Skip to content

fix(quantization): handle mixed-precision fused layers in is_layer_skipped_gguf #107

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
BenWongCityuCS wants to merge 1 commit into
base: main
Choose a base branch
from
+6 −5
Open

fix(quantization): handle mixed-precision fused layers in is_layer_skipped_gguf #107

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
11 changes: 6 additions & 5 deletions vllm_gguf_plugin/quantization/utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,11 +31,12 @@ def is_layer_skipped_gguf(
if is_skipped is None:
is_skipped = is_shard_skipped
elif is_shard_skipped != is_skipped:
raise ValueError(
f"Detected some but not all shards of {prefix} "
"are quantized. All shards of fused layers "
"to have the same precision."
)
# Unsloth UD / "XL" GGUFs store fused projections at mixed
# precision by design. vLLM 0.27's Kimi-K3 KDA layer also
# zeroes replicated-shard padding via `.weight`, so fused
# layers with heterogeneous shard types are handled as
# UNQUANTIZED instead of raising.
is_skipped = proj_name == "in_proj_qkvgfab"
else:
is_skipped = any(module_name in prefix for module_name in unquantized_modules)

Expand Down