The examples in this folder shows how to use LangChain with ipex-llm
on Intel GPU.
Tip
For more information, please refer to the upstream LangChain LLM documentation with IPEX-LLM here, and upstream LangChain embedding model documentation with IPEX-LLM here.
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.
We suggest using conda to manage environment:
conda create -n llm python=3.11
conda activate llm
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
pip install --pre --upgrade ipex-llm[xpu] --extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/us/
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.
Install LangChain dependencies:
pip install -U langchain langchain-community
In the current directory, run the example with command:
python chat.py -m MODEL_PATH -q QUESTION
Additional Parameters for Configuration:
-m MODEL_PATH
: required, path to the model-q QUESTION
: question to ask. Default isWhat is AI?
.
The RAG example (rag.py) shows how to load the input text into vector database, and then use LangChain to build a retrival pipeline.
Install LangChain dependencies:
pip install -U langchain langchain-community langchain-chroma sentence-transformers==3.0.1
In the current directory, run the example with command:
python rag.py -m <path_to_llm_model> -e <path_to_embedding_model> [-q QUESTION] [-i INPUT_PATH]
Additional Parameters for Configuration:
-m LLM_MODEL_PATH
: required, path to the model.-e EMBEDDING_MODEL_PATH
: required, path to the embedding model.-q QUESTION
: question to ask. Default isWhat is IPEX-LLM?
.-i INPUT_PATH
: path to the input doc.
The low_bit example (low_bit.py) showcases how to use use LangChain with low_bit optimized model.LangChain
By save_low_bit
we save the weights of low_bit model into the target folder.
Note
save_low_bit
only saves the weights of the model.
Users could copy the tokenizer model into the target folder or specify tokenizer_id
during initialization.
Install LangChain dependencies:
pip install -U langchain langchain-community
In the current directory, run the example with command:
python low_bit.py -m <path_to_model> -t <path_to_target> [-q <your question>]
Additional Parameters for Configuration:
-m MODEL_PATH
: Required, the path to the model-t TARGET_PATH
: Required, the path to save the low_bit model-q QUESTION
: question to ask. Default isWhat is AI?
.