Upload 13 files

antioxidant_predictor_5.py

@@ -0,0 +1,100 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+class AntioxidantPredictor(nn.Module):
+    def __init__(self, input_dim, transformer_layers=3, transformer_heads=4, transformer_dropout=0.1):
+        super(AntioxidantPredictor, self).__init__()
+        self.prott5_dim = 1024
+        self.handcrafted_dim = input_dim - self.prott5_dim
+        self.seq_len = 16
+        self.prott5_feature_dim = 64  # 16 * 64 = 1024
+
+        encoder_layer = nn.TransformerEncoderLayer(
+            d_model=self.prott5_feature_dim,
+            nhead=transformer_heads,
+            dropout=transformer_dropout,
+            batch_first=True
+        )
+        self.transformer_encoder = nn.TransformerEncoder(encoder_layer, num_layers=transformer_layers)
+
+        fused_dim = self.prott5_feature_dim + self.handcrafted_dim
+        self.fusion_fc = nn.Sequential(
+            nn.Linear(fused_dim, 1024),
+            nn.ReLU(),
+            nn.Dropout(0.3),
+            nn.Linear(1024, 512),
+            nn.ReLU(),
+            nn.Dropout(0.3)
+        )
+
+        self.classifier = nn.Sequential(
+            nn.Linear(512, 256),
+            nn.ReLU(),
+            nn.Dropout(0.3),
+            nn.Linear(256, 1)
+        )
+
+        # 温度缩放参数 T
+        # 初始化为1.0，表示在校准前不改变logits
+        # requires_grad=False，因为T通常在模型训练完成后单独优化
+        self.temperature = nn.Parameter(torch.ones(1), requires_grad=False)
+
+    def forward(self, x, *args):
+        batch_size = x.size(0)
+        prot_t5_features = x[:, :self.prott5_dim]
+        handcrafted_features = x[:, self.prott5_dim:]
+        
+        prot_t5_seq = prot_t5_features.view(batch_size, self.seq_len, self.prott5_feature_dim)
+        encoded_seq = self.transformer_encoder(prot_t5_seq)
+        refined_prott5 = encoded_seq.mean(dim=1)
+        
+        fused_features = torch.cat([refined_prott5, handcrafted_features], dim=1)
+        fused_features = self.fusion_fc(fused_features)
+        
+        logits = self.classifier(fused_features)
+        
+        # 应用温度缩放: logits / T
+        # 注意：这里是在获取原始logits后，外部应用sigmoid前进行缩放
+        # 如果要直接输出校准后的概率，可以在这里除以T然后sigmoid
+        # 但通常T的优化和应用是分离的。
+        # 为了在调用模型时就能获得校准的logits（如果T已优化），我们在这里应用它。
+        # 如果T未被优化（仍为1），则此操作无影响。
+        logits_scaled = logits / self.temperature
+        
+        return logits_scaled # 返回校准后（或原始，如果T=1）的logits
+
+    def set_temperature(self, temp_value, device):
+        """用于设置优化后的温度值"""
+        self.temperature = nn.Parameter(torch.tensor([temp_value], device=device), requires_grad=False)
+        print(f"模型温度 T 设置为: {self.temperature.item()}")
+
+    def get_temperature(self):
+        """获取当前温度值"""
+        return self.temperature.item()
+
+if __name__ == "__main__":
+    dummy_input = torch.randn(8, 1914)
+    model = AntioxidantPredictor(input_dim=1914)
+    
+    print(f"初始温度: {model.get_temperature()}")
+    logits_output_initial = model(dummy_input)
+    print("初始 logits shape:", logits_output_initial.shape)
+    probs_initial = torch.sigmoid(logits_output_initial)
+    print("初始概率 (T=1.0):", probs_initial.detach().cpu().numpy()[:2])
+
+    # 模拟设置一个优化后的温度
+    model.set_temperature(1.5, device='cpu') # 假设优化得到 T=1.5
+    print(f"设置后温度: {model.get_temperature()}")
+    logits_output_scaled = model(dummy_input) # 模型内部应用了 T
+    print("缩放后 logits shape:", logits_output_scaled.shape)
+    probs_scaled = torch.sigmoid(logits_output_scaled) # 外部仍然需要 sigmoid
+    print("缩放后概率 (T=1.5):", probs_scaled.detach().cpu().numpy()[:2])
+    
+    # 验证 logits / T 的效果
+    # logits_manual_scale = logits_output_initial / 1.5
+    # probs_manual_scale = torch.sigmoid(logits_manual_scale)
+    # print("手动缩放后概率 (T=1.5):", probs_manual_scale.detach().cpu().numpy()[:2])
+    # assert torch.allclose(probs_scaled, probs_manual_scale) # 应该相等

