Huggingface 的“resume_from_checkpoint”有效吗？

我目前将我的教练设置为：

training_args = TrainingArguments(
    output_dir=f"./results_{model_checkpoint}",
    evaluation_strategy="epoch",
    learning_rate=5e-5,
    per_device_train_batch_size=4,
    per_device_eval_batch_size=4,
    num_train_epochs=2,
    weight_decay=0.01,
    push_to_hub=True,
    save_total_limit = 1,
    resume_from_checkpoint=True,
)

trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=tokenized_qa["train"],
    eval_dataset=tokenized_qa["validation"],
    tokenizer=tokenizer,
    data_collator=DataCollatorForMultipleChoice(tokenizer=tokenizer),
    compute_metrics=compute_metrics
)

训练结束后，在我的output_dir我有培训师保存的几个文件：

['README.md',
 'tokenizer.json',
 'training_args.bin',
 '.git',
 '.gitignore',
 'vocab.txt',
 'config.json',
 'checkpoint-5000',
 'pytorch_model.bin',
 'tokenizer_config.json',
 'special_tokens_map.json',
 '.gitattributes']

来自文档 https://huggingface.co/docs/transformers/main_classes/trainer#transformers.Trainer.train.resume_from_checkpoint看起来resume_from_checkpoint将从最后一个检查点继续训练模型：

resume_from_checkpoint (str or bool, optional) — If a str, local path to a saved checkpoint as saved by a previous instance of Trainer. If a bool and equals True, load the last checkpoint in args.output_dir as saved by a previous instance of Trainer. If present, training will resume from the model/optimizer/scheduler states loaded here.

但当我打电话时trainer.train()它似乎删除了最后一个检查点并开始一个新的检查点：

Saving model checkpoint to ./results_distilbert-base-uncased/checkpoint-500
...
Deleting older checkpoint [results_distilbert-base-uncased/checkpoint-5000] due to args.save_total_limit

它是否真的从最后一个检查点（即 5000）继续训练，并从 0 开始新检查点的计数（保存 500 步后的第一个检查点 - “checkpoint-500”），或者它只是不继续训练？我还没有找到测试它的方法，文档对此也不清楚。

是的，它有效！当您调用 trainer.train() 时，您隐式地告诉它覆盖所有检查点并从头开始。您应该调用 trainer.train(resume_from_checkpoint=True) 或将resume_from_checkpoint 设置为指向检查点路径的字符串。