Can't use Trainer on mps device

[smartliuhw]

System Info

version information:
transformers -- 4.48.1
torch -- 2.2.2
accelerate -- 1.3.0
system -- mac os 15.2 (24C101)

Who can help?

@muellerzr @SunMarc

Information

• The official example scripts
• My own modified scripts

Tasks

• An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
• My own task or dataset (give details below)

Reproduction

I'm using c3 dataset to finetune Qwen2.5-3B-Instruct for classification task. The code runs successfully on a cuda device, while I try to run the same code on the mps device, got error messages below

The following are the most important part of my code:

## Train the model

exp_num=0

training_args = TrainingArguments(
    output_dir=f"./output/{lora_rank}_{exp_num}",
    learning_rate=5e-5,
    per_device_train_batch_size=4,
    per_device_eval_batch_size=2,
    num_train_epochs=5,
    weight_decay=0.01,
    evaluation_strategy="epoch",
    save_strategy="epoch",
    load_best_model_at_end=True,
    fp16=True,
    bf16=False
)

trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=train_dataset,
    eval_dataset=dev_dataset,
    tokenizer=tokenizer,
    data_collator=datacollator,
    compute_metrics=compute_metrics
)
logger.info("Start training")
trainer.train()

Expected behavior

The code runs sucessfully

[SunMarc]

This is indeed not supported yet. Follow the this PR which should solve your issue. huggingface/accelerate#3373

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.