app.py

@@ -0,0 +1,364 @@
+# app.py - RLAnOxPeptide Gradio Web Application
+# This script integrates both the predictor and generator into a user-friendly web UI.
+
+import os
+import torch
+import torch.nn as nn
+import pandas as pd
+import joblib
+import numpy as np
+import gradio as gr
+from sklearn.cluster import KMeans
+from tqdm import tqdm
+import transformers
+
+# Suppress verbose logging from transformers
+transformers.logging.set_verbosity_error()
+
+# --------------------------------------------------------------------------
+# SECTION 1: CORE CLASS AND FUNCTION DEFINITIONS
+# To make this app self-contained, we copy necessary class definitions here.
+# These should match the versions used during training.
+# --------------------------------------------------------------------------
+
+# --- Vocabulary Definition (from both scripts) ---
+AMINO_ACIDS = "ACDEFGHIKLMNPQRSTVWY"
+token2id = {aa: i + 2 for i, aa in enumerate(AMINO_ACIDS)}
+token2id["<PAD>"] = 0
+token2id["<EOS>"] = 1
+id2token = {i: t for t, i in token2id.items()}
+VOCAB_SIZE = len(token2id)
+
+# --- Predictor Model Architecture (from antioxidant_predictor_5.py) ---
+class AntioxidantPredictor(nn.Module):
+    # This class definition should be an exact copy from your project
+    def __init__(self, input_dim, transformer_layers=3, transformer_heads=4, transformer_dropout=0.1):
+        super(AntioxidantPredictor, self).__init__()
+        self.input_dim = input_dim
+        self.t5_dim = 1024
+        self.hand_crafted_dim = self.input_dim - self.t5_dim
+        
+        encoder_layer = nn.TransformerEncoderLayer(
+            d_model=self.t5_dim, nhead=transformer_heads, 
+            dropout=transformer_dropout, batch_first=True
+        )
+        self.transformer_encoder = nn.TransformerEncoder(encoder_layer, num_layers=transformer_layers)
+        
+        self.mlp = nn.Sequential(
+            nn.Linear(self.input_dim, 512),
+            nn.ReLU(),
+            nn.Dropout(0.5),
+            nn.Linear(512, 256),
+            nn.ReLU(),
+            nn.Dropout(0.5),
+            nn.Linear(256, 1)
+        )
+        self.temperature = nn.Parameter(torch.ones(1))
+
+    def forward(self, fused_features):
+        tr_features = fused_features[:, :self.t5_dim]
+        hand_features = fused_features[:, self.t5_dim:]
+        tr_features_unsqueezed = tr_features.unsqueeze(1)
+        transformer_output = self.transformer_encoder(tr_features_unsqueezed)
+        transformer_output_pooled = transformer_output.mean(dim=1)
+        combined_features = torch.cat((transformer_output_pooled, hand_features), dim=1)
+        logits = self.mlp(combined_features)
+        return logits / self.temperature
+
+    def get_temperature(self):
+        return self.temperature.item()
+
+# --- Generator Model Architecture (from generator.py) ---
+class ProtT5Generator(nn.Module):
+    # This class definition should be an exact copy from your project
+    def __init__(self, vocab_size, embed_dim=512, num_layers=6, num_heads=8, dropout=0.1):
+        super(ProtT5Generator, self).__init__()
+        self.embed_tokens = nn.Embedding(vocab_size, embed_dim, padding_idx=token2id["<PAD>"])
+        encoder_layer = nn.TransformerEncoderLayer(d_model=embed_dim, nhead=num_heads, dropout=dropout, batch_first=True)
+        self.encoder = nn.TransformerEncoder(encoder_layer, num_layers=num_layers)
+        self.lm_head = nn.Linear(embed_dim, vocab_size)
+        self.vocab_size = vocab_size
+        self.eos_token_id = token2id["<EOS>"]
+        self.pad_token_id = token2id["<PAD>"]
+    
+    def forward(self, input_ids):
+        embeddings = self.embed_tokens(input_ids)
+        encoder_output = self.encoder(embeddings)
+        logits = self.lm_head(encoder_output)
+        return logits
+    
+    def sample(self, batch_size, max_length=20, device="cpu", temperature=2.5, min_decoded_length=3):
+        start_token = torch.randint(2, self.vocab_size, (batch_size, 1), device=device)
+        generated = start_token
+        for _ in range(max_length - 1):
+            logits = self.forward(generated)
+            next_logits = logits[:, -1, :] / temperature
+            if generated.size(1) < min_decoded_length:
+                next_logits[:, self.eos_token_id] = -float("inf")
+            
+            probs = torch.softmax(next_logits, dim=-1)
+            next_token = torch.multinomial(probs, num_samples=1)
+            generated = torch.cat((generated, next_token), dim=1)
+            
+            if (next_token == self.eos_token_id).all():
+                break
+        return generated
+
+    def decode(self, token_ids_batch):
+        seqs = []
+        for ids_tensor in token_ids_batch:
+            seq = ""
+            # Start from index 1 to skip the initial random start token
+            for token_id in ids_tensor.tolist()[1:]:
+                if token_id == self.eos_token_id: break
+                if token_id == self.pad_token_id: continue
+                seq += id2token.get(token_id, "?")
+            seqs.append(seq)
+        return seqs
+
+# --- Feature Extraction Logic (from feature_extract.py) ---
+# Note: You need the actual ProtT5Model and extract_features here. 
+# Assuming they are in a file named `feature_extract.py` in the same directory.
+try:
+    from feature_extract import ProtT5Model as FeatureProtT5Model, extract_features
+except ImportError:
+    raise gr.Error("Failed to import feature_extract.py. Please ensure the file is in the same directory as app.py.")
+
+# --- Clustering Logic (from generator.py) ---
+def cluster_sequences(generator, sequences, num_clusters, device):
+    if not sequences or len(sequences) < num_clusters:
+        return sequences[:num_clusters]
+    with torch.no_grad():
+        token_ids_list = []
+        max_len = max(len(seq) for seq in sequences) + 2 # Start token + EOS
+        for seq in sequences:
+            # Recreate encoding to match how generator sees it (with start token)
+            ids = [token2id.get(aa, 0) for aa in seq] + [generator.eos_token_id]
+            ids = [np.random.randint(2, VOCAB_SIZE)] + ids # Add a dummy start token
+            ids += [token2id["<PAD>"]] * (max_len - len(ids))
+            token_ids_list.append(ids)
+
+        input_ids = torch.tensor(token_ids_list, dtype=torch.long, device=device)
+        embeddings = generator.embed_tokens(input_ids)
+        mask = (input_ids != token2id["<PAD>"]).unsqueeze(-1).float()
+        embeddings = embeddings * mask
+        lengths = mask.sum(dim=1)
+        seq_embeds = embeddings.sum(dim=1) / (lengths + 1e-9)
+        seq_embeds_np = seq_embeds.cpu().numpy()
+
+    kmeans = KMeans(n_clusters=int(num_clusters), random_state=42, n_init='auto').fit(seq_embeds_np)
+    representatives = []
+    for i in range(int(num_clusters)):
+        indices = np.where(kmeans.labels_ == i)[0]
+        if len(indices) == 0: continue
+        cluster_center = kmeans.cluster_centers_[i]
+        cluster_embeddings = seq_embeds_np[indices]
+        distances = np.linalg.norm(cluster_embeddings - cluster_center, axis=1)
+        representative_index = indices[np.argmin(distances)]
+        representatives.append(sequences[representative_index])
+    return representatives
+
+
+# --------------------------------------------------------------------------
+# SECTION 2: GLOBAL MODEL LOADING
+# Load all models and dependencies once when the app starts.
+# --------------------------------------------------------------------------
+print("Loading all models and dependencies. Please wait...")
+DEVICE = "cpu"  # Use CPU for compatibility with Hugging Face free tier
+
+try:
+    # --- Define all required file paths here ---
+    # !! IMPORTANT: Ensure these are relative paths to the files in your Space !!
+    PREDICTOR_CHECKPOINT_PATH = "checkpoints/final_rl_model_logitp0.1_calibrated_FINETUNED_PROTT5.pth"
+    SCALER_PATH = "checkpoints/scaler_FINETUNED_PROTT5.pkl"
+    GENERATOR_CHECKPOINT_PATH = "generator_checkpoints_v3.6/final_generator_model.pth"
+    PROTT5_BASE_MODEL_PATH = "prott5/model/"
+    FINETUNED_PROTT5_FOR_FEATURES_PATH = "prott5/model/finetuned_prott5.bin"
+
+    # --- Load Predictor Components ---
+    print("Loading Predictor Model...")
+    PREDICTOR_MODEL = AntioxidantPredictor(
+        input_dim=1914, transformer_layers=3, transformer_heads=4, transformer_dropout=0.1
+    )
+    PREDICTOR_MODEL.load_state_dict(torch.load(PREDICTOR_CHECKPOINT_PATH, map_location=DEVICE))
+    PREDICTOR_MODEL.to(DEVICE)
+    PREDICTOR_MODEL.eval()
+    print("✅ Predictor model loaded.")
+
+    print("Loading Scaler...")
+    SCALER = joblib.load(SCALER_PATH)
+    print("✅ Scaler loaded.")
+
+    print("Loading ProtT5 Feature Extractor...")
