From 1cf2c9012d1d8d696e3c87bbd9d59a530656baf0 Mon Sep 17 00:00:00 2001 From: Disty0 Date: Wed, 15 Jul 2026 03:20:43 +0300 Subject: [PATCH] add min block size check to sdnq atten --- modules/attention.py | 1 + 1 file changed, 1 insertion(+) diff --git a/modules/attention.py b/modules/attention.py index 00456d871..f52973e6a 100644 --- a/modules/attention.py +++ b/modules/attention.py @@ -26,6 +26,7 @@ def set_sdnq_attention(): def sdpa_sdnq_atten(query: torch.FloatTensor, key: torch.FloatTensor, value: torch.FloatTensor, attn_mask: torch.Tensor | None = None, dropout_p: float = 0.0, is_causal: bool = False, scale: float | None = None, enable_gqa: bool = False, **kwargs) -> torch.FloatTensor: if ( query.device.type != "cpu" + and (query.shape[-2] >= 32 and key.shape[-2] >= 32 and value.shape[-2] >= 32) and (query.shape[-2] >= 512 or key.shape[-2] >= 512) # Skip TE and query.shape[-3] > 1 # Skip VAE ):