From 76b184a9fed07d6e70cb3ecfc454fbc89865c8d0 Mon Sep 17 00:00:00 2001 From: Disty0 Date: Wed, 11 Mar 2026 03:55:43 +0300 Subject: [PATCH] fix fp16 mm is selected instead of fp8 mm --- modules/sdnq/quantizer.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/modules/sdnq/quantizer.py b/modules/sdnq/quantizer.py index a9a8cd733..59017a29c 100644 --- a/modules/sdnq/quantizer.py +++ b/modules/sdnq/quantizer.py @@ -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"