Skip to content
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

[Perf] Reduce peak memory usage of llama #10339

Merged
merged 1 commit into from
Nov 15, 2024
Merged

[Perf] Reduce peak memory usage of llama #10339

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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 vllm/model_executor/models/llama.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -90,8 +90,8 @@ def __init__(
self.act_fn = SiluAndMul()

def forward(self, x):
gate_up, _ = self.gate_up_proj(x)
Copy link
Member

Choose a reason for hiding this comment

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

I think torch.compile can do something similar, without renaming variables.

to keep the original semantic, maybe adding del x would be more intuitive.

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

I think torch.compile can do something similar, without renaming variables.

Yes, it can completely alleviate this problem, even when we consider cross-function refcounting which I'll cover in my investigation write-up.

to keep the original semantic, maybe adding del x would be more intuitive.

I think you might mean in this case del gate_up? Yes indeed we can add dels and make the variable names more descriptive. I just kept it as x to avoid adding extra dels and be similar to style of the rest of the function.

x = self.act_fn(gate_up)
x, _ = self.gate_up_proj(x)
x = self.act_fn(x)
x, _ = self.down_proj(x)
return x

Expand Down