+    # This extractor must use the fine-tuned model for features, as per your training logic
+    PROTT5_EXTRACTOR = FeatureProtT5Model(
+        model_path=PROTT5_BASE_MODEL_PATH,
+        finetuned_model_file=FINETUNED_PROTT5_FOR_FEATURES_PATH
+    )
+    print("✅ ProtT5 Feature Extractor loaded.")
+
+    # --- Load Generator Model ---
+    print("Loading Generator Model...")
+    GENERATOR_MODEL = ProtT5Generator(
+        vocab_size=VOCAB_SIZE, embed_dim=512, num_layers=6, num_heads=8, dropout=0.1
+    )
+    GENERATOR_MODEL.load_state_dict(torch.load(GENERATOR_CHECKPOINT_PATH, map_location=DEVICE))
+    GENERATOR_MODEL.to(DEVICE)
+    GENERATOR_MODEL.eval()
+    print("✅ Generator model loaded.")
+
+    print("\n--- All models loaded successfully! Gradio app is ready. ---\n")
+
+except Exception as e:
+    print(f"💥 FATAL ERROR: Failed to load a model or dependency file: {e}")
+    raise gr.Error(f"Model or dependency loading failed! Check file paths and integrity. Error: {e}")
+
+# --------------------------------------------------------------------------
+# SECTION 3: WRAPPER FUNCTIONS FOR GRADIO
+# These functions connect the UI to our model's logic.
+# --------------------------------------------------------------------------
+
+def predict_peptide_wrapper(sequence_str):
+    """Takes a peptide sequence string and returns its predicted probability and class."""
+    if not sequence_str or not isinstance(sequence_str, str) or any(c not in AMINO_ACIDS for c in sequence_str.upper()):
+        return "0.0000", "Error: Please enter a valid sequence with standard amino acids."
+    
+    try:
+        # 1. Extract features using the same logic as training/prediction scripts
+        features = extract_features(sequence_str, PROTT5_EXTRACTOR)
+        
+        # 2. Scale features
+        scaled_features = SCALER.transform(features.reshape(1, -1))
+        
+        # 3. Predict with the model
+        with torch.no_grad():
+            features_tensor = torch.tensor(scaled_features, dtype=torch.float32).to(DEVICE)
+            logits = PREDICTOR_MODEL(features_tensor)
+            probability = torch.sigmoid(logits).squeeze().item()
+        
+        classification = "Antioxidant" if probability >= 0.5 else "Non-Antioxidant"
+        return f"{probability:.4f}", classification
+
+    except Exception as e:
+        print(f"Prediction error for sequence '{sequence_str}': {e}")
+        return "N/A", f"An error occurred during processing: {e}"
+
+def generate_peptide_wrapper(num_to_generate, min_len, max_len, temperature, diversity_factor, progress=gr.Progress(track_tqdm=True)):
+    """Generates, validates, and clusters sequences."""
+    num_to_generate = int(num_to_generate)
+    min_len = int(min_len)
+    max_len = int(max_len)
+    
+    try:
+        # STEP 1: Generate an initial pool of unique sequences
+        target_pool_size = int(num_to_generate * diversity_factor)
+        unique_seqs = set()
+        progress(0, desc="Generating initial peptide pool...")
+        
+        max_attempts = 10
+        attempts = 0
+        while len(unique_seqs) < target_pool_size and attempts < max_attempts:
+            batch_size = (target_pool_size - len(unique_seqs)) * 2 # Generate extra to account for duplicates/short ones
+            with torch.no_grad():
+                generated_tokens = GENERATOR_MODEL.sample(
+                    batch_size=batch_size, 
+                    max_length=max_len, 
+                    device=DEVICE,
+                    temperature=temperature, 
+                    min_decoded_length=min_len
+                )
+            decoded = GENERATOR_MODEL.decode(generated_tokens.cpu())
+            for seq in decoded:
+                if min_len <= len(seq) <= max_len:
+                    unique_seqs.add(seq)
+            attempts += 1
+            progress(len(unique_seqs) / target_pool_size, desc=f"Generated {len(unique_seqs)} unique sequences...")
+
+        candidate_seqs = list(unique_seqs)
+        if not candidate_seqs:
+            return pd.DataFrame({"Sequence": ["Failed to generate valid sequences."], "Predicted Probability": ["N/A"]})
+
+        # STEP 2: Validate the generated sequences
+        validated_pool = {}
+        for seq in tqdm(candidate_seqs, desc="Validating generated sequences"):
+            prob_str, _ = predict_peptide_wrapper(seq)
+            try:
+                prob = float(prob_str)
+                if prob > 0.90: # Filter for high-quality peptides as in generator.py
+                    validated_pool[seq] = prob
+            except (ValueError, TypeError):
+                continue
+
+        if not validated_pool:
+            return pd.DataFrame({"Sequence": ["No high-activity peptides (>0.9 prob) were generated."], "Predicted Probability": ["N/A"]})
+        
+        high_quality_sequences = list(validated_pool.keys())
+        
+        # STEP 3: Cluster to ensure diversity
+        progress(1.0, desc="Clustering for diversity...")
+        final_diverse_seqs = cluster_sequences(GENERATOR_MODEL, high_quality_sequences, num_to_generate, DEVICE)
+        
+        # STEP 4: Format final results
+        final_results = [(seq, f"{validated_pool[seq]:.4f}") for seq in final_diverse_seqs]
+        final_results.sort(key=lambda x: float(x[1]), reverse=True)
+        
+        return pd.DataFrame(final_results, columns=["Sequence", "Predicted Probability"])
+
+    except Exception as e:
+        print(f"Generation error: {e}")
+        return pd.DataFrame({"Sequence": [f"An error occurred during generation: {e}"], "Predicted Probability": ["N/A"]})
+
+
+# --------------------------------------------------------------------------
+# SECTION 4: GRADIO UI CONSTRUCTION
+# Building the web interface. All text is in English.
+# --------------------------------------------------------------------------
+with gr.Blocks(theme=gr.themes.Soft(), title="RLAnOxPeptide") as demo:
+    gr.Markdown("# RLAnOxPeptide: Intelligent Peptide Design and Prediction Platform")
+    gr.Markdown("An integrated framework combining reinforcement learning and a Transformer model for the efficient prediction and innovative design of antioxidant peptides.")
+
+    with gr.Tabs():
+        with gr.TabItem("Peptide Activity Predictor"):
+            gr.Markdown("### Enter an amino acid sequence to predict its antioxidant activity.")
+            with gr.Row():
+                peptide_input = gr.Textbox(label="Peptide Sequence", placeholder="e.g., WHYHDYKY", scale=3)
+                predict_button = gr.Button("Predict", variant="primary", scale=1)
+            with gr.Row():
+                probability_output = gr.Textbox(label="Predicted Probability")
+                class_output = gr.Textbox(label="Predicted Class")
+            
+            predict_button.click(
+                fn=predict_peptide_wrapper,
+                inputs=peptide_input,
+                outputs=[probability_output, class_output]
+            )
+            gr.Examples(
+                examples=[["WHYHDYKY"], ["YPGG"], ["LVLHEHGGN"], ["INVALIDSEQUENCE"]],
+                inputs=peptide_input,
+                outputs=[probability_output, class_output],
+                fn=predict_peptide_wrapper,
+                cache_examples=False,
+            )
+
+        with gr.TabItem("Novel Sequence Generator"):
+            gr.Markdown("### Set parameters to generate novel, high-activity antioxidant peptides.")
+            with gr.Column():
+                with gr.Row():
+                    num_input = gr.Slider(minimum=1, maximum=50, value=10, step=1, label="Number of Final Peptides to Generate")
+                    min_len_input = gr.Slider(minimum=2, maximum=10, value=3, step=1, label="Minimum Length")
+                    max_len_input = gr.Slider(minimum=10, maximum=20, value=20, step=1, label="Maximum Length")
+                with gr.Row():
+                    temp_input = gr.Slider(minimum=0.5, maximum=3.0, value=2.5, step=0.1, label="Temperature (Higher = More random)")
+                    diversity_input = gr.Slider(minimum=1.0, maximum=3.0, value=1.2, step=0.1, label="Diversity Factor (Higher = Larger initial pool for clustering)")
+            
+            generate_button = gr.Button("Generate Peptides", variant="primary")
+            results_output = gr.DataFrame(headers=["Sequence", "Predicted Probability"], label="Generated & Validated Peptides", wrap=True)
+
+            generate_button.click(
+                fn=generate_peptide_wrapper,
+                inputs=[num_input, min_len_input, max_len_input, temp_input, diversity_input],
+                outputs=results_output
+            )
+
+if __name__ == "__main__":
+    demo.launch()

