Skip to content

[NVBUG: 5701937]Clear GPU cache for 3D weight tensors #649

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

Open
cjluo-nv wants to merge 1 commit into
base: main
Choose a base branch
from
Open

[NVBUG: 5701937]Clear GPU cache for 3D weight tensors #649

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
8 changes: 5 additions & 3 deletions modelopt/torch/export/quant_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -756,6 +756,11 @@ def to_quantized_weight(
if isinstance(weight, QTensorWrapper):
return weight.data

if weight.dim() == 3:
# for MOE stacked weights
# Clear GPU cache to avoid pontential GPU OOM issues for large models.
clear_cuda_cache()

if quantization == QUANTIZATION_FP8:
# Fix RuntimeError: Promotion for Float8 Types is not supported, attempted to promote Float8_e4m3fn and Float
# in speculative decoding fp8 model export
Expand All @@ -764,9 +769,6 @@ def to_quantized_weight(
return weight

if weight.dim() == 3:
# for MOE stacked weights
# Clear GPU cache to avoid pontential GPU OOM issues for large models.
clear_cuda_cache()
return (weight / weights_scaling_factor.unsqueeze(-1)).to(torch.float8_e4m3fn)
return (weight / weights_scaling_factor).to(torch.float8_e4m3fn)

Expand Down