whisper : add SYCL support (#1863)

* add changes from llama upstream

* add sycl abstraction

* add sycl build

* update cmake

* add sycl build config

* fix bug

* fix bug

* refactor build

* fix bug

* update build

* call build

* use sycl header

* add examples

* add target

* fix typecast in quant.c

* readd fp16 and readme

* fix quant typecast

* add sample

* add readme

* remove cxx file check
This commit is contained in:
Abhilash Majumder 2024-02-23 12:52:24 +05:30 committed by GitHub
parent a2506909b1
commit a0ddd8392c
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
10 changed files with 510 additions and 6 deletions

View File

@ -150,6 +150,106 @@ jobs:
make
ctest -L gh --output-on-failure'
ubuntu-22-cmake-sycl:
runs-on: ubuntu-22.04
strategy:
fail-fast: false
matrix:
dwhisper_sycl: [ON]
dcmake_c_compiler: [icx]
dcmake_cxx_compiler: [icpx]
arch: [linux/amd64, linux/arm64, linux/arm/v7, linux/ppc64le]
continue-on-error: true
steps:
- name: Clone
uses: actions/checkout@v3
- name: add oneAPI to apt
shell: bash
run: |
cd /tmp
wget https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB
sudo apt-key add GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB
rm GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB
sudo add-apt-repository "deb https://apt.repos.intel.com/oneapi all main"
- name: install oneAPI dpcpp compiler
shell: bash
run: |
sudo apt update
sudo apt install intel-oneapi-compiler-dpcpp-cpp
- name: install oneAPI MKL library
shell: bash
run: |
sudo apt install intel-oneapi-mkl-devel
- name: Clone
id: checkout
uses: actions/checkout@v3
- name: Build
id: cmake_build
run: |
source /opt/intel/oneapi/setvars.sh
mkdir build
cd build
cmake -DWHISPER_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx ..
cmake --build . --config Release -j $(nproc)
ubuntu-22-cmake-sycl-fp16:
runs-on: ubuntu-22.04
strategy:
fail-fast: false
matrix:
dwhisper_sycl: [ON]
dcmake_c_compiler: [icx]
dcmake_cxx_compiler: [icpx]
arch: [linux/amd64, linux/arm64, linux/arm/v7, linux/ppc64le]
continue-on-error: true
steps:
- name: Clone
uses: actions/checkout@v3
- name: add oneAPI to apt
shell: bash
run: |
cd /tmp
wget https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB
sudo apt-key add GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB
rm GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB
sudo add-apt-repository "deb https://apt.repos.intel.com/oneapi all main"
- name: install oneAPI dpcpp compiler
shell: bash
run: |
sudo apt update
sudo apt install intel-oneapi-compiler-dpcpp-cpp
- name: install oneAPI MKL library
shell: bash
run: |
sudo apt install intel-oneapi-mkl-devel
- name: Clone
id: checkout
uses: actions/checkout@v3
- name: Build
id: cmake_build
run: |
source /opt/intel/oneapi/setvars.sh
mkdir build
cd build
cmake -DWHISPER_SYCL_F16=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx ..
cmake --build . --config Release -j $(nproc)
windows:
runs-on: windows-latest

