In this directory, you will find examples on how you could apply IPEX-LLM INT4 optimizations on CodeGeeX2 models which is implemented based on the ChatGLM2 architecture trained on more code data on Intel GPUs. For illustration purposes, we utilize the THUDM/codegeex-6b as a reference CodeGeeX2 model.
To run these examples with IPEX-LLM on Intel GPUs, we have some recommended requirements for your machine, please refer to here for more information.
In the example generate.py, we show a basic use case for a CodeGeeX2 model to predict the next N tokens using generate()
API, with IPEX-LLM INT4 optimizations on Intel GPUs.
We suggest using conda to manage environment:
conda create -n llm python=3.11
conda activate llm
# below command will install intel_extension_for_pytorch==2.1.10+xpu as default
pip install --pre --upgrade ipex-llm[xpu] --extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/us/
We suggest using conda to manage environment:
conda create -n llm python=3.11 libuv
conda activate llm
# below command will install intel_extension_for_pytorch==2.1.10+xpu as default
pip install --pre --upgrade ipex-llm[xpu] --extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/us/
If you select the codegeex2-6b model (THUDM/codegeex-6b), please note that their code (tokenization_chatglm.py
) initialized tokenizer after the call of __init__
of its parent class, which may result in error during loading tokenizer. To address issue, we have provided an updated file (tokenization_chatglm.py)
def __init__(self, vocab_file, padding_side="left", clean_up_tokenization_spaces=False, **kwargs):
self.tokenizer = SPTokenizer(vocab_file)
super().__init__(padding_side=padding_side, clean_up_tokenization_spaces=clean_up_tokenization_spaces, **kwargs)
You could download the model from THUDM/codegeex-6b, and replace the file tokenization_chatglm.py
with tokenization_chatglm.py.
Note
Skip this step if you are running on Windows.
This is a required step on Linux for APT or offline installed oneAPI. Skip this step for PIP-installed oneAPI.
source /opt/intel/oneapi/setvars.sh
For optimal performance, it is recommended to set several environment variables. Please check out the suggestions based on your device.
For Intel Arc™ A-Series Graphics and Intel Data Center GPU Flex Series
export USE_XETLA=OFF
export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1
export SYCL_CACHE_PERSISTENT=1
For Intel Data Center GPU Max Series
export LD_PRELOAD=${LD_PRELOAD}:${CONDA_PREFIX}/lib/libtcmalloc.so
export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1
export SYCL_CACHE_PERSISTENT=1
export ENABLE_SDP_FUSION=1
Note: Please note that
libtcmalloc.so
can be installed byconda install -c conda-forge -y gperftools=2.10
.
For Intel iGPU
export SYCL_CACHE_PERSISTENT=1
For Intel iGPU and Intel Arc™ A-Series Graphics
set SYCL_CACHE_PERSISTENT=1
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.
python ./generate.py --repo-id-or-model-path REPO_ID_OR_MODEL_PATH --prompt PROMPT --n-predict N_PREDICT
Arguments info:
--repo-id-or-model-path REPO_ID_OR_MODEL_PATH
: argument defining the huggingface repo id for the CodeGeeX2 model to be downloaded, or the path to the huggingface checkpoint folder. It is default to be'THUDM/codegeex-6b'
.--prompt PROMPT
: argument defining the prompt to be infered (with integrated prompt format for chat). It is default to be'# language: Python\n# write a bubble sort function\n'
.--n-predict N_PREDICT
: argument defining the max number of tokens to predict. It is default to be128
.
Inference time: xxxx s
-------------------- Prompt --------------------
# language: Python
# write a bubble sort function
-------------------- Output --------------------
# language: Python
# write a bubble sort function
def bubble_sort(lst):
for i in range(len(lst) - 1):
for j in range(len(lst) - 1 - i):
if lst[j] > lst[j + 1]:
lst[j], lst[j + 1] = lst[j + 1], lst[j]
return lst
print(bubble_sort([5, 2, 3, 4, 1]))