Skip to content

change torch.cuda.amp.GradScaler to torch.GradScaler("cuda") #3257

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
malfet merged 4 commits into from
Jan 30, 2025
Merged

change torch.cuda.amp.GradScaler to torch.GradScaler("cuda") #3257

Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
85c31ba
change torch.cuda.amp.GradScaler to torch.GradScaler("cuda")
0ssamaak0 Jan 24, 2025
0c71332
changing torch.GradScaler to torch.amp.GradScaler
0ssamaak0 Jan 26, 2025
627983b
Merge branch 'main' into patch-1
0ssamaak0 Jan 28, 2025
50173b2
Merge branch 'main' into patch-1
svekars Jan 28, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions recipes_source/recipes/amp_recipe.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -150,7 +150,7 @@ def make_model(in_size, out_size, num_layers):
# The same ``GradScaler`` instance should be used for the entire convergence run.
# If you perform multiple convergence runs in the same script, each run should use
# a dedicated fresh ``GradScaler`` instance. ``GradScaler`` instances are lightweight.
scaler = torch.cuda.amp.GradScaler()
scaler = torch.GradScaler("cuda")

for epoch in range(0): # 0 epochs, this section is for illustration only
for input, target in zip(data, targets):
Expand Down Expand Up @@ -182,7 +182,7 @@ def make_model(in_size, out_size, num_layers):

net = make_model(in_size, out_size, num_layers)
opt = torch.optim.SGD(net.parameters(), lr=0.001)
scaler = torch.cuda.amp.GradScaler(enabled=use_amp)
scaler = torch.GradScaler("cuda" ,enabled=use_amp)
Copy link
Contributor

@spzala spzala Jan 24, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Same as above.

Copy link
Contributor Author

@0ssamaak0 0ssamaak0 Jan 25, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I followed this
image

Source

Copy link
Contributor

@spzala spzala Jan 26, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for the quick reply. GradScaler can be used on its own but because you are modifying amp recipe, I would suggest to use torch.amp.GradScaler as it's also mentioned in the code - https://github.com/pytorch/pytorch/blob/main/torch/cuda/amp/grad_scaler.py#L12
Also, if you run the recipe, you will notice the warning message suggesting to use torch.amp.GradScaler

Copy link
Contributor Author

@0ssamaak0 0ssamaak0 Jan 26, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I have changed the code accordingly.

Thank you

Copy link
Contributor

@spzala spzala Jan 28, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks!


start_timer()
for epoch in range(epochs):
Expand Down