loss_biodist": 1.0060182809829712,
+      "loss_toxic": 0.6082571670413017,
+      "acc_pdi": 0.8235294117647058,
+      "acc_ee": 0.7450980392156863,
+      "acc_toxic": 0.851063829787234
+    },
+    {
+      "loss": 4.225732326507568,
+      "loss_size": 0.7807798981666565,
+      "loss_pdi": 0.57969930768013,
+      "loss_ee": 0.7046914398670197,
+      "loss_delivery": 0.5619150400161743,
+      "loss_biodist": 1.0033797025680542,
+      "loss_toxic": 0.5952669233083725,
+      "acc_pdi": 0.7843137254901961,
+      "acc_ee": 0.7450980392156863,
+      "acc_toxic": 0.851063829787234
+    },
+    {
+      "loss": 4.122166633605957,
+      "loss_size": 0.727344423532486,
+      "loss_pdi": 0.5855642706155777,
+      "loss_ee": 0.7065156400203705,
+      "loss_delivery": 0.5201583206653595,
+      "loss_biodist": 0.9953437149524689,
+      "loss_toxic": 0.5872401967644691,
+      "acc_pdi": 0.803921568627451,
+      "acc_ee": 0.7450980392156863,
+      "acc_toxic": 0.851063829787234
+    }
+  ]
+}

