From 2ad67a18b10bc5d7670c737f6c09bafd9cc9e576 Mon Sep 17 00:00:00 2001 From: binbin Deng <108676127+plusbang@users.noreply.github.com> Date: Wed, 11 Oct 2023 13:38:15 +0800 Subject: [PATCH] LLM: add mistral examples (#9121) --- .../Model/README.md | 1 + .../Model/mistral/README.md | 73 ++++++++++++++++++ .../Model/mistral/generate.py | 65 ++++++++++++++++ .../CPU/PyTorch-Models/Model/README.md | 15 ++-- .../PyTorch-Models/Model/mistral/README.md | 69 +++++++++++++++++ .../PyTorch-Models/Model/mistral/generate.py | 64 ++++++++++++++++ .../Model/README.md | 29 +++---- .../Model/mistral/README.md | 64 ++++++++++++++++ .../Model/mistral/generate.py | 76 +++++++++++++++++++ .../example/GPU/PyTorch-Models/Model/.keep | 0 .../GPU/PyTorch-Models/Model/README.md | 31 ++++++++ .../PyTorch-Models/Model/mistral/README.md | 64 ++++++++++++++++ .../PyTorch-Models/Model/mistral/generate.py | 74 ++++++++++++++++++ 13 files changed, 604 insertions(+), 21 deletions(-) create mode 100644 python/llm/example/CPU/HF-Transformers-AutoModels/Model/mistral/README.md create mode 100644 python/llm/example/CPU/HF-Transformers-AutoModels/Model/mistral/generate.py create mode 100644 python/llm/example/CPU/PyTorch-Models/Model/mistral/README.md create mode 100644 python/llm/example/CPU/PyTorch-Models/Model/mistral/generate.py create mode 100644 python/llm/example/GPU/HF-Transformers-AutoModels/Model/mistral/README.md create mode 100644 python/llm/example/GPU/HF-Transformers-AutoModels/Model/mistral/generate.py delete mode 100644 python/llm/example/GPU/PyTorch-Models/Model/.keep create mode 100644 python/llm/example/GPU/PyTorch-Models/Model/README.md create mode 100644 python/llm/example/GPU/PyTorch-Models/Model/mistral/README.md create mode 100644 python/llm/example/GPU/PyTorch-Models/Model/mistral/generate.py diff --git a/python/llm/example/CPU/HF-Transformers-AutoModels/Model/README.md b/python/llm/example/CPU/HF-Transformers-AutoModels/Model/README.md index 497e7e6c..3d55bbe0 100644 --- a/python/llm/example/CPU/HF-Transformers-AutoModels/Model/README.md +++ b/python/llm/example/CPU/HF-Transformers-AutoModels/Model/README.md @@ -22,6 +22,7 @@ You can use BigDL-LLM to run any Huggingface Transformer models with INT4 optimi | Whisper | [link](whisper) | | Qwen | [link](qwen) | | Aquila | [link](aquila) | +| Mistral | [link](mistral) | ## Recommended Requirements To run the examples, we recommend using Intel® Xeon® processors (server), or >= 12th Gen Intel® Core™ processor (client). diff --git a/python/llm/example/CPU/HF-Transformers-AutoModels/Model/mistral/README.md b/python/llm/example/CPU/HF-Transformers-AutoModels/Model/mistral/README.md new file mode 100644 index 00000000..793e01af --- /dev/null +++ b/python/llm/example/CPU/HF-Transformers-AutoModels/Model/mistral/README.md @@ -0,0 +1,73 @@ +# Mistral +In this directory, you will find examples on how you could apply BigDL-LLM INT4 optimizations on Mistral models. For illustration purposes, we utilize the [mistralai/Mistral-7B-Instruct-v0.1](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1) and [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) as reference Mistral models. + +## Requirements +To run these examples with BigDL-LLM, we have some recommended requirements for your machine, please refer to [here](../README.md#recommended-requirements) for more information. + +## Example: Predict Tokens using `generate()` API +In the example [generate.py](./generate.py), we show a basic use case for a Mistral 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](https://docs.conda.io/en/latest/miniconda.html#). + +After installing conda, create a Python environment for BigDL-LLM: +```bash +conda create -n llm python=3.9 # recommend to use Python 3.9 +conda activate llm + +pip install --pre --upgrade bigdl-llm[all] # install the latest bigdl-llm nightly build with 'all' option + +# Refer to https://huggingface.co/mistralai/Mistral-7B-v0.1#troubleshooting, please make sure you are