UniMol2_embedding.py

import os
import pickle
import numpy as np
import pandas as pd
import torch
import torch.nn as nn
from tqdm import tqdm
from unimol_tools import UniMolRepr
import os, sys, contextlib

@contextlib.contextmanager
def suppress_stdout_stderr():
    """with 块内所有 print / tqdm / warning 都不会显示"""
    with open(os.devnull, 'w') as devnull:
        old_out, old_err = sys.stdout, sys.stderr
        sys.stdout, sys.stderr = devnull, devnull
        try:
            yield
        finally:
            sys.stdout, sys.stderr = old_out, old_err

def get_unimol_embeddings(smiles_list, output_file="UniMol_emb512.pkl", 
                         model_name='unimolv1', model_size='84m', 
                         remove_hs=False, batch_size=32):
    """
    使用Uni-Mol模型为SMILES列表生成分子嵌入，并保存为pickle文件
    
    参数:
    smiles_list (list): SMILES字符串列表
    output_file (str): 输出pickle文件路径
    model_name (str): 模型名称，可选'unimolv1'或'unimolv2'
    model_size (str): 模型大小，仅在使用unimolv2时有效
    remove_hs (bool): 是否移除氢原子
    batch_size (int): 批处理大小
    
    返回:
    dict: 包含SMILES及其对应嵌入的字典
    """
    # 初始化模型
    clf = UniMolRepr(
        data_type='molecule',
        remove_hs=remove_hs,
        model_name=model_name,
        model_size=model_size
    )
    
    # 用于存储结果的字典
    embeddings_dict = {}
    error_smiles = []
    
    # 批处理SMILES
    total_batches = (len(smiles_list) + batch_size - 1) // batch_size
    
    print(f"开始生成{len(smiles_list)}个SMILES的嵌入表示...")
    
    print("开始")
    # with suppress_stdout_stderr():
    for i in tqdm(range(total_batches), desc="处理批次"):
        batch = smiles_list[i*batch_size : (i+1)*batch_size]
        
        try:
            # 获取嵌入表示
            batch_repr = clf.get_repr(batch, return_atomic_reprs=False)
            
            # 将结果存入字典 (使用CLS token作为分子表示)
            for idx, smiles in enumerate(batch):
                embeddings_dict[smiles] = batch_repr['cls_repr'][idx]
                
        except Exception as e:
            print(f"处理批次 {i+1}/{total_batches} 时发生错误: {str(e)}")
            # 记录处理失败的SMILES
            error_smiles.extend(batch)
        
    # 保存嵌入结果
    with open(output_file, 'wb') as f:
        pickle.dump(embeddings_dict, f)
    
    print(f"嵌入生成完成！共处理 {len(smiles_list)} 个SMILES，"
          f"{len(smiles_list) - len(error_smiles)} 个成功，"
          f"{len(error_smiles)} 个失败。")
    print(f"嵌入结果已保存至 {output_file}")
    
    if error_smiles:
        print(f"处理失败的SMILES已记录。")
    
    return embeddings_dict

# 使用示例
if __name__ == "__main__":
    
    # 假设unique_smiles是你的SMILES列表
    unique_smiles = [
    "CC(=O)OC1=CC=CC=C1C(=O)O",  # Aspirin
    "CN1C=NC2=C1C(=O)N(C(=O)N2C)C" # Caffeine
    ]  # 示例SMILES列表
    unique_smiles = pd.read_csv('./LINCS2020/LINCS2020_smiles.csv')['SMILES'].tolist()

    # UniMol V1
    # embeddings = get_unimol_embeddings(
    #     smiles_list=unique_smiles,
    #     output_file="embeddings/UniMol_emb512.pkl",
    #     model_name='unimolv1',
    #     model_size='84m',
    #     remove_hs=False,
    #     batch_size=32
    # )
    
    # UniMol V2
    embeddings = get_unimol_embeddings(
    smiles_list=unique_smiles,
    output_file="embeddings/UniMolV2_emb1024.pkl", # save path
    model_name='unimolv2',      # 修改为 v2 版本
    model_size='310m',          # 指定 v2 模型大小（根据需要选择）
    remove_hs=False,
    batch_size=32
    )

    # 打印样例嵌入
    sample_smiles = unique_smiles[0]
    print(f"SMILES: {sample_smiles}")
    print(f"嵌入向量: {embeddings[sample_smiles]}")
    print(f"嵌入维度: {len(embeddings[sample_smiles])}")