models/finetune_cv/fold_4/history.json

@@ -0,0 +1,384 @@
+{
+  "train": [
+    {
+      "loss": 19.51876787705855,
+      "loss_size": 14.430041096427225,
+      "loss_pdi": 1.3163191405209629,
+      "loss_ee": 1.080450177192688,
+      "loss_delivery": 0.8366337662393396,
+      "loss_biodist": 1.2393495603041216,
+      "loss_toxic": 0.6159739873626016
+    },
+    {
+      "loss": 7.450059110468084,
+      "loss_size": 3.0973785898902197,
+      "loss_pdi": 1.0574429847977378,
+      "loss_ee": 0.972686382857236,
+      "loss_delivery": 0.8131051341241057,
+      "loss_biodist": 1.0609121160073713,
+      "loss_toxic": 0.4485339197245511
+    },
+    {
+      "loss": 4.450224074450406,
+      "loss_size": 0.4227097576314753,
+      "loss_pdi": 0.8444447679953142,
+      "loss_ee": 0.9470934163440358,
+      "loss_delivery": 1.0848771997473456,
+      "loss_biodist": 0.8853748061440208,
+      "loss_toxic": 0.26572414352135226
+    },
+    {
+      "loss": 3.697209119796753,
+      "loss_size": 0.27599087357521057,
+      "loss_pdi": 0.7228363969109275,
+      "loss_ee": 0.9080322655764493,
+      "loss_delivery": 0.8549216186458414,
+      "loss_biodist": 0.7190906730565158,
+      "loss_toxic": 0.21633729677308688
+    },
+    {
+      "loss": 3.5547448938543145,
+      "loss_size": 0.29429094425656577,
+      "loss_pdi": 0.6620089682665738,
+      "loss_ee": 0.8595339439131997,
+      "loss_delivery": 0.9978644847869873,
+      "loss_biodist": 0.5649206800894304,
+      "loss_toxic": 0.17612605541944504
+    },
+    {
+      "loss": 3.069189115004106,
+      "loss_size": 0.30328830602494156,
+      "loss_pdi": 0.6169473230838776,
+      "loss_ee": 0.8132463910362937,
+      "loss_delivery": 0.72551099143245,
+      "loss_biodist": 0.46287189017642627,
+      "loss_toxic": 0.14732415906407617
+    },
+    {
+      "loss": 3.1349260156804863,
+      "loss_size": 0.27535233172503387,
+      "loss_pdi": 0.5893999202684923,
+      "loss_ee": 0.807813747362657,
+      "loss_delivery": 0.9538742283528502,
+      "loss_biodist": 0.4018077091737227,
+      "loss_toxic": 0.1066781035201116
+    },
+    {
+      "loss": 2.6963415037501943,
+      "loss_size": 0.276088840582154,
+      "loss_pdi": 0.5490301495248621,
+      "loss_ee": 0.757920276034962,
+      "loss_delivery": 0.6706068068742752,
+      "loss_biodist": 0.34504769336093555,
+      "loss_toxic": 0.09764780781485817
+    },
+    {
+      "loss": 2.418043158271096,
+      "loss_size": 0.25799565559083765,
+      "loss_pdi": 0.5286295684901151,
+      "loss_ee": 0.73835120417855,
+      "loss_delivery": 0.5078353543173183,
+      "loss_biodist": 0.2939920425415039,
+      "loss_toxic": 0.09123929352922873
+    },
+    {
+      "loss": 2.294130650433627,
+      "loss_size": 0.20914554325017062,
+      "loss_pdi": 0.5159178945151243,
+      "loss_ee": 0.7331724437800321,
+      "loss_delivery": 0.50414734875614,
+      "loss_biodist": 0.2559647980061444,
+      "loss_toxic": 0.07578264227644964
+    },
+    {
+      "loss": 2.260723189874129,
+      "loss_size": 0.21194299920038742,
+      "loss_pdi": 0.51129734787074,
+      "loss_ee": 0.71018939668482,
+      "loss_delivery": 0.506003974513574,
+      "loss_biodist": 0.24507361785932022,
+      "loss_toxic": 0.07621594924818385
+    },
+    {
+      "loss": 2.180013732476668,
+      "loss_size": 0.21782933243296362,
+      "loss_pdi": 0.5094848789952018,
+      "loss_ee": 0.6991291533816945,
+      "loss_delivery": 0.4610004154118625,
+      "loss_biodist": 0.22516351396387274,
+      "loss_toxic": 0.06740644057704644
+    },
+    {
+      "loss": 2.131091995672746,
+      "loss_size": 0.22081551971760663,
+      "loss_pdi": 0.4984923790801655,
+      "loss_ee": 0.6744041009382769,
+      "loss_delivery": 0.44286114688624034,
+      "loss_biodist": 0.2276345125653527,
+      "loss_toxic": 0.0668843225999312
+    },
+    {
+      "loss": 2.075855114243247,
+      "loss_size": 0.1967273937030272,
+      "loss_pdi": 0.4761518023230813,
+      "loss_ee": 0.6580501876094125,
+      "loss_delivery": 0.4651151258837093,
+      "loss_biodist": 0.22098628905686465,
+      "loss_toxic": 0.05882429365407337
+    },
+    {
+      "loss": 2.070832209153609,
+      "loss_size": 0.22619221427223898,
+      "loss_pdi": 0.46735330332409253,
+      "loss_ee": 0.658088050105355,
+      "loss_delivery": 0.46364551173015073,
+      "loss_biodist": 0.1992648494514552,
+      "loss_toxic": 0.0562882690097798
+    },
+    {
+      "loss": 2.0497749502008613,
+      "loss_size": 0.2018005665053021,
+      "loss_pdi": 0.44933846592903137,
+      "loss_ee": 0.6409396637569774,
+      "loss_delivery": 0.4817944710904902,
+      "loss_biodist": 0.21351950141516599,
+      "loss_toxic": 0.0623823038556359
+    },
+    {
+      "loss": 1.998817953196439,
+      "loss_size": 0.19020642136985605,
+      "loss_pdi": 0.4579390991817821,
+      "loss_ee": 0.6345709101720289,
+      "loss_delivery": 0.45773128284649417,
+      "loss_biodist": 0.20633393864740024,
+      "loss_toxic": 0.052036324177276
+    },
+    {
+      "loss": 1.9732873006300493,
+      "loss_size": 0.18214904246005145,
+      "loss_pdi": 0.46314583312381397,
+      "loss_ee": 0.6480948545716025,
+      "loss_delivery": 0.43798652359030465,
+      "loss_biodist": 0.19162344119765543,
+      "loss_toxic": 0.05028762956234542
+    }
+  ],
+  "val": [
+    {
+      "loss": 11.670351346333822,
+      "loss_size": 7.447449843088786,
+      "loss_pdi": 1.082938313484192,
+      "loss_ee": 0.9422469735145569,
+      "loss_delivery": 0.7185012102127075,
+      "loss_biodist": 0.945093979438146,
+      "loss_toxic": 0.5341211756070455,
+      "acc_pdi": 0.8484848484848485,
+      "acc_ee": 0.8333333333333334,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.7281323273976645,
+      "loss_size": 0.4031377931435903,
+      "loss_pdi": 0.7136062383651733,
+      "loss_ee": 0.7757165829340616,
+      "loss_delivery": 0.6889261901378632,
+      "loss_biodist": 0.8803721169630686,
+      "loss_toxic": 0.2663734555244446,
+      "acc_pdi": 0.8484848484848485,
+      "acc_ee": 0.8333333333333334,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 2.996154228846232,
+      "loss_size": 0.08968868106603622,
+      "loss_pdi": 0.5251734455426534,
+      "loss_ee": 0.705430785814921,
+      "loss_delivery": 0.7259814739227295,
+      "loss_biodist": 0.8580058316389719,
+      "loss_toxic": 0.09187404563029607,
+      "acc_pdi": 0.8484848484848485,
+      "acc_ee": 0.8333333333333334,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.0241867701212564,
+      "loss_size": 0.06373066206773122,
+      "loss_pdi": 0.5075281461079916,
+      "loss_ee": 0.6890556613604227,
+      "loss_delivery": 0.8740459084510803,
+      "loss_biodist": 0.8290574749310812,
+      "loss_toxic": 0.060768917202949524,
+      "acc_pdi": 0.8484848484848485,
+      "acc_ee": 0.8333333333333334,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.2901877562204995,
+      "loss_size": 0.20215384662151337,
+      "loss_pdi": 0.5433482428391775,
+      "loss_ee": 0.7523069183031718,
+      "loss_delivery": 0.9290379285812378,
+      "loss_biodist": 0.8056556979815165,
+      "loss_toxic": 0.057684975365797676,
+      "acc_pdi": 0.8484848484848485,
+      "acc_ee": 0.8333333333333334,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.11362091700236,
+      "loss_size": 0.11133595556020737,
+      "loss_pdi": 0.5679469505945841,
+      "loss_ee": 0.8252793351809183,
+      "loss_delivery": 0.8343843619028727,
+      "loss_biodist": 0.7218613227208456,
+      "loss_toxic": 0.052812947581211724,
+      "acc_pdi": 0.8484848484848485,
+      "acc_ee": 0.8181818181818182,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.247321446736654,
+      "loss_size": 0.09277657171090443,
+      "loss_pdi": 0.6567125717798868,
+      "loss_ee": 1.0444208979606628,
+      "loss_delivery": 0.7062844236691793,
+      "loss_biodist": 0.6986102362473806,
+      "loss_toxic": 0.04851680745681127,
+      "acc_pdi": 0.8181818181818182,
+      "acc_ee": 0.45454545454545453,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.168424208958944,
+      "loss_size": 0.05177713930606842,
+      "loss_pdi": 0.5932339330514272,
+      "loss_ee": 0.968136191368103,
+      "loss_delivery": 0.7594618300596873,
+      "loss_biodist": 0.7631273567676544,
+      "loss_toxic": 0.032687741021315254,
+      "acc_pdi": 0.8484848484848485,
+      "acc_ee": 0.45454545454545453,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.1226733525594077,
+      "loss_size": 0.16442706187566122,
+      "loss_pdi": 0.4861932198206584,
+      "loss_ee": 0.8927785356839498,
+      "loss_delivery": 0.809323231379191,
+      "loss_biodist": 0.7391951779524485,
+      "loss_toxic": 0.030756143853068352,
+      "acc_pdi": 0.8484848484848485,
+      "acc_ee": 0.4696969696969697,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.4750285943349204,
+      "loss_size": 0.07188746457298596,
+      "loss_pdi": 0.635328451792399,
+      "loss_ee": 1.0510863463083904,
+      "loss_delivery": 0.9019280473391215,
+      "loss_biodist": 0.7850770453612009,
+      "loss_toxic": 0.02972123461465041,
+      "acc_pdi": 0.8181818181818182,
+      "acc_ee": 0.3787878787878788,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.3856334686279297,
+      "loss_size": 0.0857744167248408,
+      "loss_pdi": 0.5498887598514557,
+      "loss_ee": 0.9292206565539042,
+      "loss_delivery": 1.003569980462392,
+      "loss_biodist": 0.7962689697742462,
+      "loss_toxic": 0.020910644593338173,
+      "acc_pdi": 0.8333333333333334,
+      "acc_ee": 0.3787878787878788,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.594546397527059,
+      "loss_size": 0.058230139315128326,
+      "loss_pdi": 0.7094775040944418,
+      "loss_ee": 1.134681224822998,
+      "loss_delivery": 0.8755488395690918,
+      "loss_biodist": 0.7928893665472666,
+      "loss_toxic": 0.023719362293680508,
+      "acc_pdi": 0.6363636363636364,
+      "acc_ee": 0.21212121212121213,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.34331480662028,
+      "loss_size": 0.09952588627735774,
+      "loss_pdi": 0.5444782872994741,
+      "loss_ee": 0.9434934655825297,
+      "loss_delivery": 0.9477877616882324,
+      "loss_biodist": 0.7897172272205353,
+      "loss_toxic": 0.018312191901107628,
+      "acc_pdi": 0.7878787878787878,
+      "acc_ee": 0.36363636363636365,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.4212222894032798,
+      "loss_size": 0.08121616393327713,
+      "loss_pdi": 0.5517565310001373,
+      "loss_ee": 1.0685155193010967,
+      "loss_delivery": 0.874200721581777,
+      "loss_biodist": 0.828464408715566,
+      "loss_toxic": 0.017068898615737755,
+      "acc_pdi": 0.8181818181818182,
+      "acc_ee": 0.25757575757575757,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.6395487785339355,
+      "loss_size": 0.07888962080081303,
+      "loss_pdi": 0.5913220842679342,
+      "loss_ee": 1.0437468489011128,
+      "loss_delivery": 1.0660852392514546,
+      "loss_biodist": 0.8421931266784668,
+      "loss_toxic": 0.01731194742023945,
+      "acc_pdi": 0.7878787878787878,
+      "acc_ee": 0.2727272727272727,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.5140305360158286,
+      "loss_size": 0.0599971575041612,
+      "loss_pdi": 0.5561938285827637,
+      "loss_ee": 1.0674984057744343,
+      "loss_delivery": 0.9739653070767721,
+      "loss_biodist": 0.8400343159834543,
+      "loss_toxic": 0.01634151643762986,
+      "acc_pdi": 0.7424242424242424,
+      "acc_ee": 0.22727272727272727,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.628753344217936,
+      "loss_size": 0.08887146785855293,
+      "loss_pdi": 0.582503984371821,
+      "loss_ee": 1.114095131556193,
+      "loss_delivery": 0.9745903412501017,
+      "loss_biodist": 0.8516232868035635,
+      "loss_toxic": 0.017069284183283646,
+      "acc_pdi": 0.7121212121212122,
+      "acc_ee": 0.19696969696969696,
+      "acc_toxic": 1.0
+    },
+    {
+      "loss": 3.6391177972157798,
+      "loss_size": 0.08282352735598882,
+      "loss_pdi": 0.581497848033905,
+      "loss_ee": 1.141685386498769,
+      "loss_delivery": 0.9548555612564087,
+      "loss_biodist": 0.8643284440040588,
+      "loss_toxic": 0.013927079737186432,
+      "acc_pdi": 0.7424242424242424,
+      "acc_ee": 0.16666666666666666,
+      "acc_toxic": 1.0
+    }
+  ]
+}

