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

llama : remove LLAMA_MAX_DEVICES and LLAMA_SUPPORTS_GPU_OFFLOAD (#5240)

Browse source 
* llama : remove LLAMA_MAX_DEVICES from llama.h

ggml-ci

* Update llama.cpp

Co-authored-by: slaren <slarengh@gmail.com>

* server : remove LLAMA_MAX_DEVICES

ggml-ci

* llama : remove LLAMA_SUPPORTS_GPU_OFFLOAD

ggml-ci

* train : remove LLAMA_SUPPORTS_GPU_OFFLOAD

* readme : add deprecation notice

* readme : change deprecation notice to "remove" and fix url

* llama : remove gpu includes from llama.h

ggml-ci

---------

Co-authored-by: slaren <slarengh@gmail.com>
This commit is contained in:
Georgi Gerganov 2024-01-31 17:30:17 +02:00 • committed by GitHub
parent efb7bdbbd0
commit 5cb04dbc16
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
9 changed files with 143 additions and 124 deletions
Download patch file Download diff file Expand all files Collapse all files

39
llama.cpp
View file

@ -10090,18 +10090,45 @@ struct llama_model_quantize_params llama_model_quantize_default_params() {
return result;
}
int32_t llama_max_devices(void) {
return LLAMA_MAX_DEVICES;
size_t llama_max_devices(void) {
#if defined(GGML_USE_METAL)
return 1;
#elif defined(GGML_USE_CUBLAS)
return GGML_CUDA_MAX_DEVICES;
#elif defined(GGML_USE_SYCL)
return GGML_SYCL_MAX_DEVICES;
#else
return 1;
#endif
}
bool llama_mmap_supported(void) {
bool llama_supports_mmap(void) {
return llama_mmap::SUPPORTED;
}
bool llama_mlock_supported(void) {
bool llama_supports_mlock(void) {
return llama_mlock::SUPPORTED;
}
bool llama_supports_gpu_offload(void) {
#if defined(GGML_USE_CUBLAS) || defined(GGML_USE_CLBLAST) || defined(GGML_USE_METAL) || defined(GGML_USE_VULKAN) || \
defined(GGML_USE_SYCL) || defined(GGML_USE_KOMPUTE)
// Defined when llama.cpp is compiled with support for offloading model layers to GPU.
return true;
#else
return false;
#endif
}
// deprecated:
bool llama_mmap_supported(void) {
return llama_supports_mmap();
}
bool llama_mlock_supported(void) {
return llama_supports_mlock();
}
void llama_backend_init(bool numa) {
ggml_time_init();
@ -10133,8 +10160,8 @@ int64_t llama_time_us(void) {
}
struct llama_model * llama_load_model_from_file(
const char * path_model,
struct llama_model_params params) {
const char * path_model,
struct llama_model_params params) {
ggml_time_init();
llama_model * model = new llama_model;