Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-25 05:48:47 +01:00
Code Issues Packages Projects Releases Wiki Activity

fallback to CPU buffer if host buffer alloc fails (#4610)

Browse Source
This commit is contained in:
slaren 2023-12-23 16:10:51 +01:00 committed by GitHub
parent 925e5584a0
commit 708e179e85
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 17 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
ggml-cuda.cu
View File

@ -6729,8 +6729,7 @@ void * ggml_cuda_host_malloc(size_t size) {
void * ptr = nullptr; void * ptr = nullptr;
cudaError_t err = cudaMallocHost((void **) &ptr, size); cudaError_t err = cudaMallocHost((void **) &ptr, size);
if (err != cudaSuccess) { if (err != cudaSuccess) {
// The allocation error can be bypassed. A null ptr will assigned out of this function. // clear the error
// This can fixed the OOM error in WSL.
cudaGetLastError(); cudaGetLastError();
fprintf(stderr, "WARNING: failed to allocate %.2f MB of pinned memory: %s\n", fprintf(stderr, "WARNING: failed to allocate %.2f MB of pinned memory: %s\n",
size/1024.0/1024.0, cudaGetErrorString(err)); size/1024.0/1024.0, cudaGetErrorString(err));
@ -9674,12 +9673,14 @@ ggml_backend_buffer_type_t ggml_backend_cuda_buffer_type(int device) {
// host buffer type // host buffer type
static void ggml_backend_cuda_host_buffer_free_buffer(ggml_backend_buffer_t buffer) { static void ggml_backend_cuda_host_buffer_free_buffer(ggml_backend_buffer_t buffer) {
CUDA_CHECK(cudaFreeHost(buffer->context)); ggml_cuda_host_free(buffer->context);
} }
static ggml_backend_buffer_t ggml_backend_cuda_host_buffer_type_alloc_buffer(ggml_backend_buffer_type_t buft, size_t size) { static ggml_backend_buffer_t ggml_backend_cuda_host_buffer_type_alloc_buffer(ggml_backend_buffer_type_t buft, size_t size) {
void * ptr; void * ptr = ggml_cuda_host_malloc(size);
CUDA_CHECK(cudaMallocHost(&ptr, size)); if (ptr == nullptr) {
return nullptr;
}
// FIXME: this is a hack to avoid having to implement a new buffer type // FIXME: this is a hack to avoid having to implement a new buffer type
ggml_backend_buffer_t buffer = ggml_backend_cpu_buffer_from_ptr(ptr, size); ggml_backend_buffer_t buffer = ggml_backend_cpu_buffer_from_ptr(ptr, size);

16
llama.cpp
View File

@ -1177,21 +1177,27 @@ static std::string llama_token_to_piece(const struct llama_context * ctx, llama_
} }
static ggml_backend_buffer_type_t llama_default_buffer_type(int n_gpu_layers) { static ggml_backend_buffer_type_t llama_default_buffer_type(int n_gpu_layers) {
ggml_backend_buffer_type_t buft = nullptr;
#ifdef GGML_USE_METAL #ifdef GGML_USE_METAL
if (n_gpu_layers > 0) { if (n_gpu_layers > 0) {
return ggml_backend_metal_buffer_type(); buft = ggml_backend_metal_buffer_type();
} }
#elif defined(GGML_USE_CUBLAS) && defined(LLAMA_GGML_BACKEND_CUDA_TEST) #elif defined(GGML_USE_CUBLAS) && defined(LLAMA_GGML_BACKEND_CUDA_TEST)
if (n_gpu_layers > 0) { if (n_gpu_layers > 0) {
return ggml_backend_cuda_buffer_type(0); buft = ggml_backend_cuda_buffer_type(0);
} }
#elif defined(GGML_USE_CUBLAS) #elif defined(GGML_USE_CUBLAS)
return ggml_backend_cuda_host_buffer_type(); buft = ggml_backend_cuda_host_buffer_type();
#elif defined(GGML_USE_CPU_HBM) #elif defined(GGML_USE_CPU_HBM)
return ggml_backend_cpu_hbm_buffer_type(); buft = ggml_backend_cpu_hbm_buffer_type();
#endif #endif
return ggml_backend_cpu_buffer_type(); if (buft == nullptr) {
buft = ggml_backend_cpu_buffer_type();
}
return buft;
GGML_UNUSED(n_gpu_layers); GGML_UNUSED(n_gpu_layers);
} }