ipex-llm/python/llm/example/NPU/HF-Transformers-AutoModels/LLM/README.md

Run Large Language Model on Intel NPU

In this directory, you will find examples on how you could apply IPEX-LLM INT4 or INT8 optimizations on LLM models on Intel NPUs. For illustration purposes, we utilize the meta-llama/Llama-2-7b-chat-hf as reference Llama2 models. In this directory, you will find examples on how you could apply IPEX-LLM INT4 or INT8 optimizations on LLM models on Intel NPUs. See the table blow for verified models.

Verification Models

Model	Model Link
Llama2	meta-llama/Llama-2-7b-chat-hf
Llama3	meta-llama/Meta-Llama-3-8B-Instruct
Chatglm3	THUDM/chatglm3-6b
Qwen2	Qwen/Qwen2-7B-Instruct
MiniCPM	openbmb/MiniCPM-2B-sft-bf16
Phi-3	microsoft/Phi-3-mini-4k-instruct
Stablelm	stabilityai/stablelm-zephyr-3b

0. Requirements

To run these examples with IPEX-LLM on Intel NPUs, make sure to install the newest driver version of Intel NPU. Go to https://www.intel.com/content/www/us/en/download/794734/intel-npu-driver-windows.html to download and unzip the driver. Then go to Device Manager, find Neural Processors -> Intel(R) AI Boost. Right click and select Update Driver. And then manually select the folder unzipped from the driver.

Example: Predict Tokens using generate() API

In the example generate.py, we show a basic use case for a Llama2 model to predict the next N tokens using generate() API, with IPEX-LLM INT4 optimizations on Intel NPUs.

1. Install

1.1 Installation on Windows

We suggest using conda to manage environment:

conda create -n llm python=3.10 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/

# below command will install intel_npu_acceleration_library
pip install intel-npu-acceleration-library==1.3

pip install transformers==4.40

2. Runtime Configurations

For optimal performance, it is recommended to set several environment variables. Please check out the suggestions based on your device.

2.1 Configurations for Windows

Following envrionment variables are required:

set BIGDL_USE_NPU=1

3. Running examples

python ./generate.py

Arguments info:

• --repo-id-or-model-path REPO_ID_OR_MODEL_PATH: argument defining the huggingface repo id for the Llama2 model (e.g. meta-llama/Llama-2-7b-chat-hf and meta-llama/Llama-2-13b-chat-hf) to be downloaded, or the path to the huggingface checkpoint folder. It is default to be 'meta-llama/Llama-2-7b-chat-hf', and more verified models please see the list in Verification Models.
• --prompt PROMPT: argument defining the prompt to be infered (with integrated prompt format for chat). It is default to be 'Once upon a time, there existed a little girl who liked to have adventures. She wanted to go to places and meet new people, and have fun'.
• --n-predict N_PREDICT: argument defining the max number of tokens to predict. It is default to be 32.
• --load_in_low_bit: argument defining the load_in_low_bit format used. It is default to be sym_int8, sym_int4 can also be used.

Sample Output

meta-llama/Llama-2-7b-chat-hf

Inference time: xxxx s
-------------------- Output --------------------
<s> Once upon a time, there existed a little girl who liked to have adventures. She wanted to go to places and meet new people, and have fun. But her parents were always telling her to stay at home and be careful. They were worried about her safety, and they didn't want her to
--------------------------------------------------------------------------------
done