Skip to content

[Bugfix] fix for deepseek w4a16 #8906

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
merged 2 commits into from
Sep 27, 2024
Merged

[Bugfix] fix for deepseek w4a16 #8906

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3f2a262
fix for deepseek w4a16
LucasWilkinson Sep 27, 2024
5640d27
Format
mgoin Sep 27, 2024
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
9 changes: 5 additions & 4 deletions vllm/model_executor/layers/quantization/kernels/marlin.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,10 +38,11 @@ def can_implement(cls,
"Marlin, supported group sizes are: "\
f"{MARLIN_SUPPORTED_GROUP_SIZES}"

return check_marlin_supports_shape(c.partition_weight_shape[0],
c.partition_weight_shape[1],
c.full_weight_shape[1],
c.group_size)
return check_marlin_supports_shape(
c.partition_weight_shape[1], # out_features
c.partition_weight_shape[0], # in_features
c.full_weight_shape[0], # in_features
c.group_size)

# note assumes that
# `weight_packed` is: {input_dim = 0, output_dim = 1, packed_dim = 0}
Expand Down