View File

@ -70,12 +70,14 @@ if (APPLE)
option(WHISPER_COREML_ALLOW_FALLBACK "whisper: allow non-CoreML fallback" OFF)
option(WHISPER_METAL_EMBED_LIBRARY "whisper: embed Metal library" OFF)
else()
option(WHISPER_BLAS "whisper: use BLAS libraries" OFF)
option(WHISPER_BLAS_VENDOR "whisper: BLAS library vendor" Generic)
option(WHISPER_OPENBLAS "whisper: prefer OpenBLAS" OFF)
option(WHISPER_CUBLAS "whisper: support for cuBLAS" OFF)
option(WHISPER_HIPBLAS "whisper: support for hipBLAS" OFF)
option(WHISPER_CLBLAST "whisper: use CLBlast" OFF)
option(WHISPER_BLAS "whisper: use BLAS libraries" OFF)
option(WHISPER_BLAS_VENDOR "whisper: BLAS library vendor" Generic)
option(WHISPER_OPENBLAS "whisper: prefer OpenBLAS" OFF)
option(WHISPER_CUBLAS "whisper: support for cuBLAS" OFF)
option(WHISPER_HIPBLAS "whisper: support for hipBLAS" OFF)
option(WHISPER_CLBLAST "whisper: use CLBlast" OFF)
option(WHISPER_SYCL "whisper: use SYCL" OFF)
option(WHISPER_SYCL_F16 "whisper: use 16 bit floats for sycl calculations" OFF)
endif()
option(WHISPER_PERF "whisper: enable perf timings" OFF)
@ -106,6 +108,13 @@ endif()
find_package(Threads REQUIRED)
#compile flag sycl
if (WHISPER_SYCL)
set(CMAKE_CXX_STANDARD 17)
else()
set(CMAKE_CXX_STANDARD 11)
endif()
# on APPLE
if (APPLE)
# include Accelerate framework
@ -309,6 +318,30 @@ if( WHISPER_OPENVINO )
find_package(OpenVINO REQUIRED COMPONENTS Runtime)
endif()
if (WHISPER_SYCL)
if ( NOT DEFINED ENV{ONEAPI_ROOT})
message(FATAL_ERROR "Not detect ENV {ONEAPI_ROOT}, please install oneAPI & source it, like: source /opt/intel/oneapi/setvars.sh")
endif()
#todo: AOT
find_package(IntelSYCL REQUIRED)
if (WHISPER_SYCL_F16)
add_compile_definitions(GGML_SYCL_F16)
endif()
add_compile_definitions(GGML_USE_SYCL)
add_compile_options(-I./) #include DPCT
add_compile_options(-I/${SYCL_INCLUDE_DIR})
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-narrowing")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O3")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsycl -L${MKLROOT}/lib")
set(GGML_HEADERS_SYCL ggml-sycl.h)
set(GGML_SOURCES_SYCL ggml-sycl.cpp)
set(WHISPER_EXTRA_LIBS ${WHISPER_EXTRA_LIBS} sycl OpenCL mkl_core pthread m dl mkl_sycl_blas mkl_intel_ilp64 mkl_tbb_thread)
endif()
# compiler flags
if (NOT CMAKE_BUILD_TYPE AND NOT CMAKE_CONFIGURATION_TYPES)
@ -503,6 +536,8 @@ add_library(${TARGET}
${GGML_SOURCES_METAL}
${GGML_SOURCES_CUDA}
${GGML_SOURCES_OPENCL}
${GGML_SOURCES_SYCL}
${GGML_HEADERS_SYCL}
whisper.h
whisper.cpp
)