models/finetune_cv/test_results.json

@@ -0,0 +1,294 @@
+{
+  "fold_results": [
+    {
+      "fold_idx": 0,
+      "n_samples": 95,
+      "size": {
+        "n": 95,
+        "rmse": 0.5909209144067168,
+        "mae": 0.376253614927593,
+        "r2": 0.005712927161997228
+      },
+      "delivery": {
+        "n": 66,
+        "rmse": 1.3280577883458438,
+        "mae": 0.5195405159964028,
+        "r2": 0.03195999739694366
+      },
+      "pdi": {
+        "n": 95,
+        "accuracy": 0.6105263157894737,
+        "precision": 0.20350877192982456,
+        "recall": 0.3333333333333333,
+        "f1": 0.25272331154684097
+      },
+      "ee": {
+        "n": 95,
+        "accuracy": 0.6736842105263158,
+        "precision": 0.22456140350877193,
+        "recall": 0.3333333333333333,
+        "f1": 0.26834381551362685
+      },
+      "toxic": {
+        "n": 66,
+        "accuracy": 0.8939393939393939,
+        "precision": 0.44696969696969696,
+        "recall": 0.5,
+        "f1": 0.472
+      },
+      "biodist": {
+        "n": 66,
+        "kl_divergence": 0.851655784204727,
+        "js_divergence": 0.21404831573756974
+      }
+    },
+    {
+      "fold_idx": 1,
+      "n_samples": 195,
+      "size": {
+        "n": 193,
+        "rmse": 0.4425801645813746,
+        "mae": 0.26432527161632796,
+        "r2": -0.026225211870033682
+      },
+      "delivery": {
+        "n": 123,
+        "rmse": 0.7771322048436382,
+        "mae": 0.6133777339870822,
+        "r2": -0.128644776760948
+      },
+      "pdi": {
+        "n": 195,
+        "accuracy": 0.7076923076923077,
+        "precision": 0.35384615384615387,
+        "recall": 0.5,
+        "f1": 0.4144144144144144
+      },
+      "ee": {
+        "n": 195,
+        "accuracy": 0.4205128205128205,
+        "precision": 0.14017094017094017,
+        "recall": 0.3333333333333333,
+        "f1": 0.19735258724428398
+      },
+      "toxic": {
+        "n": 123,
+        "accuracy": 1.0,
+        "precision": 1.0,
+        "recall": 1.0,
+        "f1": 1.0
+      },
+      "biodist": {
+        "n": 123,
+        "kl_divergence": 0.9336461102028436,
+        "js_divergence": 0.24870266224462317
+      }
+    },
+    {
+      "fold_idx": 2,
+      "n_samples": 51,
+      "size": {
+        "n": 51,
+        "rmse": 0.6473513298834871,
+        "mae": 0.5600235602434944,
+        "r2": -9.27515642706235
+      },
+      "delivery": {
+        "n": 44,
+        "rmse": 0.7721077356414991,
+        "mae": 0.6167582499593581,
+        "r2": -0.4822886602727561
+      },
+      "pdi": {
+        "n": 51,
+        "accuracy": 0.8823529411764706,
+        "precision": 0.29411764705882354,
+        "recall": 0.3333333333333333,
+        "f1": 0.3125
+      },
+      "ee": {
+        "n": 51,
+        "accuracy": 0.8431372549019608,
+        "precision": 0.28104575163398693,
+        "recall": 0.3333333333333333,
+        "f1": 0.3049645390070922
+      },
+      "toxic": {
+        "n": 47,
+        "accuracy": 0.851063829787234,
+        "precision": 0.425531914893617,
+        "recall": 0.5,
+        "f1": 0.4597701149425288
+      },
+      "biodist": {
+        "n": 45,
+        "kl_divergence": 1.1049896129018548,
+        "js_divergence": 0.25485248115851133
+      }
+    },
+    {
+      "fold_idx": 3,
+      "n_samples": 66,
+      "size": {
+        "n": 66,
+        "rmse": 0.2407212117920812,
+        "mae": 0.19363613562150436,
+        "r2": -0.11204941379936861
+      },
+      "delivery": {
+        "n": 62,
+        "rmse": 1.0041711455927012,
+        "mae": 0.7132550483914993,
+        "r2": -0.63265374674746
+      },
+      "pdi": {
+        "n": 66,
+        "accuracy": 0.8484848484848485,
+        "precision": 0.42424242424242425,
+        "recall": 0.5,
+        "f1": 0.4590163934426229
+      },
+      "ee": {
+        "n": 66,
+        "accuracy": 0.8181818181818182,
+        "precision": 0.27692307692307694,
+        "recall": 0.32727272727272727,
+        "f1": 0.3
+      },
+      "toxic": {
+        "n": 62,
+        "accuracy": 1.0,
+        "precision": 1.0,
+        "recall": 1.0,
+        "f1": 1.0
+      },
+      "biodist": {
+        "n": 62,
+        "kl_divergence": 0.9677978984139058,
+        "js_divergence": 0.2020309307244639
+      }
+    },
+    {
+      "fold_idx": 4,
+      "n_samples": 27,
+      "size": {
+        "n": 27,
+        "rmse": 0.23392834445509142,
+        "mae": 0.19066280788845485,
+        "r2": -0.2667651950955112
+      },
+      "delivery": {
+        "n": 15,
+        "rmse": 1.9603892288630869,
+        "mae": 1.3892907698949177,
+        "r2": -0.29760739742916287
+      },
+      "pdi": {
+        "n": 27,
+        "accuracy": 0.8888888888888888,
+        "precision": 0.4444444444444444,
+        "recall": 0.5,
+        "f1": 0.47058823529411764
+      },
+      "ee": {
+        "n": 27,
+        "accuracy": 0.5925925925925926,
+        "precision": 0.19753086419753085,
+        "recall": 0.3333333333333333,
+        "f1": 0.24806201550387597
+      },
+      "toxic": {
+        "n": 15,
+        "accuracy": 1.0,
+        "precision": 1.0,
+        "recall": 1.0,
+        "f1": 1.0
+      },
+      "biodist": {
+        "n": 15,
+        "kl_divergence": 0.9389607012315264,
+        "js_divergence": 0.2470218476598176
+      }
+    }
+  ],
+  "summary_stats": {
+    "size": {
+      "rmse_mean": 0.43110039302375025,
+      "rmse_std": 0.17179051271013462,
+      "r2_mean": -1.9348966641330534,
+      "r2_std": 3.6713441784129
+    },
+    "delivery": {
+      "rmse_mean": 1.1683716206573538,
+      "rmse_std": 0.4449374578352648,
+      "r2_mean": -0.30184691676267666,
+      "r2_std": 0.23809090378746706
+    },
+    "pdi": {
+      "accuracy_mean": 0.7875890604063979,
+      "accuracy_std": 0.11016791908756088,
+      "f1_mean": 0.3818484709395992,
+      "f1_std": 0.08529090446864619
+    },
+    "ee": {
+      "accuracy_mean": 0.6696217393431015,
+      "accuracy_std": 0.15503740047242787,
+      "f1_mean": 0.2637445914537758,
+      "f1_std": 0.039213602228007696
+    },
+    "toxic": {
+      "accuracy_mean": 0.9490006447453256,
+      "accuracy_std": 0.06391582554207781,
+      "f1_mean": 0.7863540229885058,
+      "f1_std": 0.26169039387919035
+    },
+    "biodist": {
+      "kl_mean": 0.9594100213909715,
+      "kl_std": 0.08240959093662605,
+      "js_mean": 0.23333124750499712,
+      "js_std": 0.021158533549255752
+    }
+  },
+  "overall": {
+    "size": {
+      "n_samples": 432,
+      "mse": 0.22604480336185886,
+      "rmse": 0.47544169291497657,
+      "mae": 0.3084443360567093,
+      "r2": -0.2313078534105617
+    },
+    "delivery": {
+      "n_samples": 310,
+      "mse": 1.0873755440675295,
+      "rmse": 1.0427730069710903,
+      "mae": 0.6513989447841361,
+      "r2": -0.09443640807387799
+    },
+    "pdi": {
+      "n_samples": 434,
+      "accuracy": 0.7396313364055299,
+      "precision": 0.18490783410138248,
+      "recall": 0.25,
+      "f1": 0.21258278145695364
+    },
+    "ee": {
+      "n_samples": 434,
+      "accuracy": 0.5967741935483871,
+      "precision": 0.1993841416474211,
+      "recall": 0.33205128205128204,
+      "f1": 0.24915824915824913
+    },
+    "toxic": {
+      "n_samples": 313,
+      "accuracy": 0.9552715654952076,
+      "precision": 0.4776357827476038,
+      "recall": 0.5,
+      "f1": 0.48856209150326796
+    },
+    "biodist": {
+      "n_samples": 311,
+      "kl_divergence": 0.9481034280166569,
+      "js_divergence": 0.23285280825310384
+    }
+  }
+}

