SDNQ set sdnq_dequantize_fp32 to True by default

This commit is contained in:
Disty0
2026-03-25 21:46:17 +03:00
parent 688485018e
commit a255e5bdcf
+1 -1
View File
@@ -159,7 +159,7 @@ def create_settings(cmd_opts):
"sdnq_use_quantized_matmul": OptionInfo(False, "Use quantized MatMul", gr.Checkbox),
"sdnq_use_quantized_matmul_conv": OptionInfo(False, "Use quantized MatMul with conv", gr.Checkbox),
"sdnq_quantize_with_gpu": OptionInfo(True, "Quantize using GPU", gr.Checkbox),
"sdnq_dequantize_fp32": OptionInfo(False, "Dequantize using full precision", gr.Checkbox),
"sdnq_dequantize_fp32": OptionInfo(True, "Dequantize using full precision", gr.Checkbox),
"sdnq_quantize_shuffle_weights": OptionInfo(False, "Shuffle weights in post mode", gr.Checkbox),
"nunchaku_sep": OptionInfo("<h2>Nunchaku Engine</h2>", "", gr.HTML),