From 95fb0aefab568348da159efdd370e064d1b35f97 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Johannes=20G=C3=A4=C3=9Fler?= Date: Wed, 22 May 2024 10:24:29 +0200 Subject: [PATCH] CUDA: remove incorrect precision check (#7454) --- ggml-cuda/fattn-tile-f32.cu | 3 --- 1 file changed, 3 deletions(-) diff --git a/ggml-cuda/fattn-tile-f32.cu b/ggml-cuda/fattn-tile-f32.cu index 130e7cbdb..54db765e2 100644 --- a/ggml-cuda/fattn-tile-f32.cu +++ b/ggml-cuda/fattn-tile-f32.cu @@ -286,9 +286,6 @@ void ggml_cuda_flash_attn_ext_tile_f32(ggml_backend_cuda_context & ctx, ggml_ten const ggml_tensor * KQV = dst; const ggml_tensor * Q = dst->src[0]; - const int32_t precision = KQV->op_params[2]; - GGML_ASSERT(precision == GGML_PREC_DEFAULT); - if (Q->ne[1] <= 16) { constexpr int cols_per_block = 16; constexpr int parallel_blocks = 4;