10000 sync : ggml (conv ops + cuda MSVC fixes) by ggerganov · Pull Request #3765 · ggml-org/llama.cpp · GitHub
[go: up one dir, main page]
Skip to content

sync : ggml (conv ops + cuda MSVC fixes) #3765

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Oct 24, 2023
Merged

sync : ggml (conv ops + cuda MSVC fixes) #3765

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
sync : ggml (conv ops + cuda MSVC fixes) 
ggml-ci
  • Loading branch information
@ggerganov
ggerganov committed Oct 24, 2023
commit 58f8ddd0f526a1ffaf8d3e83af3f7bf913af4e6f
10 changes: 5 additions & 5 deletions ggml-cuda.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5664,10 +5664,10 @@ void ggml_init_cublas() {
GGML_ASSERT(g_device_count <= GGML_CUDA_MAX_DEVICES);
int64_t total_vram = 0;
fprintf(stderr, "%s: found %d " GGML_CUDA_NAME " devices:\n", __func__, g_device_count);
for (int64_t id = 0; id < g_device_count; ++id) {
for (int id = 0; id < g_device_count; ++id) {
cudaDeviceProp prop;
CUDA_CHECK(cudaGetDeviceProperties(&prop, id));
fprintf(stderr, " Device %ld: %s, compute capability %d.%d\n", id, prop.name, prop.major, prop.minor);
fprintf(stderr, " Device %d: %s, compute capability %d.%d\n", id, prop.name, prop.major, prop.minor);

g_tensor_split[id] = total_vram;
total_vram += prop.totalGlobalMem;
Expand All @@ -5677,15 +5677,15 @@ void ggml_init_cublas() {
g_compute_capab 7632 ilities[id] = 100*prop.major + 10*prop.minor;
#endif // defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)
}
for (int64_t id = 0; id < g_device_count; ++id) {
for (int id = 0; id < g_device_count; ++id) {
g_tensor_split[id] /= total_vram;
}

for (int64_t id = 0; id < g_device_count; ++id) {
for (int id = 0; id < g_device_count; ++id) {
CUDA_CHECK(ggml_cuda_set_device(id));

// create cuda streams
for (int64_t is = 0; is < MAX_STREAMS; ++is) {
for (int is = 0; is < MAX_STREAMS; ++is) {
CUDA_CHECK(cudaStreamCreateWithFlags(&g_cudaStreams[id][is], cudaStreamNonBlocking));
}

Expand Down
Loading
0