Skip to content

Commit b253af8

Browse files
Honor configured A16 weight ranges (pytorch#22769) (pytorch#22769)
Summary: Forward the existing weight_qmin and weight_qmax arguments into the Arm symmetric A16W8 quantization configuration in both the fbcode and xplat mirrors. Callers that omit the arguments retain the existing [-127, 127] weight range; restricted-range recipes receive the requested qspec. Reviewed By: WongJohnson Differential Revision: D119779205
1 parent 3235f58 commit b253af8

1 file changed

Lines changed: 2 additions & 0 deletions

File tree

‎backends/arm/quantizer/arm_quantizer.py‎

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -484,6 +484,8 @@ def get_symmetric_a16w8_quantization_config(
484484
is_per_channel=is_per_channel,
485485
is_qat=is_qat,
486486
is_dynamic=is_dynamic,
487+
weight_qmin=weight_qmin,
488+
weight_qmax=weight_qmax,
487489
)
488490

489491
if is_dynamic:

0 commit comments

Comments
 (0)