Public/automatic
1
0
Fork 0
mirror of https://github.com/vladmandic/automatic synced 2026-08-26 15:16:01 +02:00
Code Issues Packages Projects Releases Wiki Activity

SDNQ fix quantization_device getting ignored on post load quant

Browse Source
This commit is contained in:
Disty0
2025-12-08 01:29:52 +03:00
parent 064b64c76c
commit 3ae7ecdbad
1 changed files with 2 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/sdnq/quantizer.py
+2 -3
View File
@@ -396,11 +396,10 @@ def sdnq_quantize_layer(layer, weights_dtype="int8", quantized_matmul_dtype=None
return layer
use_quantized_matmul = use_quantized_matmul_conv
layer.weight.requires_grad = False
layer.weight.requires_grad_(False)
if return_device is None:
return_device = layer.weight.device
if quantization_device is None:
quantization_device = layer.weight.device
if quantization_device is not None:
layer.weight.data = layer.weight.to(quantization_device, non_blocking=non_blocking)
(