single_dcu_train.py

import torch
from torch.utils.data import Dataset, DataLoader
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
from transformers import AdamW, get_scheduler
from tqdm.auto import tqdm
from rouge import Rouge
import random
import numpy as np
import os
import json
from torch import nn
from datetime import datetime

def seed_everything(seed=1029):
    random.seed(seed)
    os.environ['PYTHONHASHSEED'] = str(seed)
    np.random.seed(seed)
    torch.manual_seed(seed)
    torch.cuda.manual_seed(seed)
    torch.cuda.manual_seed_all(seed)
    torch.backends.cudnn.deterministic = True

class LCSTS(Dataset):
    def __init__(self, data_file):
        self.data = self.load_data(data_file)
    
    def load_data(self, data_file):
        Data = {}
        with open(data_file, 'rt', encoding='utf-8') as f:
            for idx, line in enumerate(f):
                if idx >= max_dataset_size:
                    break
                items = line.strip().split('!=!')
                assert len(items) == 2
                Data[idx] = {
                    'title': items[0],
                    'content': items[1]
                }
        return Data
    
    def __len__(self):
        return len(self.data)

    def __getitem__(self, idx):
        return self.data[idx]



def collate_fn(batch_samples):
    batch_inputs, batch_targets = [], []
    for sample in batch_samples:
        batch_inputs.append(sample['content'])
        batch_targets.append(sample['title'])
    batch_data = tokenizer(
        batch_inputs, 
        padding=True, 
        max_length=max_input_length,
        truncation=True, 
        return_tensors="pt"
    )
    with tokenizer.as_target_tokenizer():
        labels = tokenizer(
            batch_targets, 
            padding=True, 
            max_length=max_target_length,
            truncation=True, 
            return_tensors="pt"
        )["input_ids"]
        batch_data['decoder_input_ids'] = model.prepare_decoder_input_ids_from_labels(labels)
        end_token_index = torch.where(labels == tokenizer.eos_token_id)[1]
        for idx, end_idx in enumerate(end_token_index):
            labels[idx][end_idx+1:] = -100
        batch_data['labels'] = labels
    return batch_data


def train_loop(dataloader, model, optimizer, lr_scheduler, epoch, total_loss):
    progress_bar = tqdm(range(len(dataloader)))
    progress_bar.set_description(f'loss: {0:>7f}')
    finish_batch_num = (epoch-1) * len(dataloader)
    
    model.train()
    for batch, batch_data in enumerate(dataloader, start=1):
        batch_data = {k: v.to(device) for k, v in batch_data.items()}
        outputs = model(**batch_data)
        loss = outputs.loss
        loss = loss.mean()

        optimizer.zero_grad()
        loss.backward()
        optimizer.step()
        lr_scheduler.step()

        total_loss += loss.item()
        progress_bar.set_description(f'loss: {total_loss/(finish_batch_num + batch):>7f}')
        progress_bar.update(1)
    return total_loss

def test_loop(dataloader, model):
    preds, labels = [], []
    
    model.eval()
    for batch_data in tqdm(dataloader):
        batch_data = {k: v.to(device) for k, v in batch_data.items()}
        with torch.no_grad():
            # 如果你使用了 DataParallel，你可以通过访问 model.module 来获取原始模型
            generated_tokens = model.generate(
                batch_data["input_ids"],
                attention_mask=batch_data["attention_mask"],
                max_length=max_target_length,
                num_beams=beam_size,
                no_repeat_ngram_size=no_repeat_ngram_size,
            ).cpu().numpy()
        if isinstance(generated_tokens, tuple):
            generated_tokens = generated_tokens[0]
        label_tokens = batch_data["labels"].cpu().numpy()

        decoded_preds = tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)
        label_tokens = np.where(label_tokens != -100, label_tokens, tokenizer.pad_token_id)
        decoded_labels = tokenizer.batch_decode(label_tokens, skip_special_tokens=True)

        preds += [' '.join(pred.strip()) for pred in decoded_preds]
        labels += [' '.join(label.strip()) for label in decoded_labels]
    scores = rouge.get_scores(hyps=preds, refs=labels, avg=True)
    result = {key: value['f'] * 100 for key, value in scores.items()}
    result['avg'] = np.mean(list(result.values()))
    print(f"Rouge1: {result['rouge-1']:>0.2f} Rouge2: {result['rouge-2']:>0.2f} RougeL: {result['rouge-l']:>0.2f}\n")
    return result


if __name__=='__main__':

    os.environ["HIP_VISIBLE_DEVICES"] = "0"

    device = 'cuda' if torch.cuda.is_available() else 'cpu'
    print(f'Using {device} device')
    seed_everything(5)

    rouge = Rouge()

    max_dataset_size = 200000
    max_input_length = 512
    max_target_length = 32

    batch_size = 16
    learning_rate = 1e-5
    epoch_num = 1

    beam_size = 4
    no_repeat_ngram_size = 2

    folder_path = "/saves/train_dtk_weights"

    # 检查文件夹是否存在
    if not os.path.exists(folder_path):
        # 如果不存在，则创建文件夹
        os.makedirs(folder_path)

    train_data = LCSTS('/umt5/data/lcsts_tsv/data1.tsv')
    valid_data = LCSTS('/umt5/data/lcsts_tsv/data2.tsv')

    model_checkpoint = "/umt5/umt5_base"
    tokenizer = AutoTokenizer.from_pretrained(model_checkpoint)
    model = AutoModelForSeq2SeqLM.from_pretrained(model_checkpoint)

        # 检查是否有多个 GPU 可用
    if torch.cuda.device_count() > 1:
        print("Let's use", torch.cuda.device_count(), "GPUs!")
        # 如果有多个 GPUs，使用 nn.DataParallel 包装模型
        model = nn.DataParallel(model).to(device)
    else:
        model = model.to(device)
         
    
    train_dataloader = DataLoader(train_data, batch_size=batch_size, shuffle=True, collate_fn=collate_fn)
    valid_dataloader = DataLoader(valid_data, batch_size=batch_size, shuffle=False, collate_fn=collate_fn)


    optimizer = AdamW(model.parameters(), lr=learning_rate)
    lr_scheduler = get_scheduler(
        "linear",
        optimizer=optimizer,
        num_warmup_steps=0,
        num_training_steps=epoch_num*len(train_dataloader),
    )

    total_loss = 0.
    best_avg_rouge = 0.
    for t in range(epoch_num):
        print(f"Epoch {t+1}/{epoch_num}\n-------------------------------")
        total_loss = train_loop(train_dataloader, model, optimizer, lr_scheduler, t+1, total_loss)
        valid_rouge = test_loop(valid_dataloader, model)
        rouge_avg = valid_rouge['avg']
        if rouge_avg > best_avg_rouge:
            best_avg_rouge = rouge_avg
            print('saving new weights...\n')
            weight_path = f'/utm5/saves/train_dtk_weights/epoch_{t+1}_valid_rouge_{rouge_avg:0.4f}_model_dtk_weights.bin'
            torch.save(model.state_dict(), weight_path)
            # 加载训练后的权重
            state_dict = torch.load(weight_path)
            model.load_state_dict(state_dict)
            # 获取当前的日期和时间
            now = datetime.now()
            timestamp = now.strftime("%Y%m%d_%H%M%S")
            new_model_path = f'saves/umt5_{timestamp}'
            model.module.save_pretrained(new_model_path)
            tokenizer.save_pretrained(new_model_path)
    print("Done!")