using a stable version of Transformers, 4.34.0 or newer. +pip install transformers==4.34.0 +``` + +### 2. Run +After setting up the Python environment, you could run the example by following steps. + +> **Note**: When loading the model in 4-bit, BigDL-LLM converts linear layers in the model into INT4 format. In theory, a *X*B model saved in 16-bit will requires approximately 2*X* GB of memory for loading, and ~0.5*X* GB memory for further inference. +> +> Please select the appropriate size of the Mistral model based on the capabilities of your machine. + +#### 2.1 Client +On client Windows machines, it is recommended to run directly with full utilization of all cores: +```powershell +python ./generate.py --prompt 'What is AI?' +``` +More information about arguments can be found in [Arguments Info](#23-arguments-info) section. The expected output can be found in [Sample Output](#24-sample-output) section. + +#### 2.2 Server +For optimal performance on server, it is recommended to set several environment variables (refer to [here](../README.md#best-known-configuration-on-linux) for more information), and run the example with all the physical cores of a single socket. + +E.g. on Linux, +```bash +# set BigDL-Nano env variables +source bigdl-nano-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 --prompt 'What is AI?' +``` +More information about arguments can be found in [Arguments Info](#23-arguments-info) section. The expected output can be found in [Sample Output](#24-sample-output) section. + +#### 2.3 Arguments Info +In the example, several arguments can be passed to satisfy your requirements: + +- `--repo-id-or-model-path REPO_ID_OR_MODEL_PATH`: argument defining the huggingface repo id for the Mistral model (e.g. `mistralai/Mistral-7B-Instruct-v0.1` and `mistralai/Mistral-7B-v0.1`) to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'mistralai/Mistral-7B-Instruct-v0.1'`. +- `--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`. + +#### 2.3 Sample Output +#### [mistralai/Mistral-7B-Instruct-v0.1](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1) +```log +Inference time: xxxx s +-------------------- Output -------------------- +[INST] What is AI? [/INST] AI stands for Artificial Intelligence. It is a branch of computer science that focuses on the development of intelligent machines that work, react, and even think like humans +``` + +#### [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) +```log +Inference time: xxxx s +-------------------- Output -------------------- +[INST] What is AI? [/INST] + +[INST] Artificial Intelligence (AI) is a branch of computer science that deals with the simulation of intelligent behavior in computers. It is a broad +``` diff --git a/python/llm/example/CPU/HF-Transformers-AutoModels/Model/mistral/generate.py b/python/llm/example/CPU/HF-Transformers-AutoModels/Model/mistral/generate.py new file mode 100644 index 00000000..72e4e269 --- /dev/null +++ b/python/llm/example/CPU/HF-Transformers-AutoModels/Model/mistral/generate.py @@ -0,0 +1,65 @@ +# +# Copyright 2016 The BigDL Authors. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# + +import torch +import time +import argparse + +from bigdl.llm.transformers import AutoModelForCausalLM +from transformers import AutoTokenizer + +# you could tune the prompt based on your own model, +# here the prompt tuning refers to https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1#instruction-format +MISTRAL_PROMPT_FORMAT = """[INST] {prompt} [/INST]""" + +if __name__ == '__main__': + parser = argparse.ArgumentParser(description='Predict Tokens using `generate()` API for Mistral model') + parser.add_argument('--repo-id-or-model-path', type=str, default="mistralai/Mistral-7B-Instruct-v0.1", + help='The huggingface repo id for the Mistral (e.g. `mistralai/Mistral-7B-Instruct-v0.1` and `mistralai/Mistral-7B-v0.1`) to be downloaded' + ', or the path to the huggingface checkpoint