Skip to content

Commit

Permalink
metal : disable FA kernel for HS=256 (#7556)
Browse files Browse the repository at this point in the history 
ggml-ci
  • Loading branch information
@ggerganov
ggerganov authored May 27, 2024
1 parent eaf6e03 commit 62bfef5
Show file tree
Hide file tree
Showing 2 changed files with 11 additions and 8 deletions.
15 changes: 9 additions & 6 deletions ggml-metal.m
View file
Original file line number Diff line number Diff line change
Expand Up @@ -184,9 +184,9 @@
GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_F16_H96,
GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_F16_H112,
GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_F16_H128,
GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_F16_H256,
//GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_F16_H256, // https://github.com/ggerganov/llama.cpp/issues/7261
GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_VEC_F16_H128,
GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_VEC_F16_H256,
//GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_VEC_F16_H256, // https://github.com/ggerganov/llama.cpp/issues/7261
GGML_METAL_KERNEL_TYPE_CPY_F32_F16,
GGML_METAL_KERNEL_TYPE_CPY_F32_F32,
GGML_METAL_KERNEL_TYPE_CPY_F32_Q8_0,
Expand Down Expand Up @@ -634,9 +634,9 @@ static void ggml_metal_log(enum ggml_log_level level, const char * format, ...){
GGML_METAL_ADD_KERNEL(GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_F16_H96, flash_attn_ext_f16_h96, ctx->support_simdgroup_mm);
GGML_METAL_ADD_KERNEL(GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_F16_H112, flash_attn_ext_f16_h112, ctx->support_simdgroup_mm);
GGML_METAL_ADD_KERNEL(GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_F16_H128, flash_attn_ext_f16_h128, ctx->support_simdgroup_mm);
GGML_METAL_ADD_KERNEL(GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_F16_H256, flash_attn_ext_f16_h256, ctx->support_simdgroup_mm);
//GGML_METAL_ADD_KERNEL(GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_F16_H256, flash_attn_ext_f16_h256, ctx->support_simdgroup_mm);
GGML_METAL_ADD_KERNEL(GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_VEC_F16_H128, flash_attn_ext_vec_f16_h128, ctx->support_simdgroup_reduction);
GGML_METAL_ADD_KERNEL(GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_VEC_F16_H256, flash_attn_ext_vec_f16_h256, ctx->support_simdgroup_reduction);
//GGML_METAL_ADD_KERNEL(GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_VEC_F16_H256, flash_attn_ext_vec_f16_h256, ctx->support_simdgroup_reduction);
GGML_METAL_ADD_KERNEL(GGML_METAL_KERNEL_TYPE_CPY_F32_F16, cpy_f32_f16, true);
GGML_METAL_ADD_KERNEL(GGML_METAL_KERNEL_TYPE_CPY_F32_F32, cpy_f32_f32, true);
GGML_METAL_ADD_KERNEL(GGML_METAL_KERNEL_TYPE_CPY_F32_Q8_0, cpy_f32_q8_0, true);
Expand Down Expand Up @@ -770,6 +770,9 @@ static bool ggml_metal_supports_op(const struct ggml_metal_context * ctx, const
case GGML_OP_LEAKY_RELU:
return true;
case GGML_OP_FLASH_ATTN_EXT:
if (op->src[0]->ne[0] == 256) {
return false;
}
return ctx->support_simdgroup_mm; // TODO: over-restricted for vec-kernels
case GGML_OP_MUL_MAT:
case GGML_OP_MUL_MAT_ID:
Expand Down Expand Up @@ -2573,7 +2576,7 @@ static enum ggml_status ggml_metal_graph_compute(
case 96: pipeline = ctx->kernels[GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_F16_H96 ].pipeline; break;
case 112: pipeline = ctx->kernels[GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_F16_H112].pipeline; break;
case 128: pipeline = ctx->kernels[GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_F16_H128].pipeline; break;
case 256: pipeline = ctx->kernels[GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_F16_H256].pipeline; break;
//case 256: pipeline = ctx->kernels[GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_F16_H256].pipeline; break;
default:
{
GGML_METAL_LOG_ERROR("unsupported size: %lld\n", ne00);
Expand All @@ -2586,7 +2589,7 @@ static enum ggml_status ggml_metal_graph_compute(

switch (ne00) {
case 128: pipeline = ctx->kernels[GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_VEC_F16_H128].pipeline; break;
case 256: pipeline = ctx->kernels[GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_VEC_F16_H256].pipeline; break;
//case 256: pipeline = ctx->kernels[GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_VEC_F16_H256].pipeline; break;
default:
{
GGML_METAL_LOG_ERROR("unsupported size: %lld\n", ne00);
Expand Down
4 changes: 2 additions & 2 deletions ggml-metal.metal
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2418,7 +2418,7 @@ template [[host_name("kernel_flash_attn_ext_f16_h80" )]] kernel flash_attn_ext_f
template [[host_name("kernel_flash_attn_ext_f16_h96" )]] kernel flash_attn_ext_f16_t kernel_flash_attn_ext_f16<96>;
template [[host_name("kernel_flash_attn_ext_f16_h112")]] kernel flash_attn_ext_f16_t kernel_flash_attn_ext_f16<112>;
template [[host_name("kernel_flash_attn_ext_f16_h128")]] kernel flash_attn_ext_f16_t kernel_flash_attn_ext_f16<128>;
template [[host_name("kernel_flash_attn_ext_f16_h256")]] kernel flash_attn_ext_f16_t kernel_flash_attn_ext_f16<256>;
//template [[host_name("kernel_flash_attn_ext_f16_h256")]] kernel flash_attn_ext_f16_t kernel_flash_attn_ext_f16<256>;

template<int64_t D, int64_t Q = 1, int64_t C = 32> // head size, queries per threadgroup, cache items per threadgroup
kernel void kernel_flash_attn_ext_vec_f16(
Expand Down Expand Up @@ -2696,7 +2696,7 @@ kernel void kernel_flash_attn_ext_vec_f16(
}

template [[host_name("kernel_flash_attn_ext_vec_f16_h128")]] kernel flash_attn_ext_f16_t kernel_flash_attn_ext_vec_f16<128>;
template [[host_name("kernel_flash_attn_ext_vec_f16_h256")]] kernel flash_attn_ext_f16_t kernel_flash_attn_ext_vec_f16<256>;
//template [[host_name("kernel_flash_attn_ext_vec_f16_h256")]] kernel flash_attn_ext_f16_t kernel_flash_attn_ext_vec_f16<256>;

kernel void kernel_cpy_f16_f16(
device const half * src0,
Expand Down

0 comments on commit 62bfef5

Please sign in to comment.