models/pretrain_history.json

@@ -1,310 +1,206 @@
 {
   "train": [
     {
-      "loss": 0.8244676398801744,
-      "n_samples": 8721
+      "loss": 0.7730368412685099,
+      "n_samples": 6783
     },
     {
-      "loss": 0.6991508170533461,
-      "n_samples": 8721
+      "loss": 0.658895703010919,
+      "n_samples": 6783
     },
     {
-      "loss": 0.6388374940987616,
-      "n_samples": 8721
+      "loss": 0.6059015260392299,
+      "n_samples": 6783
     },
     {
-      "loss": 0.6008581508669937,
-      "n_samples": 8721
+      "loss": 0.5744731174349416,
+      "n_samples": 6783
     },
     {
-      "loss": 0.584832567446085,
-      "n_samples": 8721
+      "loss": 0.5452056020458733,
+      "n_samples": 6783
     },
     {
-      "loss": 0.5481657371815157,
-      "n_samples": 8721
+      "loss": 0.5138543470936083,
+      "n_samples": 6783
     },
     {
-      "loss": 0.5368926340308079,
-      "n_samples": 8721
+      "loss": 0.4885380559178135,
+      "n_samples": 6783
     },
     {
-      "loss": 0.5210388793613561,
-      "n_samples": 8721
+      "loss": 0.47587182296687974,
+      "n_samples": 6783
     },
     {
-      "loss": 0.49758357966374045,
-      "n_samples": 8721
+      "loss": 0.4671051038255316,
+      "n_samples": 6783
     },
     {
-      "loss": 0.49256294099457043,
-      "n_samples": 8721
+      "loss": 0.46794115915756107,
+      "n_samples": 6783
     },
     {
-      "loss": 0.4697267088016886,
-      "n_samples": 8721
+      "loss": 0.4293930456997915,
+      "n_samples": 6783
     },
     {
-      "loss": 0.45763822707571084,
-      "n_samples": 8721
+      "loss": 0.42624105651716415,
+      "n_samples": 6783
     },
     {
-      "loss": 0.4495221330627172,
-      "n_samples": 8721
+      "loss": 0.4131358770446828,
+      "n_samples": 6783
     },
     {
-      "loss": 0.446159594079631,
-      "n_samples": 8721
+      "loss": 0.3946074267790835,
+      "n_samples": 6783
     },
     {
-      "loss": 0.4327090857889029,
-      "n_samples": 8721
+      "loss": 0.3898155013755344,
+      "n_samples": 6783
     },
     {
-      "loss": 0.4249273364101852,
-      "n_samples": 8721
+      "loss": 0.37861797005733383,
+      "n_samples": 6783
     },
     {
-      "loss": 0.4216959138704459,
-      "n_samples": 8721
+      "loss": 0.3775682858392304,
+      "n_samples": 6783
     },
     {
-      "loss": 0.416526201182502,
-      "n_samples": 8721
+      "loss": 0.3800349080262064,
+      "n_samples": 6783
     },
     {
-      "loss": 0.40368679039741573,
-      "n_samples": 8721
+      "loss": 0.36302345173031675,
+      "n_samples": 6783
     },
     {
-      "loss": 0.4051084730032182,
-      "n_samples": 8721
+      "loss": 0.3429561740842766,
+      "n_samples": 6783
     },
     {
-      "loss": 0.38971701020385785,
-      "n_samples": 8721
+      "loss": 0.3445638883004898,
+      "n_samples": 6783
     },
     {
-      "loss": 0.39155546386038786,
-      "n_samples": 8721
+      "loss": 0.318970229203733,
+      "n_samples": 6783
     },
     {
-      "loss": 0.37976963541784114,
-      "n_samples": 8721
+      "loss": 0.30179278279904437,
+      "n_samples": 6783
     },
     {
-      "loss": 0.36484339719805037,
-      "n_samples": 8721
+      "loss": 0.2887343142006437,
+      "n_samples": 6783
     },
     {
-      "loss": 0.36232607571196496,
-      "n_samples": 8721
-    },
-    {
-      "loss": 0.3345973272380199,
-      "n_samples": 8721
-    },
-    {
-      "loss": 0.31767916518768957,
-      "n_samples": 8721
-    },
-    {
-      "loss": 0.32065429246052457,
-      "n_samples": 8721
-    },
-    {
-      "loss": 0.3171297926146043,
-      "n_samples": 8721
-    },
-    {
-      "loss": 0.3122120894173009,
-      "n_samples": 8721
-    },
-    {
-      "loss": 0.3135035038404461,
-      "n_samples": 8721
-    },
-    {
-      "loss": 0.2987745178222875,
-      "n_samples": 8721
-    },
-    {
-      "loss": 0.2914867957853393,
-      "n_samples": 8721
-    },
-    {
-      "loss": 0.2983839795507705,
-      "n_samples": 8721
-    },
-    {
-      "loss": 0.2826709597875678,
-      "n_samples": 8721
-    },
-    {
-      "loss": 0.2731766632569382,
-      "n_samples": 8721
-    },
-    {
-      "loss": 0.27726896305742266,
-      "n_samples": 8721
-    },
-    {
-      "loss": 0.27864557847067956,
-      "n_samples": 8721
+      "loss": 0.29240367556855545,
+      "n_samples": 6783
     }
   ],
   "val": [
     {
-      "loss": 0.7601077516012517,
-      "n_samples": 969
+      "loss": 0.7350345371841441,
+      "n_samples": 2907
     },
     {
-      "loss": 0.7119935319611901,
-      "n_samples": 969
+      "loss": 0.7165568811318536,
+      "n_samples": 2907
     },
     {
-      "loss": 0.6461842978148269,
-      "n_samples": 969
+      "loss": 0.7251406249862214,
+      "n_samples": 2907
     },
     {
-      "loss": 0.7006978391063226,
-      "n_samples": 969
+      "loss": 0.6836505264587159,
+      "n_samples": 2907
     },
     {
-      "loss": 0.6533874032943979,
-      "n_samples": 969
+      "loss": 0.6747132955771933,
+      "n_samples": 2907
     },
     {
-      "loss": 0.6413641451743611,
-      "n_samples": 969
+      "loss": 0.6691136244936912,
+      "n_samples": 2907
     },
     {
-      "loss": 0.6168395132979742,
-      "n_samples": 969
+      "loss": 0.6337480902323249,
+      "n_samples": 2907
     },
     {
-      "loss": 0.6095251602162025,
-      "n_samples": 969
+      "loss": 0.6600317959527934,
+      "n_samples": 2907
     },
     {
-      "loss": 0.5887809592626905,
-      "n_samples": 969
+      "loss": 0.6439923948855346,
+      "n_samples": 2907
     },
     {
-      "loss": 0.5655298325376368,
-      "n_samples": 969
+      "loss": 0.643800035575267,
+      "n_samples": 2907
     },
     {
-      "loss": 0.5809201743872788,
-      "n_samples": 969
+      "loss": 0.6181512585221839,
+      "n_samples": 2907
     },
     {
-      "loss": 0.5897585974912033,
-      "n_samples": 969
+      "loss": 0.6442458634939151,
+      "n_samples": 2907
     },
     {
-      "loss": 0.5732012489662573,
-      "n_samples": 969
+      "loss": 0.6344759362359862,
+      "n_samples": 2907
     },
     {
-      "loss": 0.5607388911786094,
-      "n_samples": 969
+      "loss": 0.6501405371457472,
+      "n_samples": 2907
     },
     {
-      "loss": 0.5717580675371414,
-      "n_samples": 969
+      "loss": 0.6098835162990152,
+      "n_samples": 2907
     },
     {
-      "loss": 0.5553950037657291,
-      "n_samples": 969
+      "loss": 0.6366627322138894,
+      "n_samples": 2907
     },
     {
-      "loss": 0.5778171792857049,
-      "n_samples": 969
+      "loss": 0.6171610150646417,
+      "n_samples": 2907
     },
     {
-      "loss": 0.5602665468127734,
-      "n_samples": 969
+      "loss": 0.6358801012273748,
+      "n_samples": 2907
     },
     {
-      "loss": 0.5475307451359259,
-      "n_samples": 969
+      "loss": 0.6239976831059871,
+      "n_samples": 2907
     },
     {
-      "loss": 0.551515599314827,
-      "n_samples": 969
+      "loss": 0.6683828232827201,
+      "n_samples": 2907
     },
     {
-      "loss": 0.5755438121541243,
-      "n_samples": 969
+      "loss": 0.6655785786478143,
+      "n_samples": 2907
     },
     {
-      "loss": 0.5798238261811381,
-      "n_samples": 969
+      "loss": 0.6152775046503088,
+      "n_samples": 2907
     },
     {
-      "loss": 0.5739961433828923,
-      "n_samples": 969
+      "loss": 0.6202247662153858,
+      "n_samples": 2907
     },
     {
-      "loss": 0.5742599932540312,
-      "n_samples": 969
+      "loss": 0.648199727435189,
+      "n_samples": 2907
     },
     {
-      "loss": 0.5834948123885382,
-      "n_samples": 969
-    },
-    {
-      "loss": 0.554078846570139,
-      "n_samples": 969
-    },
-    {
-      "loss": 0.5714933996322354,
-      "n_samples": 969
-    },
-    {
-      "loss": 0.5384107524350331,
-      "n_samples": 969
-    },
-    {
-      "loss": 0.570854394451568,
-      "n_samples": 969
-    },
-    {
-      "loss": 0.5767292551642478,
-      "n_samples": 969
-    },
-    {
-      "loss": 0.5660079547556808,
-      "n_samples": 969
-    },
-    {
-      "loss": 0.5608972411514312,
-      "n_samples": 969
-    },
-    {
-      "loss": 0.5620947442987263,
-      "n_samples": 969
-    },
-    {
-      "loss": 0.5706970894361305,
-      "n_samples": 969
-    },
-    {
-      "loss": 0.5702376298690974,
-      "n_samples": 969
-    },
-    {
-      "loss": 0.5758474825259579,
-      "n_samples": 969
-    },
-    {
-      "loss": 0.5673816067284844,
-      "n_samples": 969
-    },
-    {
-      "loss": 0.5671441179879925,
-      "n_samples": 969
+      "loss": 0.6473217075085124,
+      "n_samples": 2907
     }
   ]
 }