folder') + parser.add_argument('--prompt', type=str, default="What is AI?", + help='Prompt to infer') + parser.add_argument('--n-predict', type=int, default=32, + help='Max tokens to predict') + + args = parser.parse_args() + model_path = args.repo_id_or_model_path + + # Load model in 4 bit, + # which convert the relevant layers in the model into INT4 format + model = AutoModelForCausalLM.from_pretrained(model_path, + load_in_4bit=True, + trust_remote_code=True) + + # Load tokenizer + tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True) + + # Generate predicted tokens + with torch.inference_mode(): + prompt = MISTRAL_PROMPT_FORMAT.format(prompt=args.prompt) + input_ids = tokenizer.encode(prompt, return_tensors="pt") + st = time.time() + # if your selected model is capable of utilizing previous key/value attentions + # to enhance decoding speed, but has `"use_cache": false` in its model config, + # it is important to set `use_cache=True` explicitly in the `generate` function + # to obtain optimal performance with BigDL-LLM INT4 optimizations + output = model.generate(input_ids, + max_new_tokens=args.n_predict) + end = time.time() + output_str = tokenizer.decode(output[0], skip_special_tokens=True) + print(f'Inference time: {end-st} s') + print('-'*20, 'Output', '-'*20) + print(output_str) diff --git a/python/llm/example/CPU/PyTorch-Models/Model/README.md b/python/llm/example/CPU/PyTorch-Models/Model/README.md index 004ac550..268534df 100644 --- a/python/llm/example/CPU/PyTorch-Models/Model/README.md +++ b/python/llm/example/CPU/PyTorch-Models/Model/README.md @@ -2,13 +2,14 @@ You can use `optimize_model` API to accelerate general PyTorch models on Intel servers and PCs. This directory contains example scripts to help you quickly get started using BigDL-LLM to run some popular open-source models in the community. Each model has its own dedicated folder, where you can find detailed instructions on how to install and run it. # Verified models -| Model | Example | -|-----------|----------------------------------------------------------| -| LLaMA 2 | [link](llama2) | -| ChatGLM | [link](chatglm) | -| Openai Whisper | [link](openai-whisper) | -| BERT | [link](bert) | -| Bark | [link](bark) | +| Model | Example | +|----------------|----------------------------------------------------------| +| LLaMA 2 | [link](llama2) | +| ChatGLM | [link](chatglm) | +| Openai Whisper | [link](openai-whisper) | +| BERT | [link](bert) | +| Bark | [link](bark) | +| Mistral | [link](mistral) | ## Recommended Requirements To run the examples, we recommend using Intel® Xeon® processors (server), or >= 12th Gen Intel® Core™ processor (client). diff --git a/python/llm/example/CPU/PyTorch-Models/Model/mistral/README.md b/python/llm/example/CPU/PyTorch-Models/Model/mistral/README.md new file mode 100644 index 00000000..9f6ff419 --- /dev/null +++ b/python/llm/example/CPU/PyTorch-Models/Model/mistral/README.md @@ -0,0 +1,69 @@ +# Mistral +In this directory, you will find examples on how you could use BigDL-LLM `optimize_model` API to accelerate Mistral models. For illustration purposes, we utilize the [mistralai/Mistral-7B-Instruct-v0.1](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1) and [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) as reference Mistral models. + +## Requirements +To run these examples with BigDL-LLM, we have some recommended requirements for your machine, please refer to [here](../README.md#recommended-requirements) for more information. + +## Example: Predict Tokens using `generate()` API +In the example [generate.py](./generate.py), we show a basic use case for a Mistral 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](https://docs.conda.io/en/latest/miniconda.html#). + +After installing conda, create a Python environment for BigDL-LLM: +```bash +conda create -n llm python=3.9 # recommend to use Python 3.9 +conda activate llm + +pip install --pre --upgrade bigdl-llm[all] # install the latest bigdl-llm nightly build with 