出现运行时错误：使用 mps 时视图大小与输入张量的大小和步幅不兼容

Question

运行时错误：视图大小与输入张量的大小和步幅不兼容（至少一维跨越两个连续的子空间）。使用 .reshape(...) 代替。

仅在使用 mps 而不是 cpu 时出现此错误

fine_tuned_decoder_path = "/path/fine_tuned_decoder"
model = VisionEncoderDecoderModel.from_encoder_decoder_pretrained(
    encoder_pretrained_model_name_or_path="google/vit-base-patch16-224-in21k",
    decoder_pretrained_model_name_or_path=fine_tuned_decoder_path,
    tie_encoder_decoder=True,
    cache_dir="/path/datasets/"+"models"  # Directory for caching models
)
os.environ["WANDB_MODE"] = "disabled"

# Set batch size and number of training epochs
BATCH_SIZE = 16
TRAIN_EPOCHS = 5

# Define the output directory for storing training outputs
output_directory = os.path.join("path", "captioning_outputs")

# Check if MPS is available
device = torch.device("mps" if torch.backends.mps.is_available() else "cuda" if torch.cuda.is_available() else "cpu")

# Move your model to the correct device
model.to(device)

# Set mixed precision and device handling
fp16 = False  # Disable fp16 entirely
mixed_precision = None  # Disable mixed precision (default)

# Training Arguments
training_args = TrainingArguments(
    output_dir=output_directory,
    per_device_train_batch_size=BATCH_SIZE,
    do_train=True,
    num_train_epochs=TRAIN_EPOCHS,
    overwrite_output_dir=True,
    use_cpu=False,  # Ensure you're not using CPU
    dataloader_pin_memory=False,
    fp16=fp16,  # Disable fp16 if using MPS
    bf16=False,  # Disable bf16 if using MPS
    optim="adamw_torch",  # Use AdamW Torch optimizer (more stable with mixed precision)
    gradient_checkpointing=False,  # Disable gradient checkpointing if necessary
    logging_dir=os.path.join(output_directory, 'logs'),
    report_to="none",  # Disable reporting
    
)

# Use the Trainer with the model on the correct device
trainer = Trainer(
    processing_class=feature_extractor,  # Tokenizer
    model=model,  # Model to train
    args=training_args,  # Training arguments
    train_dataset=train_dataset,  # Training dataset
    data_collator=default_data_collator  # Data collator
)
# Start the training process
trainer.train()

尝试设置 use_cpu=True，它工作正常，但不适用于 mps

Answer 1

PyTorch 2.5 存在一个已知问题。

您可以在这里阅读更多相关信息：

https://github.com/pytorch/pytorch/issues/142344

一种解决方案是下载源代码，按照malfet描述修改这两个文件，然后编译源代码。这个问题预计会在 PyTorch 2.6 中得到解决。

但是，我发现的最简单的解决方案是安装 PyTorch 的夜间版本：

pip install --pre torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cpu

致以诚挚的问候，

--KC