checkpoints/final_rl_model_logitp0.1_calibrated_FINETUNED_PROTT5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1cb2c7cf6a59a5d028af4678ac5aec83305d3638bf9f43fee6df14f162f4a4e6
+size 9930081

checkpoints/scaler_FINETUNED_PROTT5.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:50cfe7b0204a4e1dd5f22dc1440d6e8448f7ee323bb20957c70ce1058ca78475
+size 31127

feature_extract.py

@@ -0,0 +1,341 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+import os
+import numpy as np
+import random
+import pandas as pd
+from Bio.SeqUtils.ProtParam import ProteinAnalysis
+from sklearn.model_selection import train_test_split
+# from sklearn.preprocessing import StandardScaler # 不再使用 StandardScaler
+from sklearn.preprocessing import RobustScaler # 导入 RobustScaler
+import torch
+from transformers import T5EncoderModel, T5Tokenizer 
+
+# ProtT5Model, load_fasta, load_fasta_with_labels, 
+# compute_amino_acid_composition, compute_reducing_aa_ratio, 
+# compute_physicochemical_properties, compute_electronic_features,
+# compute_dimer_frequency, positional_encoding, perturb_sequence,
+# generate_adversarial_samples, extract_features 函数与您之前提供的版本相同。
+# 为保持简洁，此处省略这些函数的代码。请确保它们在您的文件中是完整的。
+# 您可以从之前的日志或您本地的文件中复制这些函数。
+# 以下是 prepare_features 函数的修改版，以及其他函数的占位符。
+
+class ProtT5Model:
+    """
+    从本地加载 ProtT5 模型。如果 finetuned_model_file 不为空，则加载微调后的权重（使用 strict=False）。
+    """
+    def __init__(self, model_path, finetuned_model_file=None):
+        self.device = "cuda" if torch.cuda.is_available() else "cpu"
+        # 尝试加载本地文件，如果失败，transformers库可能会尝试从hub下载（取决于配置）
+        try:
+            self.tokenizer = T5Tokenizer.from_pretrained(model_path, do_lower_case=False, local_files_only=True)
+            self.model = T5EncoderModel.from_pretrained(model_path, local_files_only=True)
+        except OSError: # OSError: Can't load tokenizer for '...'. If you were trying to load it from 'https://huggingface.co/models', make sure you don't have a local directory with the same name. Otherwise, make sure '...' is the correct path to a directory containing all relevant files for a T5Tokenizer tokenizer.
+            print(f"警告: 无法从本地路径 {model_path} 加载ProtT5模型/分词器。尝试从HuggingFace Hub下载（如果transformers配置允许）。")
+            self.tokenizer = T5Tokenizer.from_pretrained(model_path.split('/')[-1] if '/' in model_path else model_path, do_lower_case=False) # 尝试使用模型名下载
+            self.model = T5EncoderModel.from_pretrained(model_path.split('/')[-1] if '/' in model_path else model_path)
+
+
+        if finetuned_model_file is not None and os.path.exists(finetuned_model_file):
+            try:
+                state_dict = torch.load(finetuned_model_file, map_location=self.device)
+                missing_keys, unexpected_keys = self.model.load_state_dict(state_dict, strict=False)
+                print(f"加载微调权重 {finetuned_model_file}：缺失键 {missing_keys}, 意外键 {unexpected_keys}")
+            except Exception as e:
+                print(f"加载微调权重 {finetuned_model_file} 失败: {e}")
+        
+        self.model.to(self.device)
+        self.model.eval()
+
+    def encode(self, sequence):
+        if not sequence or not isinstance(sequence, str): # 增加对空序列或非字符串的检查
+            print(f"警告: ProtT5Model.encode 接收到无效序列: {sequence}")
+            # 返回一个零向量或根据需要处理错误
+            # 假设 ProtT5 输出维度为 1024 (embedding.shape[1])
+            # 假设序列处理后平均池化，所以返回 (1024,)
+            # 但 encode 返回的是 (seq_len, hidden_dim)，所以这里返回一个模拟的短序列零嵌入
+            return np.zeros((1, 1024), dtype=np.float32) # (1, hidden_dim)
+
+        seq_spaced = " ".join(list(sequence)) # 修改变量名以避免覆盖外部seq
+        try:
+            encoded_input = self.tokenizer(seq_spaced, return_tensors='pt', padding=True, truncation=True, max_length=1022) # ProtT5通常最大长度1024，tokenized后可能更长
+        except Exception as e:
+            print(f"分词失败序列 '{sequence[:30]}...': {e}")
+            return np.zeros((1, 1024), dtype=np.float32)
+
+        encoded_input = {k: v.to(self.device) for k, v in encoded_input.items()}
+        with torch.no_grad():
+            try:
+                embedding = self.model(**encoded_input).last_hidden_state  # (batch_size, seq_len, hidden_dim)
+            except Exception as e:
+                print(f"ProtT5模型推理失败序列 '{sequence[:30]}...': {e}")
+                return np.zeros((1, 1024), dtype=np.float32)
+
+        emb = embedding.squeeze(0).cpu().numpy()  # (seq_len, hidden_dim)
+        if emb.shape[0] == 0: # 如果由于某种原因序列长度为0
+             return np.zeros((1, 1024), dtype=np.float32)
+        return emb
+
+# --- (此处应包含您之前版本中所有其他的特征提取辅助函数) ---
+# load_fasta, load_fasta_with_labels, compute_amino_acid_composition, ... extract_features
+# 为确保完整性��请从您本地的 feature_extract.py 文件中复制这些函数到这里。
+# 下面是这些函数的一个简化占位符，您需要用实际的函数替换它们。
+
+def load_fasta(fasta_file):
+    # (您的 load_fasta 实现)
+    sequences = []
+    try:
+        with open(fasta_file, 'r') as f:
+            current_seq_lines = []
+            for line in f:
+                line = line.strip()
+                if not line: continue
+                if line.startswith(">"):
+                    if current_seq_lines: sequences.append("".join(current_seq_lines))
+                    current_seq_lines = []
+                else: current_seq_lines.append(line)
+            if current_seq_lines: sequences.append("".join(current_seq_lines))
+    except FileNotFoundError: print(f"文件未找到: {fasta_file}"); return []
+    return sequences
+
+def load_fasta_with_labels(fasta_file):
+    # (您的 load_fasta_with_labels 实现)
+    sequences, labels = [], []
+    try:
+        with open(fasta_file, 'r') as f:
+            current_seq_lines, current_label = [], None
+            for line in f:
+                line = line.strip()
+                if not line: continue
+                if line.startswith(">"):
+                    if current_seq_lines:
+                        sequences.append("".join(current_seq_lines))
+                        labels.append(current_label if current_label is not None else 0) # Default label 0
+                    current_seq_lines = []
+                    current_label = int(line[1]) if len(line) > 1 and line[1] in ['0', '1'] else 0
+                else: current_seq_lines.append(line)
+            if current_seq_lines:
+                sequences.append("".join(current_seq_lines))
+                labels.append(current_label if current_label is not None else 0)
+    except FileNotFoundError: print(f"文件未找到: {fasta_file}"); return [],[]
+    return sequences, labels
+
+
+def compute_amino_acid_composition(seq):
+    if not seq: return {aa: 0.0 for aa in "ACDEFGHIKLMNPQRSTVWY"}
+    # (您的 compute_amino_acid_composition 实现)
+    amino_acids = "ACDEFGHIKLMNPQRSTVWY"
+    seq_len = len(seq)
+    return {aa: seq.upper().count(aa) / seq_len for aa in amino_acids}
+
+
+def compute_reducing_aa_ratio(seq):
+    if not seq: return 0.0
+    # (您的 compute_reducing_aa_ratio 实现)
+    reducing = ['C', 'M', 'W']
+    return sum(seq.upper().count(aa) for aa in reducing) / len(seq) if len(seq) > 0 else 0.0
+
+def compute_physicochemical_properties(seq):
+    if not seq or not all(c.upper() in "ACDEFGHIKLMNPQRSTVWYXUBZ" for c in seq): # ProteinAnalysis might fail on invalid chars
+        return 0.0, 0.0, 0.0 # Default values
+    try:
+        analysis = ProteinAnalysis(str(seq).upper().replace('X','A').replace('U','C').replace('B','N').replace('Z','Q')) # Replace non-standard with common ones for analysis
+        return analysis.gravy(), analysis.isoelectric_point(), analysis.molecular_weight()
+    except Exception: # Catch any error from ProteinAnalysis
+        return 0.0, 7.0, 110.0 * len(seq) # Rough defaults
+
+def compute_electronic_features(seq):
+    if not seq: return 0.0, 0.0
+    # (您的 compute_electronic_features 实现)
+    electronegativity = {'A':1.8,'C':2.5,'D':3.0,'E':3.2,'F':2.8,'G':1.6,'H':2.4,'I':4.5,'K':3.0,'L':4.2,'M':4.5,'N':2.0,'P':3.5,'Q':3.5,'R':2.5,'S':1.8,'T':2.5,'V':4.0,'W':5.0,'Y':4.0}
+    values = [electronegativity.get(aa.upper(), 2.5) for aa in seq]
+    avg_val = sum(values) / len(values) if values else 2.5
+    return avg_val + 0.1, avg_val - 0.1
+
+
+def compute_dimer_frequency(seq):
+    if len(seq) < 2: return np.zeros(400) # 20*20
+    # (您的 compute_dimer_frequency 实现)
