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

try HIP fix

Browse source
This commit is contained in:
JohannesGaessler 2023-12-31 17:51:53 +01:00
parent 44d626239f
commit 8dd90131d3
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
ggml-cuda.cu
View file

@ -9316,8 +9316,8 @@ static ggml_tensor_extra_gpu * ggml_cuda_alloc_temp_tensor_extra() {
if (g_temp_tensor_extras == nullptr) { if (g_temp_tensor_extras == nullptr) {
g_temp_tensor_extras = new ggml_tensor_extra_gpu[GGML_CUDA_MAX_NODES]; g_temp_tensor_extras = new ggml_tensor_extra_gpu[GGML_CUDA_MAX_NODES];
for (int64_t i = 0; i < GGML_CUDA_MAX_NODES; ++i) { for (int64_t i = 0; i < GGML_CUDA_MAX_NODES; ++i) {
CUDA_CHECK(cudaEventCreate(&g_temp_tensor_extras[i].src0_done, cudaEventDisableTiming)); CUDA_CHECK(cudaEventCreateWithFlags(&g_temp_tensor_extras[i].src0_done, cudaEventDisableTiming));
CUDA_CHECK(cudaEventCreate(&g_temp_tensor_extras[i].src1_done, cudaEventDisableTiming)); CUDA_CHECK(cudaEventCreateWithFlags(&g_temp_tensor_extras[i].src1_done, cudaEventDisableTiming));
} }
} }