scripts/process_data_cv.py

@@ -1,232 +1,224 @@
-"""处理 cross-validation 数据脚本：将 CV splits 转换为模型所需的 parquet 格式"""
+"""内部数据 Cross-Validation 划分脚本：基于 Amine 的分组划分"""
 
 from pathlib import Path
-from typing import Dict, List, Tuple
+from typing import List
 
+import numpy as np
 import pandas as pd
 import typer
 from loguru import logger
 
-from lnp_ml.config import EXTERNAL_DATA_DIR, PROCESSED_DATA_DIR
+from lnp_ml.config import INTERIM_DATA_DIR, PROCESSED_DATA_DIR
 from lnp_ml.dataset import (
-    LNPDatasetConfig,
+    process_dataframe,
+    SMILES_COL,
     COMP_COLS,
     HELP_COLS,
+    TARGET_REGRESSION,
+    TARGET_CLASSIFICATION_PDI,
+    TARGET_CLASSIFICATION_EE,
+    TARGET_TOXIC,
+    TARGET_BIODIST,
     get_phys_cols,
     get_exp_cols,
-    EXP_ONEHOT_SPECS,
-    PHYS_ONEHOT_SPECS,
 )
 
-
 app = typer.Typer()
 
 
-# CV extra_x 列名到模型列名的映射
-CV_COL_MAPPING = {
-    # Batch_or_individual_or_barcoded -> Sample_organization_type (for Value_name related)
-    "Batch_or_individual_or_barcoded_Barcoded": "Batch_or_individual_or_barcoded_Barcoded",
-    "Batch_or_individual_or_barcoded_Individual": "Batch_or_individual_or_barcoded_Individual",
-    # Helper_lipid_ID_None 不在模型中使用，忽略
-}
-
-
-def load_cv_split(cv_dir: Path) -> Tuple[pd.DataFrame, pd.DataFrame, pd.DataFrame]:
+def amine_based_cv_split(
+    df: pd.DataFrame,
+    n_folds: int = 5,
+    seed: int = 42,
+    amine_col: str = "Amine",
+) -> List[dict]:
     """
-    加载单个 CV split 的数据。
+    基于 Amine 列进行 Cross-Validation 划分。
+    
+    步骤：
+    1. 按 amine_col 分组
+    2. 打乱分组顺序
+    3. 将分组 round-robin 分配到 n_folds 个容器
+    4. 对于每个 fold i：
+       - validation = container[i]
+       - test = container[(i+1) % n_folds]
+       - train = 其余所有
     
     Args:
-        cv_dir: CV split 目录，如 cv_0/
+        df: 输入 DataFrame
+        n_folds: 折数
+        seed: 随机种子
+        amine_col: 用于分组的列名
         
