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

try MUSA fix

Browse source
This commit is contained in:
Johannes Gäßler 2025-01-05 21:35:12 +01:00
parent ab20aa99c2
commit 6a5cdad219
1 changed files with 3 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

3
ggml/src/ggml-cuda/vendors/musa.h vendored
View file

@ -3,6 +3,7 @@
#include <musa_runtime.h> #include <musa_runtime.h>
#include <musa.h> #include <musa.h>
#include <mublas.h> #include <mublas.h>
#include <musa_bf16.h>
#include <musa_fp16.h> #include <musa_fp16.h>
#define CUBLAS_COMPUTE_16F CUDA_R_16F #define CUBLAS_COMPUTE_16F CUDA_R_16F
#define CUBLAS_COMPUTE_32F CUDA_R_32F #define CUBLAS_COMPUTE_32F CUDA_R_32F
@ -132,3 +133,5 @@
#define cudaKernelNodeParams musaKernelNodeParams #define cudaKernelNodeParams musaKernelNodeParams
#define cudaStreamCaptureModeRelaxed musaStreamCaptureModeRelaxed #define cudaStreamCaptureModeRelaxed musaStreamCaptureModeRelaxed
#define cudaStreamEndCapture musaStreamEndCapture #define cudaStreamEndCapture musaStreamEndCapture
typedef musa_bfloat16 nv_bfloat16;