From d48c88cbd563b6cf0ce972e2f56796896e240736 Mon Sep 17 00:00:00 2001 From: Georgi Gerganov Date: Thu, 23 May 2024 10:00:44 +0300 Subject: ggml : remove ggml_flash_attn and ggml_flash_ff (#7463) ggml-ci --- ggml.h | 18 +----------------- 1 file changed, 1 insertion(+), 17 deletions(-) (limited to 'ggml.h') diff --git a/ggml.h b/ggml.h index 08835042..be81e0c5 100644 --- a/ggml.h +++ b/ggml.h @@ -481,9 +481,7 @@ extern "C" { GGML_OP_ARGSORT, GGML_OP_LEAKY_RELU, - GGML_OP_FLASH_ATTN, GGML_OP_FLASH_ATTN_EXT, - GGML_OP_FLASH_FF, GGML_OP_FLASH_ATTN_BACK, GGML_OP_SSM_CONV, GGML_OP_SSM_SCAN, @@ -1761,13 +1759,6 @@ extern "C" { struct ggml_tensor * a, int k); - GGML_API struct ggml_tensor * ggml_flash_attn( - struct ggml_context * ctx, - struct ggml_tensor * q, - struct ggml_tensor * k, - struct ggml_tensor * v, - bool masked); - #define GGML_KQ_MASK_PAD 32 // q: [n_embd, n_batch, n_head, 1] @@ -1788,6 +1779,7 @@ extern "C" { struct ggml_tensor * a, enum ggml_prec prec); + // TODO: needs to be adapted to ggml_flash_attn_ext GGML_API struct ggml_tensor * ggml_flash_attn_back( struct ggml_context * ctx, struct ggml_tensor * q, @@ -1796,14 +1788,6 @@ extern "C" { struct ggml_tensor * d, bool masked); - GGML_API struct ggml_tensor * ggml_flash_ff( - struct ggml_context * ctx, - struct ggml_tensor * a, - struct ggml_tensor * b0, - struct ggml_tensor * b1, - struct ggml_tensor * c0, - struct ggml_tensor * c1); - GGML_API struct ggml_tensor * ggml_ssm_conv( struct ggml_context * ctx, struct ggml_tensor * s, -- cgit v1.2.3