     Returns:
-        (train_df, valid_df, test_df) 合并后的 DataFrame
+        List of dicts，每个 dict 包含 train_df, val_df, test_df
     """
-    splits = {}
-    for split_name in ["train", "valid", "test"]:
-        # 加载主数据（smiles, quantified_delivery）
-        main_path = cv_dir / f"{split_name}.csv"
-        extra_x_path = cv_dir / f"{split_name}_extra_x.csv"
-        metadata_path = cv_dir / f"{split_name}_metadata.csv"
+    # 获取唯一的 amine 并打乱
+    unique_amines = df[amine_col].unique()
+    rng = np.random.RandomState(seed)
+    rng.shuffle(unique_amines)
+    
+    logger.info(f"Found {len(unique_amines)} unique amines")
+    
+    # Round-robin 分配到 n_folds 个容器
+    containers = [[] for _ in range(n_folds)]
+    for i, amine in enumerate(unique_amines):
+        containers[i % n_folds].append(amine)
+    
+    # 打印每个容器的大小
+    for i, container in enumerate(containers):
+        container_samples = df[df[amine_col].isin(container)]
+        logger.info(f"  Container {i}: {len(container)} amines, {len(container_samples)} samples")
+    
+    # 生成每个 fold 的数据
+    fold_splits = []
+    for i in range(n_folds):
+        val_amines = set(containers[i])
+        test_amines = set(containers[(i + 1) % n_folds])
+        train_amines = set()
+        for j in range(n_folds):
+            if j != i and j != (i + 1) % n_folds:
+                train_amines.update(containers[j])
         
-        if not main_path.exists():
-            raise FileNotFoundError(f"Missing {main_path}")
+        train_df = df[df[amine_col].isin(train_amines)].reset_index(drop=True)
+        val_df = df[df[amine_col].isin(val_amines)].reset_index(drop=True)
+        test_df = df[df[amine_col].isin(test_amines)].reset_index(drop=True)
         
-        main_df = pd.read_csv(main_path)
+        fold_splits.append({
+            "train": train_df,
+            "val": val_df,
+            "test": test_df,
+        })
         
-        # 加载 extra_x（已 one-hot 编码的特征）
-        if extra_x_path.exists():
-            extra_x_df = pd.read_csv(extra_x_path)
-            # 确保行数一致
-            assert len(main_df) == len(extra_x_df), f"Row count mismatch: {len(main_df)} vs {len(extra_x_df)}"
-            # 合并（按行索引）
-            df = pd.concat([main_df, extra_x_df], axis=1)
-        else:
-            df = main_df
-            logger.warning(f"  {split_name}_extra_x.csv not found, using main data only")
-        
-        # 可选：从 metadata 获取额外信息
-        if metadata_path.exists():
-            metadata_df = pd.read_csv(metadata_path)
-            # 提取需要的列（如 Purity, Mix_type, Value_name 等）
-            for col in ["Purity", "Mix_type", "Value_name", "Target_or_delivered_gene"]:
-                if col in metadata_df.columns and col not in df.columns:
-                    df[col] = metadata_df[col]
-        
-        splits[split_name] = df
+        logger.info(
+            f"Fold {i}: train={len(train_df)} ({len(train_amines)} amines), "
+            f"val={len(val_df)} ({len(val_amines)} amines), "
+            f"test={len(test_df)} ({len(test_amines)} amines)"
+        )
     
-    return splits["train"], splits["valid"], splits["test"]
-
-
-def process_cv_dataframe(df: pd.DataFrame) -> pd.DataFrame:
-    """
-    处理 CV 数据的 DataFrame，对齐到模型所需的列格式。
-    
-    CV 数据的 extra_x 已经包含大部分 one-hot 编码，但需要：
-    1. 添加缺失的 one-hot 列（设为 0）
-    2. 从 metadata 中生成 phys token 的 one-hot 列（Purity, Mix_type, Cargo_type, Target_or_delivered_gene）
-    3. 生成 Value_name 的 one-hot 列
-    """
-    df = df.copy()
-    
-    # 1. 处理 comp 列
-    for col in COMP_COLS:
-        if col in df.columns:
-            df[col] = pd.to_numeric(df[col], errors="coerce").fillna(0.0)
-        else:
-            df[col] = 0.0
-    
-    # 2. 处理 help 列
-    for col in HELP_COLS:
-        if col not in df.columns:
-            df[col] = 0.0
-        else:
-            df[col] = df[col].fillna(0.0).astype(float)
-    
-    # 3. 处理 phys token 的 one-hot 列
-    for col, values in PHYS_ONEHOT_SPECS.items():
-        for v in values:
-            onehot_col = f"{col}_{v}"
-            if onehot_col not in df.columns:
-                # 尝试从原始列生成
-                if col in df.columns:
-                    df[onehot_col] = (df[col] == v).astype(float)
-                else:
-                    df[onehot_col] = 0.0
-            else:
-                df[onehot_col] = df[onehot_col].fillna(0.0).astype(float)
-    
-    # 4. 处理 exp token 的 one-hot 列
-    for col, values in EXP_ONEHOT_SPECS.items():
-        for v in values:
-            onehot_col = f"{col}_{v}"
-            if onehot_col not in df.columns:
-                # 尝试从原始列生成
-                if col in df.columns:
-                    df[onehot_col] = (df[col] == v).astype(float)
-                else:
-                    df[onehot_col] = 0.0
-            else:
-                df[onehot_col] = df[onehot_col].fillna(0.0).astype(float)
-    
-    # 5. 处理 quantified_delivery
-    if "quantified_delivery" in df.columns:
-        df["quantified_delivery"] = pd.to_numeric(df["quantified_delivery"], errors="coerce")
-    
-    return df
-
-
-def get_feature_columns() -> List[str]:
-    """获取所有特征列名"""
-    config = LNPDatasetConfig()
-    return (
-        ["smiles"]
-        + config.comp_cols
-        + config.phys_cols
-        + config.help_cols
-        + config.exp_cols
-        + ["quantified_delivery"]
-    )
+    return fold_splits
 
 
 @app.command()
 def main(
-    data_dir: Path = EXTERNAL_DATA_DIR / "all_amine_split_for_LiON",
+    input_path: Path = INTERIM_DATA_DIR / "internal_corrected.csv",
     output_dir: Path = PROCESSED_DATA_DIR / "cv",
     n_folds: int = 5,
+    seed: int = 42,
+    amine_col: str = "Amine",
 ):
     """
-    处理 cross-validation 数据，生成模型所需的 parquet 文件。
+    基于 Amine 分组进行 Cross-Validation 数据划分。
+    
+    采用类似 scaffold splitting 的思路，将相同 Amine 的数据放在同一组，
+    确保训练集和测试集之间没有 Amine 泄露。
+    
+    划分比例约为 train:val:test ≈ 3:1:1
     
     输出结构:
         - processed/cv/fold_0/train.parquet
-        - processed/cv/fold_0/valid.parquet
+        - processed/cv/fold_0/val.parquet  
         - processed/cv/fold_0/test.parquet
         - processed/cv/fold_1/...
         - processed/cv/feature_columns.txt
     """
-    logger.info(f"Processing CV data from {data_dir}")
+    logger.info(f"Loading data from {input_path}")
+    df = pd.read_csv(input_path)
+    logger.info(f"Loaded {len(df)} samples")
     
-    # 获取所有 cv_* 目录
-    cv_dirs = sorted([d for d in data_dir.iterdir() if d.is_dir() and d.name.startswith("cv_")])
-    
-    if len(cv_dirs) == 0:
-        logger.error(f"No cv_* directories found in {data_dir}")
+    # 检查 amine 列是否存在
+    if amine_col not in df.columns:
+        logger.error(f"Column '{amine_col}' not found in data. Available columns: {list(df.columns)}")
         raise typer.Exit(1)
     
