Multiple GPUs are being used despite `--num_processes 1`

🤗Accelerate
1

I’m launching my script via accelerate launch --num_processes 1 train.py. However, the following code is printing the following output:

# Code
logger.warning(
        "Process rank: %s, device: %s, n_gpu: %s, distributed training: %s, 16-bits training: %s",
        training_args.local_rank,
        training_args.device,
        training_args.n_gpu,
        bool(training_args.local_rank != -1),
        training_args.fp16,
)

# Output
Process rank: 0, device: cuda:0, n_gpu: 4, distributed training: True, 16-bits training: True

Why is n_gpu 4, and why is the value of local_rank not being set to the default value of -1?

The code runs fine when I set CUDA_VISIBLE_DEVICES=0, but I would think that HF Accelerate would handle this without having to set environment variables.