vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

musa: enable building fat binaries, enable unified memory, and disable Flash Attention on QY1 (MTT S80) (#9526)

Browse source 
* mtgpu: add mp_21 support

Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>

* mtgpu: disable flash attention on qy1 (MTT S80); disable q3_k and mul_mat_batched_cublas

Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>

* mtgpu: enable unified memory

Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>

* mtgpu: map cublasOperation_t to mublasOperation_t (sync code to latest)

Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>

---------

Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
This commit is contained in:
R0CKSTAR 2024-09-22 22:55:49 +08:00 • committed by GitHub
parent 912c331d3d
commit c35e586ea5
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
6 changed files with 31 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

18
ggml/src/ggml-cuda.cu
View file

@ -136,7 +136,7 @@ static cudaError_t ggml_cuda_device_malloc(void ** ptr, size_t size, int device)
return res;
#else
#if !defined(GGML_USE_HIPBLAS) && !defined(GGML_USE_MUSA)
#if !defined(GGML_USE_HIPBLAS)
cudaError_t err;
if (getenv("GGML_CUDA_ENABLE_UNIFIED_MEMORY") != nullptr)
{
@ -149,7 +149,7 @@ static cudaError_t ggml_cuda_device_malloc(void ** ptr, size_t size, int device)
return err;
#else
return cudaMalloc(ptr, size);
#endif // !defined(GGML_USE_HIPBLAS) && !defined(GGML_USE_MUSA)
#endif // !defined(GGML_USE_HIPBLAS)
#endif
}
@ -2830,6 +2830,12 @@ GGML_CALL static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, cons
if (op->op == GGML_OP_MUL_MAT && a->ne[3] != b->ne[3]) {
return false;
}
#ifdef GGML_USE_MUSA
if (b->type == GGML_TYPE_F16 && b->ne[2]*b->ne[3] > 1 &&
!ggml_is_transposed(a) && !ggml_is_transposed(b)) {
return false;
}
#endif // GGML_USE_MUSA
switch (a->type) {
case GGML_TYPE_F32:
case GGML_TYPE_F16:
@ -2853,6 +2859,11 @@ GGML_CALL static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, cons
case GGML_TYPE_IQ3_XXS:
case GGML_TYPE_IQ4_NL:
case GGML_TYPE_IQ4_XS:
#ifdef GGML_USE_MUSA
if (a->type == GGML_TYPE_Q3_K) {
return false;
}
#endif // GGML_USE_MUSA
return true;
default:
return false;
@ -2978,6 +2989,9 @@ GGML_CALL static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, cons
case GGML_OP_RWKV_WKV:
return true;
case GGML_OP_FLASH_ATTN_EXT: {
#ifndef FLASH_ATTN_AVAILABLE
return false;
#endif
if (op->src[0]->ne[0] == 64 && op->src[1]->type == GGML_TYPE_F16) {
return true;
}