Missing positional arguments when try to use multiple GPUs with accelerator

🤗Accelerate
1

First time using accelerator and notebook_launcher.
I have 2 GPUs in the desktop.

When I do this:
notebook_launcher(training_process, (train_loader, val_loader), num_processes=1) → working great on my CPU.

Then I try this:
notebook_launcher(training_process, (train_loader, val_loader), num_processes=2)
Got error message: TypeError: training_process() missing 2 required positional arguments: ‘train_loader’ and ‘val_loader’

Did I miss something? Plz point me a direction to resolve this… Thanks.

2

That’s because those args are not passed in this case :man_facepalming:
Will send a PR today to fix this!

3

Thanks for the explanation.

4

Here is the PR with the fix!

5

Thanks for the quick fix!