ipex-llm/python/llm/example/CPU/HF-Transformers-AutoModels/Model/yuan2/README.md

Yuan2

In this directory, you will find examples on how you could apply BigDL-LLM INT4 optimizations on Yuan2 models. For illustration purposes, we utilize the IEITYuan/Yuan2-2B-hf as a reference Yuan2 model.

0. Requirements

To run these examples with BigDL-LLM, we have some recommended requirements for your machine, please refer to here for more information.

In addition, you need to modify some files in Yuan2-2B-hf folder, since Flash attention dependency is for CUDA usage and currently cannot be installed on Intel CPUs. To manually turn it off, please refer to this issue. We also provide two modified files(config.json and yuan_hf_model.py), which can be used to replace the original content in config.json and yuan_hf_model.py.

Example: Predict Tokens using generate() API

In the example generate.py, we show a basic use case for an Yuan2 model to predict the next N tokens using generate() API, with BigDL-LLM INT4 optimizations.

1. Install

We suggest using conda to manage the Python environment. For more information about conda installation, please refer to here.

After installing conda, create a Python environment for BigDL-LLM:

conda create -n llm python=3.9
conda activate llm

pip install --pre --upgrade bigdl-llm[all] # install the latest bigdl-llm nightly build with 'all' option
pip install einops # additional package required for Yuan2 to conduct generation
pip install pandas # additional package required for Yuan2 to conduct generation

2. Run

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 Yuan2 model to be downloaded, or the path to the huggingface checkpoint folder. It is default to be 'IEITYuan/Yuan2-2B-hf'.
• --prompt PROMPT: argument defining the prompt to be infered (with integrated prompt format for chat). It is default to be 'IEITYuan/Yuan2-2B-hf'.
• --n-predict N_PREDICT: argument defining the max number of tokens to predict. It is default to be 100.

Note

: When loading the model in 4-bit, BigDL-LLM converts linear layers in the model into INT4 format. In theory, a XB model saved in 16-bit will requires approximately 2X GB of memory for loading, and ~0.5X GB memory for further inference.

Please select the appropriate size of the Yuan2 model based on the capabilities of your machine.

2.1 Client

On client Windows machine, it is recommended to run directly with full utilization of all cores:

python ./generate.py

2.2 Server

For optimal performance on server, it is recommended to set several environment variables (refer to here for more information), and run the example with all the physical cores of a single socket.

E.g. on Linux,

# set BigDL-LLM env variables
source bigdl-llm-init

# e.g. for a server with 48 cores per socket
export OMP_NUM_THREADS=48
numactl -C 0-47 -m 0 python ./generate.py

2.3 Sample Output

IEITYuan/Yuan2-2B-hf

Inference time: xxxx seconds
-------------------- Output --------------------
 
What is AI?
AI is what we call "Artificial Intelligence."<eod>