-    if len(cv_dirs) != n_folds:
-        logger.warning(f"Expected {n_folds} folds, found {len(cv_dirs)}")
+    # 处理数据（列对齐、one-hot 生成等）
+    logger.info("Processing dataframe...")
+    df = process_dataframe(df)
     
-    logger.info(f"Found {len(cv_dirs)} folds: {[d.name for d in cv_dirs]}")
+    # 确保 Amine 列仍然存在（process_dataframe 可能不会保留它）
+    # 重新加载原始数据获取 Amine 列
+    original_df = pd.read_csv(input_path)
+    if amine_col in original_df.columns and amine_col not in df.columns:
+        df[amine_col] = original_df[amine_col].values
     
-    feature_cols = get_feature_columns()
+    # 定义要保留的列
+    phys_cols = get_phys_cols()
+    exp_cols = get_exp_cols()
+    
+    keep_cols = (
+        [SMILES_COL]
+        + COMP_COLS
+        + phys_cols
+        + HELP_COLS
+        + exp_cols
+        + TARGET_REGRESSION
+        + TARGET_CLASSIFICATION_PDI
+        + TARGET_CLASSIFICATION_EE
+        + [TARGET_TOXIC]
+        + TARGET_BIODIST
+    )
+    
+    # 只保留存在的列
+    keep_cols = [c for c in keep_cols if c in df.columns]
+    
+    # 进行 CV 划分
+    logger.info(f"\nPerforming {n_folds}-fold amine-based CV split (seed={seed})...")
+    fold_splits = amine_based_cv_split(df, n_folds=n_folds, seed=seed, amine_col=amine_col)
+    
+    # 保存每个 fold
     output_dir.mkdir(parents=True, exist_ok=True)
     
-    for i, cv_dir in enumerate(cv_dirs):
-        fold_name = f"fold_{i}"
-        fold_output_dir = output_dir / fold_name
-        fold_output_dir.mkdir(parents=True, exist_ok=True)
-        
-        logger.info(f"Processing {cv_dir.name} -> {fold_name}")
-        
-        # 加载数据
-        train_df, valid_df, test_df = load_cv_split(cv_dir)
-        
-        logger.info(f"  Loaded: train={len(train_df)}, valid={len(valid_df)}, test={len(test_df)}")
-        
-        # 处理数据
-        train_df = process_cv_dataframe(train_df)
-        valid_df = process_cv_dataframe(valid_df)
-        test_df = process_cv_dataframe(test_df)
-        
-        # 确保所有列存在
-        for col in feature_cols:
-            for df in [train_df, valid_df, test_df]:
-                if col not in df.columns:
-                    df[col] = 0.0 if col != "smiles" else ""
+    for i, split in enumerate(fold_splits):
+        fold_dir = output_dir / f"fold_{i}"
+        fold_dir.mkdir(parents=True, exist_ok=True)
         
         # 只保留需要的列
-        train_df = train_df[feature_cols]
-        valid_df = valid_df[feature_cols]
-        test_df = test_df[feature_cols]
+        train_df = split["train"][keep_cols].reset_index(drop=True)
+        val_df = split["val"][keep_cols].reset_index(drop=True)
+        test_df = split["test"][keep_cols].reset_index(drop=True)
         
         # 保存
-        train_df.to_parquet(fold_output_dir / "train.parquet", index=False)
-        valid_df.to_parquet(fold_output_dir / "valid.parquet", index=False)
-        test_df.to_parquet(fold_output_dir / "test.parquet", index=False)
+        train_df.to_parquet(fold_dir / "train.parquet", index=False)
+        val_df.to_parquet(fold_dir / "val.parquet", index=False)
+        test_df.to_parquet(fold_dir / "test.parquet", index=False)
         
-        logger.success(f"  Saved to {fold_output_dir}")
+        logger.success(f"Saved fold {i} to {fold_dir}")
     
-    # 保存特征列配置
-    cols_path = output_dir / "feature_columns.txt"
-    with open(cols_path, "w") as f:
-        f.write("\n".join(feature_cols))
-    logger.success(f"Saved feature columns to {cols_path}")
+    # 保存列名配置
+    config_path = output_dir / "feature_columns.txt"
+    with open(config_path, "w") as f:
+        f.write("# Feature columns configuration\n\n")
+        f.write(f"# SMILES\n{SMILES_COL}\n\n")
+        f.write(f"# comp token [{len(COMP_COLS)}]\n")
+        f.write("\n".join(COMP_COLS) + "\n\n")
+        f.write(f"# phys token [{len(phys_cols)}]\n")
+        f.write("\n".join(phys_cols) + "\n\n")
+        f.write(f"# help token [{len(HELP_COLS)}]\n")
+        f.write("\n".join(HELP_COLS) + "\n\n")
+        f.write(f"# exp token [{len(exp_cols)}]\n")
+        f.write("\n".join(exp_cols) + "\n\n")
+        f.write("# Targets\n")
+        f.write("## Regression\n")
+        f.write("\n".join(TARGET_REGRESSION) + "\n")
+        f.write("## PDI classification\n")
+        f.write("\n".join(TARGET_CLASSIFICATION_PDI) + "\n")
+        f.write("## EE classification\n")
+        f.write("\n".join(TARGET_CLASSIFICATION_EE) + "\n")
+        f.write("## Toxic\n")
+        f.write(f"{TARGET_TOXIC}\n")
+        f.write("## Biodistribution\n")
+        f.write("\n".join(TARGET_BIODIST) + "\n")
     
+    logger.success(f"Saved feature config to {config_path}")
+    
+    # 打印汇总
     logger.info("\n" + "=" * 60)
     logger.info("CV DATA PROCESSING COMPLETE")
     logger.info("=" * 60)
     logger.info(f"Output directory: {output_dir}")
-    logger.info(f"Number of folds: {len(cv_dirs)}")
+    logger.info(f"Number of folds: {n_folds}")
+    logger.info(f"Splitting method: Amine-based (column: {amine_col})")
+    logger.info(f"Random seed: {seed}")
 
 
 if __name__ == "__main__":

scripts/process_external_cv.py

@@ -0,0 +1,234 @@
+"""处理 cross-validation 数据脚本：将 CV splits 转换为模型所需的 parquet 格式"""
+
+from pathlib import Path
+from typing import Dict, List, Tuple
+
+import pandas as pd
+import typer
+from loguru import logger
+
+from lnp_ml.config import EXTERNAL_DATA_DIR, PROCESSED_DATA_DIR
+from lnp_ml.dataset import (
+    LNPDatasetConfig,
+    COMP_COLS,
+    HELP_COLS,
+    get_phys_cols,
+    get_exp_cols,
+    EXP_ONEHOT_SPECS,
+    PHYS_ONEHOT_SPECS,
+)
+
+
+app = typer.Typer()
+
+
+# CV extra_x 列名到模型列名的映射
+CV_COL_MAPPING = {
+    # Batch_or_individual_or_barcoded -> Sample_organization_type (for Value_name related)
+    "Batch_or_individual_or_barcoded_Barcoded": "Batch_or_individual_or_barcoded_Barcoded",
+    "Batch_or_individual_or_barcoded_Individual": "Batch_or_individual_or_barcoded_Individual",
+    # Helper_lipid_ID_None 不在模型中使用，忽略
+}
+
+
+def load_cv_split(cv_dir: Path) -> Tuple[pd.DataFrame, pd.DataFrame, pd.DataFrame]:
+    """
+    加载单个 CV split 的数据。
+    
+    Args:
+        cv_dir: CV split 目录，如 cv_0/
+        
+    Returns:
+        (train_df, valid_df, test_df) 合并后的 DataFrame
+    """
+    splits = {}
+    for split_name in ["train", "valid", "test"]:
+        # 加载主数据（smiles, quantified_delivery）
+        main_path = cv_dir / f"{split_name}.csv"
+        extra_x_path = cv_dir / f"{split_name}_extra_x.csv"
+        metadata_path = cv_dir / f"{split_name}_metadata.csv"
+        
+        if not main_path.exists():
+            raise FileNotFoundError(f"Missing {main_path}")
+        
+        main_df = pd.read_csv(main_path)
+        
+        # 加载 extra_x（已 one-hot 编码的特征）
+        if extra_x_path.exists():
+            extra_x_df = pd.read_csv(extra_x_path)
+            # 确保行数一致
+            assert len(main_df) == len(extra_x_df), f"Row count mismatch: {len(main_df)} vs {len(extra_x_df)}"
+            # 合并（按行索引）
+            df = pd.concat([main_df, extra_x_df], axis=1)
+        else:
+            df = main_df
+            logger.warning(f"  {split_name}_extra_x.csv not found, using main data only")
+        
+        # 可选：从 metadata 获取额外信息
+        if metadata_path.exists():
+            metadata_df = pd.read_csv(metadata_path)
+            # 提取需要的列（如 Purity, Mix_type, Value_name 等）
+            for col in ["Purity", "Mix_type", "Value_name", "Target_or_delivered_gene"]:
+                if col in metadata_df.columns and col not in df.columns:
+                    df[col] = metadata_df[col]
+        
+        splits[split_name] = df
+    
+    return splits["train"], splits["valid"], splits["test"]
+
+
+def process_cv_dataframe(df: pd.DataFrame) -> pd.DataFrame:
+    """
+    处理 CV 数据的 DataFrame，对齐到模型所需的列格式。
+    
+    CV 数据的 extra_x 已经包含大部分 one-hot 编码，但需要：
+    1. 添加缺失的 one-hot 列（设为 0）
+    2. 从 metadata 中生成 phys token 的 one-hot 列（Purity, Mix_type, Cargo_type, Target_or_delivered_gene）
+    3. 生成 Value_name 的 one-hot 列
+    """
+    df = df.copy()
+    
+    # 1. 处理 comp 列
+    for col in COMP_COLS:
+        if col in df.columns:
+            df[col] = pd.to_numeric(df[col], errors="coerce").fillna(0.0)
+        else:
+            df[col] = 0.0
+    
+    # 2. 处理 help 列
+    for col in HELP_COLS:
+        if col not in df.columns:
+            df[col] = 0.0
+        else:
+            df[col] = df[col].fillna(0.0).astype(float)
+    
+    # 3. 处理 phys token 的 one-hot 列
+    for col, values in PHYS_ONEHOT_SPECS.items():
+        for v in values:
+            onehot_col = f"{col}_{v}"
+            if onehot_col not in df.columns:
+                # 尝试从原始列生成
+                if col in df.columns:
+                    df[onehot_col] = (df[col] == v).astype(float)
+                else:
+                    df[onehot_col] = 0.0
+            else:
+                df[onehot_col] = df[onehot_col].fillna(0.0).astype(float)
+    
+    # 4. 处理 exp token 的 one-hot 列
+    for col, values in EXP_ONEHOT_SPECS.items():
+        for v in values:
+            onehot_col = f"{col}_{v}"
+            if onehot_col not in df.columns:
+                # 尝试从原始列生成
+                if col in df.columns:
+                    df[onehot_col] = (df[col] == v).astype(float)
+                else:
+                    df[onehot_col] = 0.0
+            else:
+                df[onehot_col] = df[onehot_col].fillna(0.0).astype(float)
+    
+    # 5. 处理 quantified_delivery
+    if "quantified_delivery" in df.columns:
+        df["quantified_delivery"] = pd.to_numeric(df["quantified_delivery"], errors="coerce")
+    
+    return df
+
+
+def get_feature_columns() -> List[str]:
+    """获取所有特征列名"""
+    config = LNPDatasetConfig()
+    return (
+        ["smiles"]
+        + config.comp_cols
+        + config.phys_cols
+        + config.help_cols
+        + config.exp_cols
+        + ["quantified_delivery"]
+    )
+
+
+@app.command()
+def main(
+    data_dir: Path = EXTERNAL_DATA_DIR / "all_amine_split_for_LiON",
+    output_dir: Path = PROCESSED_DATA_DIR / "pretrain_cv",
+    n_folds: int = 5,
+):
+    """
+    处理 cross-validation 数据，生成模型所需的 parquet 文件。
+    
+    输出结构:
+        - processed/pretrain_cv/fold_0/train.parquet
+        - processed/pretrain_cv/fold_0/valid.parquet
+        - processed/pretrain_cv/fold_0/test.parquet
+        - processed/pretrain_cv/fold_1/...
+        - processed/pretrain_cv/feature_columns.txt
+    """
+    logger.info(f"Processing CV data from {data_dir}")
+    
+    # 获取所有 cv_* 目录
+    cv_dirs = sorted([d for d in data_dir.iterdir() if d.is_dir() and d.name.startswith("cv_")])
+    
+    if len(cv_dirs) == 0:
+        logger.error(f"No cv_* directories found in {data_dir}")
+        raise typer.Exit(1)
+    
+    if len(cv_dirs) != n_folds:
+        logger.warning(f"Expected {n_folds} folds, found {len(cv_dirs)}")
+    
+    logger.info(f"Found {len(cv_dirs)} folds: {[d.name for d in cv_dirs]}")
+    
+    feature_cols = get_feature_columns()
+    output_dir.mkdir(parents=True, exist_ok=True)
+    
+    for i, cv_dir in enumerate(cv_dirs):
+        fold_name = f"fold_{i}"
+        fold_output_dir = output_dir / fold_name
+        fold_output_dir.mkdir(parents=True, exist_ok=True)
+        
+        logger.info(f"Processing {cv_dir.name} -> {fold_name}")
+        
+        # 加载数据
+        train_df, valid_df, test_df = load_cv_split(cv_dir)
+        
+        logger.info(f"  Loaded: train={len(train_df)}, valid={len(valid_df)}, test={len(test_df)}")
+        
+        # 处理数据
+        train_df = process_cv_dataframe(train_df)
+        valid_df = process_cv_dataframe(valid_df)
+        test_df = process_cv_dataframe(test_df)
+        
+        # 确保所有列存在
+        for col in feature_cols:
+            for df in [train_df, valid_df, test_df]:
+                if col not in df.columns:
+                    df[col] = 0.0 if col != "smiles" else ""
+        
+        # 只保留需要的列
+        train_df = train_df[feature_cols]
+        valid_df = valid_df[feature_cols]
+        test_df = test_df[feature_cols]
+        
+        # 保存
+        train_df.to_parquet(fold_output_dir / "train.parquet", index=False)
+        valid_df.to_parquet(fold_output_dir / "valid.parquet", index=False)
+        test_df.to_parquet(fold_output_dir / "test.parquet", index=False)
+        
+        logger.success(f"  Saved to {fold_output_dir}")
+    
+    # 保存特征列配置
+    cols_path = output_dir / "feature_columns.txt"
+    with open(cols_path, "w") as f:
+        f.write("\n".join(feature_cols))
+    logger.success(f"Saved feature columns to {cols_path}")
+    
+    logger.info("\n" + "=" * 60)
+    logger.info("CV DATA PROCESSING COMPLETE")
+    logger.info("=" * 60)
+    logger.info(f"Output directory: {output_dir}")
+    logger.info(f"Number of folds: {len(cv_dirs)}")
+
+
+if __name__ == "__main__":
+    app()
+