Public/automatic
1
0
Fork 0
mirror of https://github.com/vladmandic/automatic synced 2026-09-19 17:24:32 +02:00
Code Issues Packages Projects Releases Wiki Activity

fix fp16 mm is selected instead of fp8 mm

Browse Source
This commit is contained in:
Disty0
2026-03-11 03:55:43 +03:00
parent d9e628574a
commit 76b184a9fe
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/sdnq/quantizer.py
+1 -1
View File
@@ -245,7 +245,7 @@ def sdnq_quantize_layer_weight(weight, layer_class_name=None, weights_dtype="int
if quantized_matmul_dtype is None:
if dtype_dict[weights_dtype]["is_integer"]:
quantized_matmul_dtype = "int8"
elif dtype_dict[weights_dtype]["num_bits"] == 8:
elif dtype_dict[weights_dtype]["num_bits"] <= 12:
quantized_matmul_dtype = "float8_e4m3fn"
else:
quantized_matmul_dtype = "float16"