add GGML_LIST_DEVICE function
This commit is contained in:
jianyuzh
Meng, Hengyu
parent
43f2c35859
commit
da752edaf5
4 changed files with 56 additions and 7 deletions
|
|
@ -474,13 +474,14 @@ if (LLAMA_SYCL)
|
|||
|
||||
if (_sycl_support)
|
||||
add_compile_definitions(GGML_USE_CUBLAS)
|
||||
#add_compile_definitions(GGML_SYCL_F16)
|
||||
#add_compile_options(-std=c++17 -O3 -fsycl)
|
||||
add_compile_options(-I/opt/intel/oneapi/compiler/2024.0/include)
|
||||
add_compile_options(-I/opt/intel/oneapi/compiler/2024.0/include/sycl)
|
||||
add_compile_options(-I/opt/intel/oneapi/dpcpp-ct/2024.0/include)
|
||||
add_compile_options(-I/opt/intel/oneapi/2024.0/include)
|
||||
|
||||
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++17")
|
||||
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++17 -Wno-narrowing")
|
||||
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O3")
|
||||
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsycl -L${MKLROOT}/lib")
|
||||
|
||||
|
|
|
|||
|
|
@ -43,7 +43,17 @@ If you want to get more binary files, please change the build prject.
|
|||
|
||||
1. Install Intel oneAPI Base toolkit.
|
||||
|
||||
2. Setup Local
|
||||
2. Build locally:
|
||||
|
||||
```
|
||||
mkdir -p build
|
||||
cd build
|
||||
source /opt/intel/oneapi/setvars.sh
|
||||
cmake .. -DLLAMA_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx
|
||||
cmake --build . --config Release -v
|
||||
```
|
||||
|
||||
or
|
||||
|
||||
```
|
||||
./setup.sh
|
||||
|
|
@ -56,17 +66,22 @@ If you want to get more binary files, please change the build prject.
|
|||
Run without parameter:
|
||||
|
||||
```
|
||||
export GGML_SYCL_LIST_DEVICE=1
|
||||
./build/bin/main
|
||||
```
|
||||
|
||||
Check the id in startup log, like:
|
||||
ggml_init_cublas: found 6 CUDA devices:
|
||||
|
||||
```
|
||||
SYCL devices:
|
||||
Device 0: Intel(R) Arc(TM) A770 Graphics, compute capability 1.3
|
||||
Device 1: Intel(R) FPGA Emulation Device, compute capability 1.2
|
||||
Device 2: 13th Gen Intel(R) Core(TM) i7-13700K, compute capability 3.0
|
||||
Device 3: Intel(R) Arc(TM) A770 Graphics, compute capability 3.0
|
||||
Device 4: Intel(R) UHD Graphics 770, compute capability 3.0
|
||||
Device 5: Intel(R) UHD Graphics 770, compute capability 1.3
|
||||
Exit for list devices task. unset GGML_SYCL_LIST_DEVICE to restore LLM work!
|
||||
```
|
||||
|
||||
#### Put model file to folder **models**
|
||||
|
||||
|
|
@ -133,17 +148,21 @@ Run the command in command line or powershell.
|
|||
Run without parameter:
|
||||
|
||||
```
|
||||
set GGML_SYCL_LIST_DEVICE=1
|
||||
.\x64\Release\llama.cpp.sycl.exe
|
||||
```
|
||||
|
||||
Check the id in startup log, like:
|
||||
ggml_init_cublas: found 6 CUDA devices:
|
||||
|
||||
```
|
||||
SYCL devices:
|
||||
Device 0: Intel(R) Arc(TM) A770 Graphics, compute capability 1.3
|
||||
Device 1: Intel(R) FPGA Emulation Device, compute capability 1.2
|
||||
Device 2: 13th Gen Intel(R) Core(TM) i7-13700K, compute capability 3.0
|
||||
Device 3: Intel(R) Arc(TM) A770 Graphics, compute capability 3.0
|
||||
Device 4: Intel(R) UHD Graphics 770, compute capability 3.0
|
||||
Device 5: Intel(R) UHD Graphics 770, compute capability 1.3
|
||||
Exit for list devices task. unset GGML_SYCL_LIST_DEVICE to restore LLM work!
|
||||
```
|
||||
|
||||
#### Put model file to folder **models**
|
||||
|
||||
|
|
@ -160,4 +179,26 @@ set GGML_SYCL_DEVICE=0
|
|||
#### Run
|
||||
```
|
||||
.\run.bat
|
||||
```
|
||||
```
|
||||
|
||||
### Environment Variable
|
||||
|
||||
#### Build
|
||||
|
||||
|Name|Value|Function|
|
||||
|-|-|-|
|
||||
|LLAMA_SYCL|ON (mandatory)|Enable build with SYCL code path|
|
||||
|CMAKE_C_COMPILER|icx|Use icx compiler for SYCL code path|
|
||||
|CMAKE_CXX_COMPILER|icpx|use icpx for SYCL code path|
|
||||
|GGML_SYCL_F16|OFF (default) or ON|Enable FP16 in computing|
|
||||
|
||||
|
||||
#### Running
|
||||
|
||||
|
||||
|Name|Value|Function|
|
||||
|-|-|-|
|
||||
|GGML_SYCL_DEVICE|0 (default) or 1|Set the device id used. List the device ids by **GGML_SYCL_LIST_DEVICE**|
|
||||
|GGML_SYCL_DEBUG|0 (default) or 1|Enable log funciton by macro: GGML_SYCL_DEBUG|
|
||||
|GGML_SYCL_LIST_DEVICE|0 (default) or 1|List the device ids only|
|
||||
|
||||
|
|
|
|||
|
|
@ -8282,7 +8282,6 @@ void print_devices(){
|
|||
|
||||
int get_sycl_env(const char* env_name, int default_val){
|
||||
char * user_device_string = getenv(env_name);
|
||||
printf("get_sycl_env=%s=%s\n", env_name, user_device_string);
|
||||
int user_number = default_val;
|
||||
|
||||
unsigned n;
|
||||
|
|
@ -8297,6 +8296,13 @@ int get_sycl_env(const char* env_name, int default_val){
|
|||
void ggml_init_cublas() try {
|
||||
static bool initialized = false;
|
||||
if (!initialized) {
|
||||
if (get_sycl_env("GGML_SYCL_LIST_DEVICE", 0)!=0){
|
||||
printf("SYCL devices:\n");
|
||||
print_devices();
|
||||
printf("Exit for list devices task. unset GGML_SYCL_LIST_DEVICE to restore LLM work!\n");
|
||||
std::exit(0);
|
||||
}
|
||||
|
||||
g_ggml_sycl_debug = get_sycl_env("GGML_SYCL_DEBUG", 0);
|
||||
|
||||
printf("g_ggml_sycl_debug=%d\n", g_ggml_sycl_debug);
|
||||
|
|
|
|||
1
run.sh
1
run.sh
|
|
@ -7,6 +7,7 @@ source /opt/intel/oneapi/setvars.sh
|
|||
|
||||
export GGML_SYCL_DEVIC=0
|
||||
export GGML_SYCL_DEBUG=1
|
||||
export GGML_SYCL_LIST_DEVICE=1
|
||||
#./build/bin/main -m models/llama-2-7b.Q4_0.gguf -p "${INPUT1}" -e -n 400 -ngl 33 -c 2048
|
||||
./build/bin/main -m models/llama-2-7b.Q4_0.gguf -p "${INPUT2}" -n 400 -e -ngl 33
|
||||
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue