LLM

Run HuggingFace transformers Models on Intel NPU

In this directory, you will find examples on how to directly run HuggingFace transformers models on Intel NPUs (leveraging Intel NPU Acceleration Library). See the table blow for verified models.

Verified Models

Model	Model Link
Llama2	meta-llama/Llama-2-7b-chat-hf
Llama3	meta-llama/Meta-Llama-3-8B-Instruct
Llama3.2-1B	meta-llama/Llama-3.2-1B-Instruct
Llama3.2-3B	meta-llama/Llama-3.2-3B-Instruct
Chatglm3	THUDM/chatglm3-6b
Chatglm2	THUDM/chatglm2-6b
Qwen2	Qwen/Qwen2-7B-Instruct, Qwen/Qwen2-1.5B-Instruct
Qwen2.5	Qwen/Qwen2.5-7B-Instruct
MiniCPM	openbmb/MiniCPM-2B-sft-bf16
Phi-3	microsoft/Phi-3-mini-4k-instruct
Stablelm	stabilityai/stablelm-zephyr-3b
Baichuan2	baichuan-inc/Baichuan2-7B-Chat
Deepseek	deepseek-ai/deepseek-coder-6.7b-instruct
Mistral	mistralai/Mistral-7B-Instruct-v0.1

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 -> Browse my computer for drivers. And then manually select the unzipped driver folder to install.

1. Install

1.1 Installation on Windows

We suggest using conda to manage environment:

conda create -n llm python=3.10
conda activate llm

:: install ipex-llm with 'npu' option
pip install --pre --upgrade ipex-llm[npu]

:: [optional] for Llama-3.2-1B-Instruct & Llama-3.2-3B-Instruct
pip install transformers==4.45.0 accelerate==0.33.0

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

Note

For optimal performance, we recommend running code in conhost rather than Windows Terminal:

• Search for conhost in the Windows search bar and run as administrator
• Run following command to use conda in conhost. Replace <your conda install location> with your conda install location.

call <your conda install location>\Scripts\activate

Following envrionment variables are required:

set BIGDL_USE_NPU=1

3. Run Models

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.

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) 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 Verified Models.
• --lowbit-path LOWBIT_MODEL_PATH: argument defining the path to save/load lowbit version of the model. If it is an empty string, the original pretrained model specified by REPO_ID_OR_MODEL_PATH will be loaded. If it is an existing path, the lowbit model in LOWBIT_MODEL_PATH will be loaded. If it is a non-existing path, the original pretrained model specified by REPO_ID_OR_MODEL_PATH will be loaded, and the converted lowbit version will be saved into LOWBIT_MODEL_PATH. It is default to be '', i.e. an empty string.
• --prompt PROMPT: argument defining the prompt to be infered. 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

4. Run Optimized Models (Experimental)

The examples below show how to run the optimized HuggingFace model implementations on Intel NPU, including

• Llama2-7B
• Llama3-8B
• Llama3.2-1B
• Llama3.2-3B
• Qwen2-1.5B
• Qwen2.5-7B
• MiniCPM-1B
• MiniCPM-2B
• Baichuan2-7B

Recommended NPU Driver Version for MTL Users

32.0.100.2540

Supported models: Llama2-7B, Llama3-8B, Qwen2-1.5B, MiniCPM-1B, MiniCPM-2B, Baichuan2-7B

Recommended NPU Driver Version for LNL Users

32.0.100.2625

Supported models: Llama2-7B, MiniCPM-1B, Baichuan2-7B

32.0.101.2715

Supported models: Llama3-8B, MiniCPM-2B, Qwen2-1.5B, Qwen2.5-7B

Run

:: to run Llama-2-7b-chat-hf
python llama.py

:: to run Meta-Llama-3-8B-Instruct (LNL driver version: 32.0.101.2715)
python llama.py --repo-id-or-model-path meta-llama/Meta-Llama-3-8B-Instruct

:: to run Llama-3.2-1B-Instruct
python llama.py --repo-id-or-model-path meta-llama/Llama-3.2-1B-Instruct

:: to run Llama-3.2-3B-Instruct
python llama.py --repo-id-or-model-path meta-llama/Llama-3.2-3B-Instruct

:: to run Qwen2-1.5B-Instruct (LNL driver version: 32.0.101.2715)
python qwen.py

:: to run Qwen2.5-7B-Instruct (LNL driver version: 32.0.101.2715)
python qwen.py --repo-id-or-model-path Qwen/Qwen2.5-7B-Instruct

