From 943f9c2f570ec39684c47bc3df427a222b680502 Mon Sep 17 00:00:00 2001 From: Disty0 Date: Wed, 15 Jul 2026 19:46:11 +0300 Subject: [PATCH] check pv_matmul_dtype = "enabled" --- modules/sdnq/kernels/triton_atten.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/modules/sdnq/kernels/triton_atten.py b/modules/sdnq/kernels/triton_atten.py index 136f36029..e23b994c8 100644 --- a/modules/sdnq/kernels/triton_atten.py +++ b/modules/sdnq/kernels/triton_atten.py @@ -203,7 +203,7 @@ def quantize_attn( ) -> tuple[torch.Tensor]: if matmul_dtype in {"auto", "enabled", "uint8"}: matmul_dtype = "int8" - if pv_matmul_dtype == "uint8": + if pv_matmul_dtype in {"enabled", "uint8"}: pv_matmul_dtype = "int8" if scale is None: scale = q.shape[-1] ** -0.5