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

tensor core info better printout (+1 squashed commits)

Browse source 
Squashed commits:

[be4ef93f] tensor core info better printout
This commit is contained in:
Concedo 2023-11-04 08:37:12 +08:00
parent 36f43ae834
commit 38471fbe06
1 changed files with 2 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

12
ggml-cuda.cu
View file

@ -5827,16 +5827,8 @@ void ggml_init_cublas() {
CUDA_CHECK(cudaGetDeviceCount(&g_device_count));
GGML_ASSERT(g_device_count <= GGML_CUDA_MAX_DEVICES);
int64_t total_vram = 0;
#if defined(GGML_CUDA_FORCE_MMQ)
fprintf(stderr, "%s: GGML_CUDA_FORCE_MMQ: yes\n", __func__);
#else
fprintf(stderr, "%s: GGML_CUDA_FORCE_MMQ: no\n", __func__);
#endif
#if defined(CUDA_USE_TENSOR_CORES)
fprintf(stderr, "%s: CUDA_USE_TENSOR_CORES: yes\n", __func__);
#else
fprintf(stderr, "%s: CUDA_USE_TENSOR_CORES: no\n", __func__);
#endif
fprintf(stderr, "%s: GGML_CUDA_FORCE_MMQ: %s\n", __func__,(g_mul_mat_q?"yes":"maybe"));
fprintf(stderr, "%s: CUDA_USE_TENSOR_CORES: %s\n", __func__,(g_mul_mat_q?"no":"maybe"));
fprintf(stderr, "%s: found %d " GGML_CUDA_NAME " devices:\n", __func__, g_device_count);
for (int id = 0; id < g_device_count; ++id) {
cudaDeviceProp prop;