'all' option + +# Refer to https://huggingface.co/mistralai/Mistral-7B-v0.1#troubleshooting, please make sure you are using a stable version of Transformers, 4.34.0 or newer. +pip install transformers==4.34.0 +``` + +### 2. Run +After setting up the Python environment, you could run the example by following steps. + +#### 2.1 Client +On client Windows machines, it is recommended to run directly with full utilization of all cores: +```powershell +python ./generate.py --prompt 'What is AI?' +``` +More information about arguments can be found in [Arguments Info](#23-arguments-info) section. The expected output can be found in [Sample Output](#24-sample-output) section. + +#### 2.2 Server +For optimal performance on server, it is recommended to set several environment variables (refer to [here](../README.md#best-known-configuration-on-linux) for more information), and run the example with all the physical cores of a single socket. + +E.g. on Linux, +```bash +# set BigDL-Nano env variables +source bigdl-nano-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 --prompt 'What is AI?' +``` +More information about arguments can be found in [Arguments Info](#23-arguments-info) section. The expected output can be found in [Sample Output](#24-sample-output) section. + +#### 2.3 Arguments Info +In the example, several arguments can be passed to satisfy your requirements: + +- `--repo-id-or-model-path REPO_ID_OR_MODEL_PATH`: argument defining the huggingface repo id for the Mistral model (e.g. `mistralai/Mistral-7B-Instruct-v0.1` and `mistralai/Mistral-7B-v0.1`) to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'mistralai/Mistral-7B-Instruct-v0.1'`. +- `--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`. + +#### 2.3 Sample Output +#### [mistralai/Mistral-7B-Instruct-v0.1](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1) +```log +Inference time: xxxx s +-------------------- Output -------------------- +[INST] What is AI? [/INST] AI stands for Artificial Intelligence. It is a branch of computer science that focuses on the development of intelligent machines that work, react, and even think like humans +``` + +#### [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) +```log +Inference time: xxxx s +-------------------- Output -------------------- +[INST] What is AI? [/INST] + +[INST] Artificial Intelligence (AI) is a branch of computer science that deals with the simulation of intelligent behavior in computers. It is a broad +``` diff --git a/python/llm/example/CPU/PyTorch-Models/Model/mistral/generate.py b/python/llm/example/CPU/PyTorch-Models/Model/mistral/generate.py new file mode 100644 index 00000000..6fa1522a --- /dev/null +++ b/python/llm/example/CPU/PyTorch-Models/Model/mistral/generate.py @@ -0,0 +1,64 @@ +# +# Copyright 2016 The BigDL Authors. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# + +import torch +import time +import argparse + +from transformers import AutoModelForCausalLM, AutoTokenizer +from bigdl.llm import optimize_model + +# you could tune the prompt based on your own model, +# here the prompt tuning refers to https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1#instruction-format +MISTRAL_PROMPT_FORMAT = """[INST] {prompt} [/INST]""" + +if __name__ == '__main__': + parser = argparse.ArgumentParser(description='Predict Tokens using `generate()` API for Mistral model') + parser.add_argument('--repo-id-or-model-path', type=str, default="mistralai/Mistral-7B-Instruct-v0.1", + help='The huggingface repo id for the Mistral (e.g. `mistralai/Mistral-7B-Instruct-v0.1` and `mistralai/Mistral-7B-v0.1`) to be downloaded' + ', or the path to the huggingface checkpoint folder') + parser.add_argument('--prompt', type=str, default="What is AI?", + help='Prompt to infer') + parser.add_argument('--n-predict', type=int, default=32, + help='Max tokens to predict') + + args = parser.parse_args() + model_path = args.repo_id_or_model_path + + # Load model + model = AutoModelForCausalLM.from_pretrained(model_path, + trust_remote_code=True, + torch_dtype='auto', + low_cpu_mem_usage=True) + + # With only one line to enable BigDL-LLM optimization on model + model = optimize_model(model) + + # Load tokenizer + tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True) + + # Generate predicted tokens + with torch.inference_mode(): + prompt = MISTRAL_PROMPT_FORMAT.format(prompt=args.prompt) + input_ids = tokenizer.encode(prompt, return_tensors="pt") + st = time.time() + output = model.generate(input_ids, + max_new_tokens=args.n_predict) + end = time.time() + output_str = tokenizer.decode(output[0], skip_special_tokens=True) + print(f'Inference time: {end-st} s') + print('-'*20, 'Output', '-'*20) + print(output_str) diff --git a/python/llm/example/GPU/HF-Transformers-AutoModels/Model/README.md b/python/llm/example/GPU/HF-Transformers-AutoModels/Model/README.md index a2164718..48533561 100644 --- a/python/llm/example/GPU/HF-Transformers-AutoModels/Model/README.md +++ b/python/llm/example/GPU/HF-Transformers-AutoModels/Model/README.md @@ -2,20 +2,21 @@ You can use BigDL-LLM to run almost every Huggingface Transformer models with INT4 optimizations on your laptops with Intel GPUs. This directory contains example scripts to help you quickly get started using BigDL-LLM to run some popular open-source models in the community. Each model has its own dedicated folder, where you can find detailed instructions on how to install and run it. ## Verified models -| Model | Example | -|------------|----------------------------------------------------------| -| Baichuan | [link](baichuan) | -| Baichuan2 | [link](baichuan2) | -| ChatGLM2 | [link](chatglm2) | -| Chinese Llama2 | [link](chinese-llama2)| -| Falcon | [link](falcon) | -| GPT-J | [link](gpt-j) | -| InternLM | [link](internlm) | -| LLaMA 2 | [link](llama2) | -| MPT | [link](mpt) | -| Qwen | [link](qwen) | -| StarCoder | [link](starcoder) | -| Whisper | [link](whisper) | +| Model | Example | +|----------------|----------------------------------------------------------| +| Baichuan | [link](baichuan) | +| Baichuan2 | [link](baichuan2) | +| ChatGLM2 | [link](chatglm2) | +| Chinese Llama2 | [link](chinese-llama2) | +| Falcon | [link](falcon) | +| GPT-J | [link](gpt-j) | +| InternLM | [link](internlm) | +| LLaMA 2 | [link](llama2) | +| Mistral | [link](mistral) | +| MPT | [link](mpt) | +| Qwen | [link](qwen) | +| StarCoder | [link](starcoder) | +| Whisper | [link](whisper) | ## Verified Hardware Platforms diff --git a/python/llm/example/GPU/HF-Transformers-AutoModels/Model/mistral/README.md b/python/llm/example/GPU/HF-Transformers-AutoModels/Model/mistral/README.md new file mode 100644 index 00000000..0487bbb0 --- /dev/null +++ b/python/llm/example/GPU/HF-Transformers-AutoModels/Model/mistral/README.md @@ -0,0 +1,64 @@ +# Mistral +In this directory, you will find examples on how you could apply BigDL-LLM INT4 optimizations on Mistral models on [Intel GPUs](../README.md). For illustration purposes, we utilize the [mistralai/Mistral-7B-Instruct-v0.1](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1) and [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) as reference Mistral models. + +## Requirements +To run these examples with BigDL-LLM on Intel GPUs, we have some recommended requirements for your machine, please refer to [here](../README.md#recommended-requirements) for more information. + +## Example: Predict Tokens using `generate()` API +In the example [generate.py](./generate.py), we show a basic use case for a Mistral model to predict the next N tokens using `generate()` API, with BigDL-LLM INT4 optimizations on Intel GPUs. +### 1. Install +We suggest using conda to manage the Python environment. For more information about conda installation, please refer to [here](https://docs.conda.io/en/latest/miniconda.html#). + +After installing conda, create a Python environment for BigDL-LLM: +```bash +conda create -n llm python=3.9 # recommend to