+    amino_acids = "ACDEFGHIKLMNPQRSTVWY"
+    dimer_counts = {aa1+aa2: 0 for aa1 in amino_acids for aa2 in amino_acids}
+    for i in range(len(seq) - 1):
+        dimer = seq[i:i+2].upper()
+        if dimer in dimer_counts: dimer_counts[dimer] += 1
+    total = max(len(seq) - 1, 1)
+    for key in dimer_counts: dimer_counts[key] /= total
+    return np.array([dimer_counts[d] for d in sorted(dimer_counts.keys())])
+
+
+def positional_encoding(seq_len_actual, L_fixed=29, d_model=16): # Pass actual sequence length or use L_fixed
+    # (您的 positional_encoding 实现)
+    # This PE is fixed length, not dependent on actual seq len if L_fixed is used.
+    # For random short sequences, this fixed PE might be an issue.
+    # A more dynamic PE or no PE for very short sequences might be better.
+    # However, to match current model input, we keep it.
+    pos_enc = np.zeros((L_fixed, d_model))
+    for pos in range(L_fixed):
+        for i in range(d_model):
+            angle = pos / (10000 ** (2 * (i // 2) / d_model))
+            pos_enc[pos, i] = np.sin(angle) if i % 2 == 0 else np.cos(angle)
+    return pos_enc.flatten()
+
+
+def perturb_sequence(seq, perturb_rate=0.1, critical=['C', 'M', 'W']):
+    # (您的 perturb_sequence 实现)
+    if not seq: return ""
+    seq_list = list(seq)
+    amino_acids = "ACDEFGHIKLMNPQRSTVWY"
+    for i, aa in enumerate(seq_list):
+        if aa.upper() not in critical and random.random() < perturb_rate:
+            seq_list[i] = random.choice([x for x in amino_acids if x != aa.upper()])
+    return "".join(seq_list)
+
+
+def extract_features(seq, prott5_model_instance, L_fixed=29, d_model_pe=16): # Renamed d_model to d_model_pe
+    if not seq or not isinstance(seq, str) or len(seq) == 0:
+        print(f"警告: extract_features 接收到空或无效序列。返回零特征。")
+        # 返回一个与预期特征维度匹配的零向量
+        # 1024 (protT5) + 20 (aac) + 1 (red_ratio) + 3 (phys) + 2 (elec) + 400 (dimer) + L_fixed*d_model_pe (pos_enc)
+        # Example: 1024 + 20 + 1 + 3 + 2 + 400 + 29*16 = 1024 + 20 + 1 + 3 + 2 + 400 + 464 = 1914
+        return np.zeros(1024 + 20 + 1 + 3 + 2 + 400 + (L_fixed * d_model_pe))
+
+
+    embedding = prott5_model_instance.encode(seq) # prott5_model is now an instance
+    prot_embed = np.mean(embedding, axis=0) if embedding.shape[0] > 0 else np.zeros(embedding.shape[1] if embedding.ndim > 1 else 1024) # Handle empty embedding
+    if prot_embed.shape[0] != 1024: # Ensure consistent ProtT5 embedding dim
+        # print(f"警告: ProtT5 嵌入维度异常 ({prot_embed.shape[0]}) for seq '{seq[:20]}...'. 使用零向量。")
+        prot_embed = np.zeros(1024)
+
+
+    aa_comp = compute_amino_acid_composition(seq)
+    aa_comp_vector = np.array([aa_comp[aa] for aa in "ACDEFGHIKLMNPQRSTVWY"])
+    red_ratio = np.array([compute_reducing_aa_ratio(seq)])
+    gravy, pI, mol_weight = compute_physicochemical_properties(seq)
+    phys_props = np.array([gravy, pI, mol_weight])
+    HOMO, LUMO = compute_electronic_features(seq)
+    electronic = np.array([HOMO, LUMO])
+    dimer_vector = compute_dimer_frequency(seq)
+    pos_enc = positional_encoding(len(seq), L_fixed, d_model_pe) # Pass actual length, though current PE uses L_fixed
+    
+    features = np.concatenate([prot_embed, aa_comp_vector, red_ratio, phys_props, electronic, dimer_vector, pos_enc])
+    return features
+
+##############################################
+# 主接口函数 prepare_features
+##############################################
+def prepare_features(neg_fasta, pos_fasta, prott5_model_path, additional_params=None):
+    neg_seqs = load_fasta(neg_fasta)
+    pos_seqs = load_fasta(pos_fasta)
+    
+    if not neg_seqs and not pos_seqs:
+        raise ValueError("未能从FASTA文件加载任何序列。请检查文件路径和内容。")
+
+    neg_labels = [0] * len(neg_seqs)
+    pos_labels = [1] * len(pos_seqs)
+    sequences = neg_seqs + pos_seqs
+    labels = neg_labels + pos_labels
+
+    combined = list(zip(sequences, labels))
+    random.shuffle(combined)
+    sequences, labels = zip(*combined)
+    sequences = list(sequences)
+    labels = list(labels)
+
+    train_seqs, val_seqs, train_labels, val_labels = train_test_split(
+        sequences, labels, test_size=0.1, random_state=42, stratify=labels if len(np.unique(labels)) > 1 else None
+    )
+    print("训练集原始样本数：", len(train_seqs))
+    print("验证集原始样本数：", len(val_seqs))
+
+    if additional_params is not None and additional_params.get("augment", False):
+        # (数据增强逻辑 - 如果启用)
+        augmented_seqs, augmented_labels = [], []
+        perturb_rate = additional_params.get("perturb_rate", 0.1)
+        for seq, label in zip(train_seqs, train_labels):
+            aug_seq = perturb_sequence(seq, perturb_rate=perturb_rate)
+            augmented_seqs.append(aug_seq)
+            augmented_labels.append(label)
+        train_seqs.extend(augmented_seqs)
+        train_labels.extend(augmented_labels)
+        print("数据增强后训练集样本数：", len(train_seqs))
+
+
+    finetuned_model_file = additional_params.get("finetuned_model_file") if additional_params else None
+    # 创建 ProtT5Model 实例
+    prott5_model_instance = ProtT5Model(prott5_model_path, finetuned_model_file=finetuned_model_file)
+
+    def process_data(seqs_list): # Renamed seqs to seqs_list
+        feature_list = []
+        for s_item in seqs_list: # Renamed s to s_item
+            # 将 ProtT5Model 实例传递给 extract_features
+            features = extract_features(s_item, prott5_model_instance) 
+            feature_list.append(features)
+        return np.array(feature_list)
+
+    X_train = process_data(train_seqs)
+    X_val = process_data(val_seqs)
+    
+    if X_train.shape[0] == 0 or X_val.shape[0] == 0:
+        raise ValueError("特征提取后训练集或验证集为空。请检查序列数据和特征提取过程。")
+
+
+    # --- **关键修改：使用 RobustScaler** ---
+    # scaler = StandardScaler() # 原来的 StandardScaler
+    scaler = RobustScaler() 
+    print("使用 RobustScaler 进行特征归一化。")
+    
+    X_train_scaled = scaler.fit_transform(X_train)
+    X_val_scaled = scaler.transform(X_val)
+    
+    return X_train_scaled, X_val_scaled, np.array(train_labels), np.array(val_labels), scaler
+
+if __name__ == "__main__":
+    # 确保测试时使用的路径是有效的，或者创建虚拟文件
+    neg_fasta_test = "dummy_data/test_neg.fasta"
+    pos_fasta_test = "dummy_data/test_pos.fasta"
+    prott5_path_test = "dummy_prott5_model/" # 需要一个包含config.json, pytorch_model.bin等的目录结构
+    
+    os.makedirs("dummy_data", exist_ok=True)
+    os.makedirs(prott5_path_test, exist_ok=True) # 创建虚拟模型目录
+
+    if not os.path.exists(neg_fasta_test):
+        with open(neg_fasta_test, "w") as f: f.write(">neg1\nKALKALKALK\n>neg2\nPEPTPEPT\n")
+    if not os.path.exists(pos_fasta_test):
+        with open(pos_fasta_test, "w") as f: f.write(">pos1\nAOPPEPTIDE\n>pos2\nTRYTRYTRY\n")
+    
+    # 为了让ProtT5Model能加载，需要模拟一个最小的transformers模型目录结构
+    # 通常至少需要 config.json, pytorch_model.bin (或 tf_model.h5), tokenizer_config.json, spiece.model
+    # 这里我们只创建目录，实际加载可能会失败，除非transformers库能从模型名下载
+    # 或者您提供一个真实的本地ProtT5模型路径
+    if not os.listdir(prott5_path_test): # 如果目录为空
+        print(f"警告: {prott5_path_test} 为空。ProtT5Model可能尝试从HuggingFace Hub下载模型。")
+        print(f"请确保您已下载Rostlab/ProstT5-XL-UniRef50或类似模型到该路径，或使用其HuggingFace名称。")
+        # 作为演示，我们假设用户会提供一个有效的路径或transformers可以处理它
+        # 如果要完全本地运行而不下载，需要填充该目录。
+
+    additional_params_test = {
+        "augment": False, 
+        "perturb_rate": 0.1, 
+        "finetuned_model_file": None # 测试时不使用微调模型
+    }
+    
+    print("开始测试 prepare_features (使用RobustScaler)...")
+    try:
+        X_train_t, X_val_t, y_train_t, y_val_t, scaler_t = prepare_features(
+            neg_fasta_test, 
+            pos_fasta_test, 
+            "Rostlab/ProstT5-XL-UniRef50", # 使用HuggingFace模型名称，如果本地路径无效
+            additional_params_test
+        )
+        print("prepare_features 测试完成。")
+        print("训练集样本数：", X_train_t.shape[0])
+        print("验证集样本数：", X_val_t.shape[0])
+        if X_train_t.shape[0] > 0:
+            print("训练集特征维度:", X_train_t.shape[1])
+            print("一个缩放后的训练样本 (前5个特征):", X_train_t[0, :5])
+        if scaler_t:
+            print("Scaler类型:", type(scaler_t))
+    except Exception as e:
+        print(f"prepare_features 测试失败: {e}")
+        print("这可能是由于无法加载ProtT5模型或FASTA文件处理问题。请检查路径和文件内容。")
+

generator_checkpoints_v3.6/final_generator_model.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d878934b93514142df36cb04c812af25ad53a79bdfa0a3edd9f65c22ba4e1d48
+size 75775354

prott5/model/config.json

@@ -0,0 +1,25 @@
+{
+  "architectures": [
+    "T5ForConditionalGeneration"
+  ],
+  "d_ff": 16384,
+  "d_kv": 128,
+  "d_model": 1024,
+  "decoder_start_token_id": 0,
+  "dropout_rate": 0.1,
+  "eos_token_id": 1,
+  "feed_forward_proj": "relu",
+  "initializer_factor": 1.0,
+  "is_encoder_decoder": true,
+  "layer_norm_epsilon": 1e-06,
+  "model_type": "t5",
+  "n_positions": 512,
+  "num_decoder_layers": 24,
+  "num_heads": 32,
+  "num_layers": 24,
+  "output_past": true,
+  "pad_token_id": 0,
+  "relative_attention_num_buckets": 32,
+  "use_cache": true,
+  "vocab_size": 128
+}

prott5/model/finetuned_prott5.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3dbe43382d7c4aa0ba6de0ec02a9680155486a3e8984c82199ae8868223eba9e
+size 75774969

prott5/model/gitattributes

@@ -0,0 +1,16 @@
+*.bin.* filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tar.gz filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text

prott5/model/special_tokens_map.json

@@ -0,0 +1 @@
+{"eos_token": "</s>", "unk_token": "<unk>", "pad_token": "<pad>", "additional_special_tokens": ["<extra_id_0>", "<extra_id_1>", "<extra_id_2>", "<extra_id_3>", "<extra_id_4>", "<extra_id_5>", "<extra_id_6>", "<extra_id_7>", "<extra_id_8>", "<extra_id_9>", "<extra_id_10>", "<extra_id_11>", "<extra_id_12>", "<extra_id_13>", "<extra_id_14>", "<extra_id_15>", "<extra_id_16>", "<extra_id_17>", "<extra_id_18>", "<extra_id_19>", "<extra_id_20>", "<extra_id_21>", "<extra_id_22>", "<extra_id_23>", "<extra_id_24>", "<extra_id_25>", "<extra_id_26>", "<extra_id_27>", "<extra_id_28>", "<extra_id_29>", "<extra_id_30>", "<extra_id_31>", "<extra_id_32>", "<extra_id_33>", "<extra_id_34>", "<extra_id_35>", "<extra_id_36>", "<extra_id_37>", "<extra_id_38>", "<extra_id_39>", "<extra_id_40>", "<extra_id_41>", "<extra_id_42>", "<extra_id_43>", "<extra_id_44>", "<extra_id_45>", "<extra_id_46>", "<extra_id_47>", "<extra_id_48>", "<extra_id_49>", "<extra_id_50>", "<extra_id_51>", "<extra_id_52>", "<extra_id_53>", "<extra_id_54>", "<extra_id_55>", "<extra_id_56>", "<extra_id_57>", "<extra_id_58>", "<extra_id_59>", "<extra_id_60>", "<extra_id_61>", "<extra_id_62>", "<extra_id_63>", "<extra_id_64>", "<extra_id_65>", "<extra_id_66>", "<extra_id_67>", "<extra_id_68>", "<extra_id_69>", "<extra_id_70>", "<extra_id_71>", "<extra_id_72>", "<extra_id_73>", "<extra_id_74>", "<extra_id_75>", "<extra_id_76>", "<extra_id_77>", "<extra_id_78>", "<extra_id_79>", "<extra_id_80>", "<extra_id_81>", "<extra_id_82>", "<extra_id_83>", "<extra_id_84>", "<extra_id_85>", "<extra_id_86>", "<extra_id_87>", "<extra_id_88>", "<extra_id_89>", "<extra_id_90>", "<extra_id_91>", "<extra_id_92>", "<extra_id_93>", "<extra_id_94>", "<extra_id_95>", "<extra_id_96>", "<extra_id_97>", "<extra_id_98>", "<extra_id_99>"]}

