ayo/ipex-llm
1
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1

LLM: add bark gpu example (#10091)

Browse source 
* add bark gpu example

* fix

* fix license

* add bark

* add example

* fix

* another way
This commit is contained in:
Jin Qiao 2024-02-07 10:47:11 +08:00 committed by GitHub
parent 8953acd7d6
commit 2f4c754759
2 changed files with 189 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

120
python/llm/example/GPU/PyTorch-Models/Model/bark/README.md Normal file
View file

@ -0,0 +1,120 @@
# Bark
In this directory, you will find examples on how you could use BigDL-LLM `optimize_model` API to accelerate Bark models. For illustration purposes, we utilize the [suno/bark-small](https://huggingface.co/suno/bark-small) as reference Bark models.
## Requirements
To run these examples with BigDL-LLM, we have some recommended requirements for your machine, please refer to [here](../README.md#recommended-requirements) for more information.
## Example: Synthesize speech with the given input text
In the example [synthesize_speech.py](./synthesize_speech.py), we show a basic use case for Bark model to synthesize speech based on the given text, with BigDL-LLM INT4 optimizations.
### 1. Install
#### 1.1 Installation on Linux
We suggest using conda to manage the Python environment. For more information about conda installation, please refer to [here](https://docs.conda.io/en/latest/miniconda.html#).
After installing conda, create a Python environment for BigDL-LLM:
```bash
conda create -n llm python=3.9 # recommend to use Python 3.9
conda activate llm
# below command will install intel_extension_for_pytorch==2.1.10+xpu as default
pip install --pre --upgrade bigdl-llm[xpu] -f https://developer.intel.com/ipex-whl-stable-xpu
pip install scipy
```
#### 1.2 Installation on Windows
We suggest using conda to manage environment:
```bash
conda create -n llm python=3.9 libuv
conda activate llm
# below command will install intel_extension_for_pytorch==2.1.10+xpu as default
pip install --pre --upgrade bigdl-llm[xpu] -f https://developer.intel.com/ipex-whl-stable-xpu
pip install scipy
```
### 2. Configures OneAPI environment variables
#### 2.1 Configurations for Linux
```bash
source /opt/intel/oneapi/setvars.sh
```
#### 2.2 Configurations for Windows
```cmd
call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat"
```
> Note: Please make sure you are using **CMD** (**Anaconda Prompt** if using conda) to run the command as PowerShell is not supported.
### 3. Runtime Configurations
For optimal performance, it is recommended to set several environment variables. Please check out the suggestions based on your device.
#### 3.1 Configurations for Linux
<details>
<summary>For Intel Arc™ A-Series Graphics and Intel Data Center GPU Flex Series</summary>
```bash
export USE_XETLA=OFF
export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1
```
</details>
<details>
<summary>For Intel Data Center GPU Max Series</summary>
```bash
export LD_PRELOAD=${LD_PRELOAD}:${CONDA_PREFIX}/lib/libtcmalloc.so
export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1
export ENABLE_SDP_FUSION=1
```
> Note: Please note that `libtcmalloc.so` can be installed by `conda install -c conda-forge -y gperftools=2.10`.
</details>
#### 3.2 Configurations for Windows
<details>
<summary>For Intel iGPU</summary>
```cmd
set SYCL_CACHE_PERSISTENT=1
set BIGDL_LLM_XMX_DISABLED=1
```
</details>
<details>
<summary>For Intel Arc™ A300-Series or Pro A60</summary>
```cmd
set SYCL_CACHE_PERSISTENT=1
```
</details>
<details>
<summary>For other Intel dGPU Series</summary>
There is no need to set further environment variables.
</details>
> Note: For the first time that each model runs on Intel iGPU/Intel Arc™ A300-Series or Pro A60, it may take several minutes to compile.
### 4. Running examples
```bash
python ./synthesize_speech.py --text 'BigDL-LLM is a library for running large language model on Intel XPU with very low latency.'
```
In the example, several arguments can be passed to satisfy your requirements:
- `--repo-id-or-model-path REPO_ID_OR_MODEL_PATH`: argument defining the huggingface repo id for the Bark model (e.g. `suno/bark-small` and `suno/bark`) to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'suno/bark-small'`.
- `--voice-preset`: argument defining the voice preset of model. It is default to be `'v2/en_speaker_6'`.
- `--text TEXT`: argument defining the text to synthesize speech. It is default to be `"BigDL-LLM is a library for running large language model on Intel XPU with very low latency."`.
#### 4.1 Sample Output
#### [suno/bark-small](https://huggingface.co/suno/bark-small)
Text: BigDL-LLM is a library for running large language model on Intel XPU with very low latency.
[Click here to hear sample output.](https://llm-assets.readthedocs.io/en/latest/_downloads/e92874986553193acbd321d1cfe29739/bark-example-output.wav)

69
python/llm/example/GPU/PyTorch-Models/Model/bark/synthesize_speech.py Normal file
View file

@ -0,0 +1,69 @@
#
# Copyright 2016 The BigDL Authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
import torch
import scipy
import time
import argparse
from bigdl.llm import optimize_model
from transformers import AutoProcessor, BarkModel
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='Synthesize speech with the given input text using Bark model')
parser.add_argument('--repo-id-or-model-path', type=str, default='suno/bark-small',
help='The huggingface repo id for the Bark model to be downloaded'
', or the path to the huggingface checkpoint folder')
parser.add_argument('--voice-preset', type=str, default='v2/en_speaker_6',
help='The voice preset of model')
parser.add_argument('--text', type=str, default="BigDL-LLM is a library for running large language model on Intel XPU with very low latency.",
help='Text to synthesize speech')
args = parser.parse_args()
model_path = args.repo_id_or_model_path
voice_preset = args.voice_preset
text = args.text
# Load model
processor = AutoProcessor.from_pretrained(model_path)
model = BarkModel.from_pretrained(model_path)
# With only one line to enable BigDL-LLM optimization on model
# When running LLMs on Intel iGPUs for Windows users, we recommend setting `cpu_embedding=True` in the optimize_model function.
# This will allow the memory-intensive embedding layer to utilize the CPU instead of iGPU.
model = optimize_model(model)
model = model.to('xpu')
inputs = processor(text, voice_preset=voice_preset).to('xpu')
with torch.inference_mode():
# ipex model needs a warmup, then inference time can be accurate
audio_array = model.generate(**inputs)
st = time.time()
audio_array = model.generate(**inputs)
torch.xpu.synchronize()
end = time.time()
print(f"Inference time: {end-st} s")
audio_array = audio_array.cpu().numpy().squeeze()
from scipy.io.wavfile import write as write_wav
sample_rate = model.generation_config.sample_rate
write_wav("bigdl_llm_bark_out.wav", sample_rate, audio_array)