Skip to content

Commit bda5305

Browse files
authored
Add scale and zp dtype in preq
Differential Revision: D73292389 Pull Request resolved: #2080
1 parent 34421b1 commit bda5305

File tree

1 file changed

+1
-1
lines changed

1 file changed

+1
-1
lines changed

torchao/quantization/GPTQ.py

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -933,7 +933,7 @@ def linear_forward_8da4w(
933933
groupsize,
934934
precision,
935935
):
936-
x = per_token_dynamic_quant(x)
936+
x = per_token_dynamic_quant(x, scale_dtype=precision, zero_point_dtype=precision)
937937
# TODO: verify and remove following reshape code
938938
# origin_x_size = x.size()
939939
# x = x.reshape(-1, origin_x_size[-1])

0 commit comments

Comments
 (0)