prott5/model/spiece.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:74da7b4afcde53faa570114b530c726135bdfcdb813dec3abfb27f9d44db7324
+size 237990

prott5/model/tokenizer_config.json

@@ -0,0 +1 @@
+{"do_lower_case": false}

requirements.txt

@@ -0,0 +1,172 @@
+absl-py @ file:///home/conda/feedstock_root/build_artifacts/absl-py_1733730548347/work
+accelerate==1.2.1
+aiohappyeyeballs @ file:///home/conda/feedstock_root/build_artifacts/aiohappyeyeballs_1733331917983/work
+aiohttp @ file:///home/conda/feedstock_root/build_artifacts/aiohttp_1734596887867/work
+aiosignal @ file:///home/conda/feedstock_root/build_artifacts/aiosignal_1734342155601/work
+alembic @ file:///home/conda/feedstock_root/build_artifacts/alembic_1733728406412/work
+AmberUtils==21.0
+astunparse @ file:///home/conda/feedstock_root/build_artifacts/astunparse_1728923142236/work
+async-timeout @ file:///home/conda/feedstock_root/build_artifacts/async-timeout_1733235340728/work
+attrs @ file:///home/conda/feedstock_root/build_artifacts/attrs_1734348785146/work
+biopython==1.79
+blinker @ file:///home/conda/feedstock_root/build_artifacts/blinker_1731096409132/work
+boto3==1.35.82
+botocore==1.35.82
+Brotli @ file:///home/conda/feedstock_root/build_artifacts/brotli-split_1725267488082/work
+cached-property @ file:///home/conda/feedstock_root/build_artifacts/cached_property_1615209429212/work
+cachetools @ file:///home/conda/feedstock_root/build_artifacts/cachetools_1733624221485/work
+catboost @ https://pypi.org/packages/cp39/c/catboost/catboost-1.2.7-cp39-cp39-manylinux2014_x86_64.whl#sha256=e58cf8966e33931acebffbc744cf640e8abd08d0fdfb0e503c107552cea6c643
+certifi @ file:///home/conda/feedstock_root/build_artifacts/certifi_1739515848642/work/certifi
+cffi @ file:///home/conda/feedstock_root/build_artifacts/cffi_1725571112467/work
+chardet @ file:///home/conda/feedstock_root/build_artifacts/chardet_1724954807150/work
+charset-normalizer @ file:///home/conda/feedstock_root/build_artifacts/charset-normalizer_1733218092148/work
+click @ file:///home/conda/feedstock_root/build_artifacts/click_1733221831880/work
+cloudpickle @ file:///home/conda/feedstock_root/build_artifacts/cloudpickle_1736947526808/work
+cmake==3.31.2
+colorama @ file:///home/conda/feedstock_root/build_artifacts/colorama_1733218098505/work
+colorlog @ file:///home/conda/feedstock_root/build_artifacts/colorlog_1733258404285/work
+contourpy @ file:///home/conda/feedstock_root/build_artifacts/contourpy_1727293517607/work
+cryptography @ file:///home/conda/feedstock_root/build_artifacts/cryptography-split_1732745941597/work
+cycler @ file:///home/conda/feedstock_root/build_artifacts/cycler_1733332471406/work
+edgembar==3.0
+filelock==3.16.1
+Flask @ file:///home/conda/feedstock_root/build_artifacts/flask_1741793020411/work
+flatbuffers @ file:///home/conda/feedstock_root/build_artifacts/python-flatbuffers_1733838640534/work
+fonttools @ file:///home/conda/feedstock_root/build_artifacts/fonttools_1733908950378/work
+freetype-py @ file:///home/conda/feedstock_root/build_artifacts/freetype-py_1650983368720/work
+frozenlist @ file:///home/conda/feedstock_root/build_artifacts/frozenlist_1729699456009/work
+fsspec==2024.10.0
+gast @ file:///home/conda/feedstock_root/build_artifacts/gast_1596839682936/work
+google-auth @ file:///home/conda/feedstock_root/build_artifacts/google-auth_1734176111475/work
+google-auth-oauthlib==1.0.0
+google-pasta @ file:///home/conda/feedstock_root/build_artifacts/google-pasta_1733852499742/work
+graphviz @ file:///home/conda/feedstock_root/build_artifacts/python-graphviz_1733791968395/work
+greenlet @ file:///home/conda/feedstock_root/build_artifacts/greenlet_1734532792566/work
+grpcio==1.68.1
+gym @ file:///home/conda/feedstock_root/build_artifacts/gym_1673317490724/work
+gym-notices @ file:///home/conda/feedstock_root/build_artifacts/gym-notices_1734709658541/work
+h2 @ file:///home/conda/feedstock_root/build_artifacts/h2_1733298745555/work
+h5py @ file:///home/conda/feedstock_root/build_artifacts/h5py_1734544937668/work
+hpack @ file:///home/conda/feedstock_root/build_artifacts/hpack_1733299205993/work
+huggingface-hub==0.26.5
+hyperframe @ file:///home/conda/feedstock_root/build_artifacts/hyperframe_1733298771451/work
+idna @ file:///home/conda/feedstock_root/build_artifacts/idna_1733211830134/work
+importlib_metadata @ file:///home/conda/feedstock_root/build_artifacts/importlib-metadata_1733223117029/work
+importlib_resources @ file:///home/conda/feedstock_root/build_artifacts/importlib_resources_1733231327252/work
+itsdangerous @ file:///home/conda/feedstock_root/build_artifacts/itsdangerous_1733308265247/work
+Jinja2==3.1.4
+jmespath==1.0.1
+joblib @ file:///home/conda/feedstock_root/build_artifacts/joblib_1733736026804/work
+keras @ file:///home/conda/feedstock_root/build_artifacts/keras_1669020828024/work/keras-2.11.0-py2.py3-none-any.whl
+Keras-Preprocessing @ file:///home/conda/feedstock_root/build_artifacts/keras-preprocessing_1610713559828/work
+kiwisolver @ file:///home/conda/feedstock_root/build_artifacts/kiwisolver_1725459266648/work
+lightgbm @ file:///home/conda/feedstock_root/build_artifacts/liblightgbm_1728547676427/work
+lit==18.1.8
+llvmlite==0.43.0
+lmdb==1.5.1
+Mako @ file:///home/conda/feedstock_root/build_artifacts/mako_1733628147227/work
+Markdown==3.7
+markdown-it-py @ file:///home/conda/feedstock_root/build_artifacts/markdown-it-py_1733250460757/work
+MarkupSafe @ file:///home/conda/feedstock_root/build_artifacts/markupsafe_1733219680183/work
+matplotlib==3.9.4
+mdurl @ file:///home/conda/feedstock_root/build_artifacts/mdurl_1733255585584/work
+ml-dtypes @ file:///home/conda/feedstock_root/build_artifacts/ml_dtypes_1725475048616/work
+MMPBSA.py==16.0
+mordred==1.2.0
+mpmath==1.3.0
+multidict @ file:///home/conda/feedstock_root/build_artifacts/multidict_1733913043842/work
+namex @ file:///home/conda/feedstock_root/build_artifacts/namex_1733858070416/work
+ndfes==3.0
+networkx @ file:///home/conda/feedstock_root/build_artifacts/networkx_1698504735452/work
+numba @ file:///home/conda/feedstock_root/build_artifacts/numba_1718888028049/work
+numpy @ file:///home/conda/feedstock_root/build_artifacts/numpy_1682210190296/work
