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

29
llama.h
View file

@ -3,15 +3,7 @@
#include "ggml.h"
#include "ggml-backend.h"
#ifdef GGML_USE_CUBLAS
#include "ggml-cuda.h"
#define LLAMA_MAX_DEVICES GGML_CUDA_MAX_DEVICES
#elif defined(GGML_USE_SYCL)
#include "ggml-sycl.h"
#define LLAMA_MAX_DEVICES GGML_SYCL_MAX_DEVICES
#else
#define LLAMA_MAX_DEVICES 1
#endif // GGML_USE_CUBLAS
#include <stddef.h>
#include <stdint.h>
#include <stdio.h>
@ -49,12 +41,6 @@
#define LLAMA_SESSION_MAGIC LLAMA_FILE_MAGIC_GGSN
#define LLAMA_SESSION_VERSION 4
#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.
#define LLAMA_SUPPORTS_GPU_OFFLOAD
#endif
#ifdef __cplusplus
extern "C" {
#endif
@ -201,7 +187,7 @@ extern "C" {
// LLAMA_SPLIT_LAYER: ignored
int32_t main_gpu;
// proportion of the model (layers or rows) to offload to each GPU, size: LLAMA_MAX_DEVICES
// proportion of the model (layers or rows) to offload to each GPU, size: llama_max_devices()
const float * tensor_split;
// Called with a progress value between 0.0 and 1.0. Pass NULL to disable.
@ -338,9 +324,14 @@ extern "C" {
LLAMA_API int64_t llama_time_us(void);
LLAMA_API int32_t llama_max_devices(void);
LLAMA_API bool llama_mmap_supported (void);
LLAMA_API bool llama_mlock_supported(void);
LLAMA_API size_t llama_max_devices(void);
LLAMA_API bool llama_supports_mmap (void);
LLAMA_API bool llama_supports_mlock (void);
LLAMA_API bool llama_supports_gpu_offload(void);
LLAMA_API DEPRECATED(bool llama_mmap_supported (void), "use llama_supports_mmap() instead");
LLAMA_API DEPRECATED(bool llama_mlock_supported(void), "use llama_supports_mlock() instead");
LLAMA_API const struct llama_model * llama_get_model(const struct llama_context * ctx);