249
README_sycl.md Normal file
View File

@ -0,0 +1,249 @@
# whisper.cpp for SYCL
[Background](#background)
[OS](#os)
[Intel GPU](#intel-gpu)
[Linux](#linux)
[Environment Variable](#environment-variable)
[Known Issue](#known-issue)
[Todo](#todo)
## Background
SYCL is a higher-level programming model to improve programming productivity on various hardware accelerators—such as CPUs, GPUs, and FPGAs. It is a single-source embedded domain-specific language based on pure C++17.
oneAPI is a specification that is open and standards-based, supporting multiple architecture types including but not limited to GPU, CPU, and FPGA. The spec has both direct programming and API-based programming paradigms.
Intel uses the SYCL as direct programming language to support CPU, GPUs and FPGAs.
To avoid re-inventing the wheel, this code refers other code paths in llama.cpp (like OpenBLAS, cuBLAS, CLBlast). We use a open-source tool [SYCLomatic](https://github.com/oneapi-src/SYCLomatic) (Commercial release [Intel® DPC++ Compatibility Tool](https://www.intel.com/content/www/us/en/developer/tools/oneapi/dpc-compatibility-tool.html)) migrate to SYCL.
The whisper.cpp for SYCL is used to support Intel GPUs.
For Intel CPU, recommend to use whisper.cpp for X86 (Intel MKL build).
## OS
|OS|Status|Verified|
|-|-|-|
|Linux|Support|Ubuntu 22.04|
|Windows|Ongoing| |
## Intel GPU
|Intel GPU| Status | Verified Model|
|-|-|-|
|Intel Data Center Max Series| Support| Max 1550|
|Intel Data Center Flex Series| Support| Flex 170|
|Intel Arc Series| Support| Arc 770|
|Intel built-in Arc GPU| Support| built-in Arc GPU in Meteor Lake|
|Intel iGPU| Support| iGPU in i5-1250P, i7-1165G7|
## Linux
### Setup Environment
1. Install Intel GPU driver.
a. Please install Intel GPU driver by official guide: [Install GPU Drivers](https://dgpu-docs.intel.com/driver/installation.html).
Note: for iGPU, please install the client GPU driver.
b. Add user to group: video, render.
```
sudo usermod -aG render username
sudo usermod -aG video username
```
Note: re-login to enable it.
c. Check
```
sudo apt install clinfo
sudo clinfo -l
```
Output (example):
```
Platform #0: Intel(R) OpenCL Graphics
`-- Device #0: Intel(R) Arc(TM) A770 Graphics
Platform #0: Intel(R) OpenCL HD Graphics
`-- Device #0: Intel(R) Iris(R) Xe Graphics [0x9a49]
```
2. Install Intel® oneAPI Base toolkit.
a. Please follow the procedure in [Get the Intel® oneAPI Base Toolkit ](https://www.intel.com/content/www/us/en/developer/tools/oneapi/base-toolkit.html).
Recommend to install to default folder: **/opt/intel/oneapi**.
Following guide use the default folder as example. If you use other folder, please modify the following guide info with your folder.
b. Check
```
source /opt/intel/oneapi/setvars.sh
sycl-ls
```
There should be one or more level-zero devices. Like **[ext_oneapi_level_zero:gpu:0]**.
Output (example):
```
[opencl:acc:0] Intel(R) FPGA Emulation Platform for OpenCL(TM), Intel(R) FPGA Emulation Device OpenCL 1.2 [2023.16.10.0.17_160000]
[opencl:cpu:1] Intel(R) OpenCL, 13th Gen Intel(R) Core(TM) i7-13700K OpenCL 3.0 (Build 0) [2023.16.10.0.17_160000]
[opencl:gpu:2] Intel(R) OpenCL Graphics, Intel(R) Arc(TM) A770 Graphics OpenCL 3.0 NEO [23.30.26918.50]
[ext_oneapi_level_zero:gpu:0] Intel(R) Level-Zero, Intel(R) Arc(TM) A770 Graphics 1.3 [1.3.26918]
```
2. Build locally:
```
mkdir -p build
cd build
source /opt/intel/oneapi/setvars.sh
#for FP16
#cmake .. -DWHISPER_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DWHISPER_SYCL_F16=ON
#for FP32
cmake .. -DWHISPER_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx
#build example/main only
#cmake --build . --config Release --target main
#build all binary
cmake --build . --config Release -v
```
or
```
./examples/sycl/build.sh
```
Note:
- By default, it will build for all binary files. It will take more time. To reduce the time, we recommend to build for **example/main** only.
### Run
1. Put model file to folder **models**
2. Enable oneAPI running environment
```
source /opt/intel/oneapi/setvars.sh
```
3. List device ID
Run without parameter:
```
./build/bin/ls-sycl-device
or
./build/bin/main
```
Check the ID in startup log, like:
```
found 4 SYCL devices:
Device 0: Intel(R) Arc(TM) A770 Graphics, compute capability 1.3,
max compute_units 512, max work group size 1024, max sub group size 32, global mem size 16225243136
Device 1: Intel(R) FPGA Emulation Device, compute capability 1.2,
max compute_units 24, max work group size 67108864, max sub group size 64, global mem size 67065057280
Device 2: 13th Gen Intel(R) Core(TM) i7-13700K, compute capability 3.0,
max compute_units 24, max work group size 8192, max sub group size 64, global mem size 67065057280
Device 3: Intel(R) Arc(TM) A770 Graphics, compute capability 3.0,
max compute_units 512, max work group size 1024, max sub group size 32, global mem size 16225243136
```
|Attribute|Note|
|-|-|
|compute capability 1.3|Level-zero running time, recommended |
|compute capability 3.0|OpenCL running time, slower than level-zero in most cases|
4. Set device ID and execute whisper.cpp
Set device ID = 0 by **GGML_SYCL_DEVICE=0**
```
GGML_SYCL_DEVICE=0 ./build/bin/main -m models/ggml-base.en.bin -f samples/jfk.wav
```
or run by script:
```
./examples/sycl/run_whisper.sh
```
5. Check the device ID in output
Like:
```
Using device **0** (Intel(R) Arc(TM) A770 Graphics) as main device
```
## Environment Variable
#### Build
|Name|Value|Function|
|-|-|-|
|WHISPER_SYCL|ON (mandatory)|Enable build with SYCL code path. <br>For FP32/FP16, WHISPER_SYCL=ON is mandatory.|
|WHISPER_SYCL_F16|ON (optional)|Enable FP16 build with SYCL code path.For FP32, do not set it.|
|CMAKE_C_COMPILER|icx|Use icx compiler for SYCL code path|
|CMAKE_CXX_COMPILER|icpx|use icpx for SYCL code path|
#### Running
|Name|Value|Function|
|-|-|-|
|GGML_SYCL_DEVICE|0 (default) or 1|Set the device id used. Check the device ids by default running output|
|GGML_SYCL_DEBUG|0 (default) or 1|Enable log function by macro: GGML_SYCL_DEBUG|
## Known Issue
- Error: `error while loading shared libraries: libsycl.so.7: cannot open shared object file: No such file or directory`.
Miss to enable oneAPI running environment.
Install oneAPI base toolkit and enable it by: `source /opt/intel/oneapi/setvars.sh`.
- Hang during startup
llama.cpp use mmap as default way to read model file and copy to GPU. In some system, memcpy will be abnormal and block.
Solution: add **--no-mmap**.
## Todo
- Support to build in Windows.
- Support multiple cards.