:: to run MiniCPM-1B-sft-bf16
python minicpm.py

:: to run MiniCPM-2B-sft-bf16 (LNL driver version: 32.0.101.2715)
python minicpm.py --repo-id-or-model-path openbmb/MiniCPM-2B-sft-bf16

:: to run Baichuan2-7B-Chat
python baichuan2.py

Arguments info:

• --repo-id-or-model-path REPO_ID_OR_MODEL_PATH: argument defining the huggingface repo id for the Llama2 model (i.e. meta-llama/Llama-2-7b-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'.
• --lowbit-path LOWBIT_MODEL_PATH: argument defining the path to save/load lowbit version of the model. If it is an empty string, the original pretrained model specified by REPO_ID_OR_MODEL_PATH will be loaded. If it is an existing path, the lowbit model in LOWBIT_MODEL_PATH will be loaded. If it is a non-existing path, the original pretrained model specified by REPO_ID_OR_MODEL_PATH will be loaded, and the converted lowbit version will be saved into LOWBIT_MODEL_PATH. It is default to be '', i.e. an empty string.
• --prompt PROMPT: argument defining the prompt to be infered (with integrated prompt format for chat). It is default to be What is AI?.
• --n-predict N_PREDICT: argument defining the max number of tokens to predict. It is default to be 32.
• --max-context-len MAX_CONTEXT_LEN: Defines the maximum sequence length for both input and output tokens. It is default to be 1024.
• --max-prompt-len MAX_PROMPT_LEN: Defines the maximum number of tokens that the input prompt can contain. It is default to be 512.
• --disable-transpose-value-cache: Disable the optimization of transposing value cache.

Troubleshooting

TypeError: can't convert meta device type tensor to numpy. Error

If you encounter TypeError: can't convert meta device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first. error when loading lowbit model, please try re-saving the lowbit model with the example script you are currently using. Please note that lowbit models saved by qwen.py, llama.py, etc. cannot be loaded by generate.py.

Output Problem

If you encounter output problem, please try to disable the optimization of transposing value cache with following command:

:: to run Llama-2-7b-chat-hf
python llama.py --disable-transpose-value-cache

:: to run Meta-Llama-3-8B-Instruct (LNL driver version: 32.0.101.2715)
python llama.py --repo-id-or-model-path meta-llama/Meta-Llama-3-8B-Instruct --disable-transpose-value-cache

:: to run Llama-3.2-1B-Instruct
python llama.py --repo-id-or-model-path meta-llama/Llama-3.2-1B-Instruct --disable-transpose-value-cache

:: to run Llama-3.2-3B-Instruct
python llama.py --repo-id-or-model-path meta-llama/Llama-3.2-3B-Instruct --disable-transpose-value-cache

:: to run Qwen2-1.5B-Instruct (LNL driver version: 32.0.101.2715)
python qwen.py --disable-transpose-value-cache

:: to run Qwen2.5-7B-Instruct LNL driver version: 32.0.101.2715)
python qwen.py --repo-id-or-model-path Qwen/Qwen2.5-7B-Instruct --disable-transpose-value-cache

:: to run MiniCPM-1B-sft-bf16
python minicpm.py --disable-transpose-value-cache

:: to run MiniCPM-2B-sft-bf16 (LNL driver version: 32.0.101.2715)
python minicpm.py --repo-id-or-model-path openbmb/MiniCPM-2B-sft-bf16 --disable-transpose-value-cache

:: to run Baichuan2-7B-Chat
python baichuan2.py --disable-transpose-value-cache

For Qwen2.5-7B, you could also try to enable mixed precision optimization when encountering output problems:

python qwen.py --repo-id-or-model-path Qwen/Qwen2.5-7B-Instruct --mixed-precision

Better Performance with High CPU Utilization

You could enable optimization by setting the environment variable with set IPEX_LLM_CPU_LM_HEAD=1 for better performance. But this will cause high CPU utilization.

Sample Output

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

Inference time: xxxx s
-------------------- Input --------------------
<s><s> [INST] <<SYS>>

<</SYS>>

What is AI? [/INST]
-------------------- Output --------------------
<s><s> [INST] <<SYS>>

<</SYS>>

What is AI? [/INST]  AI (Artificial Intelligence) is a field of computer science and engineering that focuses on the development of intelligent machines that can perform tasks