Skip to content

Fix GPTQ NF4&FP4 quant #2314

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
wants to merge 2 commits into
base: master
Choose a base branch
from
Open

Fix GPTQ NF4&FP4 quant #2314

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
2af8493
fix gptq NF4/FP4
Kaihui-intel Oct 15, 2025
5fe07d9
update int_weight
Kaihui-intel Oct 16, 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
2 changes: 2 additions & 0 deletions neural_compressor/torch/algorithms/weight_only/utility.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -516,6 +516,7 @@ def quant_weight_w_scale(weight, scale, scale_bf16_to_fp8, zp=None, group_size=-
if dtype in FLOAT_MAPPING.keys(): # NF4 FP4
int_weight_tmp = weight[:, i * group_size : (i + 1) * group_size]
quantize_4bit(int_weight_tmp, scale=scale[:, i].unsqueeze(1), dtype=dtype, return_int=True)[0]
int_weight[:, i * group_size : (i + 1) * group_size].copy_(int_weight_tmp)
else:
int_weight_tmp = weight[:, i * group_size : (i + 1) * group_size].div_(scale[:, i].unsqueeze(1))
if zp is not None:
Expand All @@ -526,6 +527,7 @@ def quant_weight_w_scale(weight, scale, scale_bf16_to_fp8, zp=None, group_size=-
if dtype in FLOAT_MAPPING.keys(): # NF4 FP4
int_weight_tmp = weight[:, leng * group_size :]
quantize_4bit(int_weight_tmp, scale=scale[:, -1].unsqueeze(1), dtype=dtype, return_int=True)[0]
int_weight[:, leng * group_size :].copy_(int_weight_tmp)
else:
int_weight_tmp = weight[:, leng * group_size :].div_(scale[:, -1].unsqueeze(1))
if zp is not None:
Expand Down