NVIDIA GPU

TrainingArguments, Trainer, and Pipeline detect NVIDIA GPUs and set the backend to cuda automatically.

Select and order GPUs

You can control which GPUs PyTorch sees and in what order. Prioritize faster GPUs or limit a run to a subset of available GPUs. It works with both DistributedDataParallel and DataParallel, and doesn’t require Accelerate or the DeepSpeed integration.

Use the CUDA_VISIBLE_DEVICES environment variable to select GPUs and set their order. Set it on the command line per run, or add it to ~/.bashrc or another startup config file.

Avoid exporting environment variables because if you forget how an environment variable was set up, you may silently train on the wrong GPUs. Set the environment variable on the same command line as the training run.

For example, to select GPUs 0 and 2 out of four:

CUDA_VISIBLE_DEVICES=0,2 torchrun trainer-program.py ...

PyTorch sees only GPUs 0 and 2, which are mapped to cuda:0 and cuda:1. To reverse the order (use GPU 2 as cuda:0 and GPU 0 as cuda:1):

CUDA_VISIBLE_DEVICES=2,0 torchrun trainer-program.py ...

To run without any GPUs:

CUDA_VISIBLE_DEVICES= python trainer-program.py ...

Control the order of CUDA devices with CUDA_DEVICE_ORDER.

• Order by PCIe bus ID (matches nvidia-smi):

  Copied

  export CUDA_DEVICE_ORDER=PCI_BUS_ID

• Order by compute capability (fastest first):

  Copied

  export CUDA_DEVICE_ORDER=FASTEST_FIRST