View File

@ -79,6 +79,9 @@ else()
add_subdirectory(talk)
add_subdirectory(talk-llama)
add_subdirectory(lsp)
if (LLAMA_SYCL)
add_subdirectory(sycl)
endif()
endif()
add_subdirectory(wchess)

View File

@ -0,0 +1,9 @@
# MIT license
# Copyright (C) 2024 Intel Corporation
# SPDX-License-Identifier: MIT
set(TARGET ls-sycl-device)
add_executable(${TARGET} ls-sycl-device.cpp)
install(TARGETS ${TARGET} RUNTIME)
target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
target_compile_features(${TARGET} PRIVATE cxx_std_17)

47
examples/sycl/README.md Normal file
View File

@ -0,0 +1,47 @@
# llama.cpp/example/sycl
This example program provide the tools for llama.cpp for SYCL on Intel GPU.
## Tool
|Tool Name| Function|Status|
|-|-|-|
|ls-sycl-device| List all SYCL devices with ID, compute capability, max work group size, ect.|Support|
### ls-sycl-device
List all SYCL devices with ID, compute capability, max work group size, ect.
1. Build the llama.cpp for SYCL for all targets.
2. Enable oneAPI running environment
```
source /opt/intel/oneapi/setvars.sh
```
3. Execute
```
./build/bin/ls-sycl-device
```
Check the ID in startup log, like:
```
found 4 SYCL devices:
Device 0: Intel(R) Arc(TM) A770 Graphics, compute capability 1.3,
max compute_units 512, max work group size 1024, max sub group size 32, global mem size 16225243136
Device 1: Intel(R) FPGA Emulation Device, compute capability 1.2,
max compute_units 24, max work group size 67108864, max sub group size 64, global mem size 67065057280
Device 2: 13th Gen Intel(R) Core(TM) i7-13700K, compute capability 3.0,
max compute_units 24, max work group size 8192, max sub group size 64, global mem size 67065057280
Device 3: Intel(R) Arc(TM) A770 Graphics, compute capability 3.0,
max compute_units 512, max work group size 1024, max sub group size 32, global mem size 16225243136
```
|Attribute|Note|
|-|-|
|compute capability 1.3|Level-zero running time, recommended |
|compute capability 3.0|OpenCL running time, slower than level-zero in most cases|

19
examples/sycl/build.sh Normal file
View File

@ -0,0 +1,19 @@
# MIT license
# Copyright (C) 2024 Intel Corporation
# SPDX-License-Identifier: MIT
mkdir -p build
cd build
source /opt/intel/oneapi/setvars.sh
#for FP16
#cmake .. -DWHISPER_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DWHISPER_SYCL_F16=ON # faster for long-prompt inference
#for FP32
cmake .. -DWHISPER_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx
#build example/main only
#cmake --build . --config Release --target main
#build all binary
cmake --build . --config Release -v

View File

@ -0,0 +1,11 @@
/*MIT license
Copyright (C) 2024 Intel Corporation
SPDX-License-Identifier: MIT
*/
#include "ggml-sycl.h"
int main(int argc, char ** argv) {
ggml_backend_sycl_print_sycl_devices();
return 0;
}

View File

@ -0,0 +1,17 @@
#!/bin/bash
# MIT license
# Copyright (C) 2024 Intel Corporation
# SPDX-License-Identifier: MIT
INPUT2="Building a website can be done in 10 simple steps:\nStep 1:"
source /opt/intel/oneapi/setvars.sh
if [ $# -gt 0 ]; then
export GGML_SYCL_DEVICE=$1
else
export GGML_SYCL_DEVICE=0
fi
echo GGML_SYCL_DEVICE=$GGML_SYCL_DEVICE
#export GGML_SYCL_DEBUG=1
./build/bin/main -m models/ggml-base.en.bin -f samples/jfk.wav

View File

@ -12,6 +12,10 @@
#include "ggml-cuda.h"
#endif
#ifdef GGML_USE_SYCL
#include "ggml-sycl.h"
#endif
#ifdef WHISPER_USE_OPENVINO
#include "openvino/whisper-openvino-encoder.h"
#endif
@ -1052,6 +1056,16 @@ static ggml_backend_t whisper_backend_init(const whisper_context_params & params
}
#endif
#ifdef GGML_USE_SYCL
if (params.use_gpu) {
WHISPER_LOG_INFO("%s: using SYCL backend\n", __func__);
backend_gpu = ggml_backend_sycl_init(params.gpu_device);
if (!backend_gpu) {
WHISPER_LOG_ERROR("%s: ggml_backend_sycl_init() failed\n", __func__);
}
}
#endif
if (backend_gpu) {
return backend_gpu;
}