+nvidia-cublas-cu11==11.10.3.66
+nvidia-cuda-cupti-cu11==11.7.101
+nvidia-cuda-nvrtc-cu11==11.7.99
+nvidia-cuda-runtime-cu11==11.7.99
+nvidia-cudnn-cu11==8.5.0.96
+nvidia-cufft-cu11==10.9.0.58
+nvidia-curand-cu11==10.2.10.91
+nvidia-cusolver-cu11==11.4.0.1
+nvidia-cusparse-cu11==11.7.4.91
+nvidia-nccl-cu11==2.14.3
+nvidia-nvtx-cu11==11.7.91
+oauthlib @ file:///home/conda/feedstock_root/build_artifacts/oauthlib_1733752848439/work
+opt_einsum @ file:///home/conda/feedstock_root/build_artifacts/opt_einsum_1733687912731/work
+optree @ file:///home/conda/feedstock_root/build_artifacts/optree_1731510511426/work
+optuna @ file:///home/conda/feedstock_root/build_artifacts/optuna_1734927470522/work
+packaging @ file:///home/conda/feedstock_root/build_artifacts/packaging_1733203243479/work
+packmol_memgen==2025.1.29
+pandas @ file:///home/conda/feedstock_root/build_artifacts/pandas_1726878406085/work
+patsy @ file:///home/conda/feedstock_root/build_artifacts/patsy_1733792384640/work
+pdb4amber==22.0
+peft==0.4.0
+peptides @ file:///opt/conda/conda-bld/peptides_1726747256415/work
+pfeature==1.4
+Pillow @ file:///home/conda/feedstock_root/build_artifacts/pillow_1688255839723/work
+plotly @ file:///home/conda/feedstock_root/build_artifacts/plotly_1733733072265/work
+propcache @ file:///home/conda/feedstock_root/build_artifacts/propcache_1733391807885/work
+protobuf==5.29.1
+psutil==6.1.0
+pyasn1 @ file:///home/conda/feedstock_root/build_artifacts/pyasn1_1733217608156/work
+pyasn1_modules @ file:///home/conda/feedstock_root/build_artifacts/pyasn1-modules_1733324602540/work
+pycairo==1.25.0
+pycparser @ file:///home/conda/feedstock_root/build_artifacts/bld/rattler-build_pycparser_1733195786/work
+Pygments @ file:///home/conda/feedstock_root/build_artifacts/pygments_1733221634316/work
+PyJWT @ file:///home/conda/feedstock_root/build_artifacts/pyjwt_1732782409051/work
+pyMSMT==22.0
+pynndescent @ file:///home/conda/feedstock_root/build_artifacts/pynndescent_1734193530377/work
+pyOpenSSL @ file:///home/conda/feedstock_root/build_artifacts/pyopenssl_1733777368795/work
+pyparsing @ file:///home/conda/feedstock_root/build_artifacts/pyparsing_1733222594562/work
+PySocks @ file:///home/conda/feedstock_root/build_artifacts/pysocks_1733217236728/work
+python-dateutil @ file:///home/conda/feedstock_root/build_artifacts/python-dateutil_1733215673016/work
+pytraj==2.0.6
+pytz @ file:///home/conda/feedstock_root/build_artifacts/pytz_1706886791323/work
+pyu2f @ file:///home/conda/feedstock_root/build_artifacts/pyu2f_1733738580568/work
+PyYAML @ file:///home/conda/feedstock_root/build_artifacts/pyyaml_1725456176299/work
+regex==2024.11.6
+reportlab @ file:///home/conda/feedstock_root/build_artifacts/reportlab_1727846838921/work
+requests @ file:///home/conda/feedstock_root/build_artifacts/requests_1733217035951/work
+requests-oauthlib @ file:///home/conda/feedstock_root/build_artifacts/requests-oauthlib_1733772243268/work
+rich @ file:///home/conda/feedstock_root/build_artifacts/rich_1733342254348/work/dist
+rlPyCairo @ file:///home/conda/feedstock_root/build_artifacts/rlpycairo_1687519531733/work
+rsa @ file:///home/conda/feedstock_root/build_artifacts/rsa_1733662684165/work
+s3transfer==0.10.4
+safetensors==0.4.5
+sander==22.0
+scikit-learn @ file:///home/conda/feedstock_root/build_artifacts/scikit-learn_1733760110140/work/dist/scikit_learn-1.6.0-cp39-cp39-linux_x86_64.whl#sha256=f4205b3762b8033782da7a40727c959075158f152dcb3f50e33f60ecfc8cf6f4
+scipy @ file:///home/conda/feedstock_root/build_artifacts/scipy-split_1716470218293/work/dist/scipy-1.13.1-cp39-cp39-linux_x86_64.whl#sha256=e6696cb8683d94467891b7648e068a3970f6bc0a1b3c1aa7f9bc89458eafd2f0
+seaborn @ file:///home/conda/feedstock_root/build_artifacts/seaborn-split_1733730015268/work
+sentencepiece @ file:///home/conda/feedstock_root/build_artifacts/sentencepiece-split_1674200849009/work/python
+shap @ file:///home/conda/feedstock_root/build_artifacts/shap_1732083651745/work
+six @ file:///home/conda/feedstock_root/build_artifacts/six_1733380938961/work
+slicer @ file:///home/conda/feedstock_root/build_artifacts/slicer_1710029110974/work
+SQLAlchemy @ file:///home/conda/feedstock_root/build_artifacts/sqlalchemy_1729066347171/work
+statsmodels @ file:///home/conda/feedstock_root/build_artifacts/statsmodels_1727986706423/work
+sympy==1.13.3
+tape-proteins==0.5
+tenacity @ file:///home/conda/feedstock_root/build_artifacts/tenacity_1733649050774/work
+tensorboard==2.13.0
+tensorboard-data-server==0.7.2
+tensorboard-plugin-wit @ file:///home/conda/feedstock_root/build_artifacts/tensorboard-plugin-wit_1641458951060/work/tensorboard_plugin_wit-1.8.1-py3-none-any.whl
+tensorboardX==2.6.2.2
+tensorflow @ file:///home/conda/feedstock_root/build_artifacts/tensorflow-split_1679857771045/work/tensorflow_pkg/tensorflow-2.11.1-cp39-cp39-linux_x86_64.whl
+tensorflow-estimator @ file:///home/conda/feedstock_root/build_artifacts/tensorflow-split_1679857771045/work/tensorflow-estimator/wheel_dir/tensorflow_estimator-2.11.0-py2.py3-none-any.whl
+termcolor @ file:///home/conda/feedstock_root/build_artifacts/termcolor_1733754631889/work
+threadpoolctl @ file:///home/conda/feedstock_root/build_artifacts/threadpoolctl_1714400101435/work
+tokenizers==0.13.3
+torch==2.0.1
+tqdm==4.65.0
+transformers==4.30.0
+triton==2.0.0
+typing_extensions @ file:///home/conda/feedstock_root/build_artifacts/typing_extensions_1733188668063/work
+tzdata @ file:///home/conda/feedstock_root/build_artifacts/python-tzdata_1733235305708/work
+umap-learn @ file:///home/conda/feedstock_root/build_artifacts/umap-learn_1730221913988/work
+unicodedata2 @ file:///home/conda/feedstock_root/build_artifacts/unicodedata2_1729704563364/work
+urllib3==1.26.20
+Werkzeug @ file:///home/conda/feedstock_root/build_artifacts/werkzeug_1733160440960/work
+wrapt @ file:///home/conda/feedstock_root/build_artifacts/wrapt_1732523603052/work
+xgboost @ file:///home/conda/feedstock_root/build_artifacts/xgboost-split_1733179637554/work/python-package
+yarl @ file:///home/conda/feedstock_root/build_artifacts/yarl_1733428811664/work
+zipp @ file:///home/conda/feedstock_root/build_artifacts/zipp_1732827521216/work
+zstandard==0.23.0