use Python 3.9 +conda activate llm + +# below command will install intel_extension_for_pytorch==2.0.110+xpu as default +# you can install specific ipex/torch version for your need +pip install --pre --upgrade bigdl-llm[xpu] -f https://developer.intel.com/ipex-whl-stable-xpu + +# Refer to https://huggingface.co/mistralai/Mistral-7B-v0.1#troubleshooting, please make sure you are using a stable version of Transformers, 4.34.0 or newer. +pip install transformers==4.34.0 +``` + +### 2. Configures OneAPI environment variables +```bash +source /opt/intel/oneapi/setvars.sh +``` + +### 3. Run + +For optimal performance on Arc, it is recommended to set several environment variables. + +```bash +export USE_XETLA=OFF +export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1 +``` + +```bash +python ./generate.py --prompt 'What is AI?' +``` + +In the example, several arguments can be passed to satisfy your requirements: + +- `--repo-id-or-model-path REPO_ID_OR_MODEL_PATH`: argument defining the huggingface repo id for the Mistral model (e.g. `mistralai/Mistral-7B-Instruct-v0.1` and `mistralai/Mistral-7B-v0.1`) to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'mistralai/Mistral-7B-Instruct-v0.1'`. +- `--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`. + +#### 2.3 Sample Output +#### [mistralai/Mistral-7B-Instruct-v0.1](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1) +```log +Inference time: xxxx s +-------------------- Output -------------------- +[INST] What is AI? [/INST] AI stands for Artificial Intelligence. It is a branch of computer science that focuses on the development of intelligent machines that work, react, and even think like humans +``` + +#### [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) +```log +Inference time: xxxx s +-------------------- Output -------------------- +[INST] What is AI? [/INST] + +[INST] Artificial Intelligence (AI) is a branch of computer science that deals with the simulation of intelligent behavior in computers. It is a broad +``` diff --git a/python/llm/example/GPU/HF-Transformers-AutoModels/Model/mistral/generate.py b/python/llm/example/GPU/HF-Transformers-AutoModels/Model/mistral/generate.py new file mode 100644 index 00000000..fcf8cca9 --- /dev/null +++ b/python/llm/example/GPU/HF-Transformers-AutoModels/Model/mistral/generate.py @@ -0,0 +1,76 @@ +# +# Copyright 2016 The BigDL Authors. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# + +import torch +import intel_extension_for_pytorch as ipex +import time +import argparse + +from bigdl.llm.transformers import AutoModelForCausalLM +from transformers import AutoTokenizer + +# you could tune the prompt based on your own model, +# here the prompt tuning refers to https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1#instruction-format +MISTRAL_PROMPT_FORMAT = """[INST] {prompt} [/INST]""" + +if __name__ == '__main__': + parser = argparse.ArgumentParser(description='Predict Tokens using `generate()` API for Mistral model') + parser.add_argument('--repo-id-or-model-path', type=str, default="mistralai/Mistral-7B-Instruct-v0.1", + help='The huggingface repo id for the Mistral (e.g. `mistralai/Mistral-7B-Instruct-v0.1` and `mistralai/Mistral-7B-v0.1`) to be downloaded' + ', or the path to the huggingface checkpoint folder') + parser.add_argument('--prompt', type=str, default="What is AI?", + help='Prompt to infer') + parser.add_argument('--n-predict', type=int, default=32, + help='Max tokens to predict') + + args = parser.parse_args() + model_path = args.repo_id_or_model_path + + # Load model in 4 bit, + # which convert the relevant layers in the model into INT4 format + model = AutoModelForCausalLM.from_pretrained(model_path, + load_in_4bit=True, + optimize_model=True, + trust_remote_code=True, + use_cache=True) + model = model.to('xpu') + + # Load tokenizer + tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True) + + # Generate predicted tokens + with torch.inference_mode(): + prompt = MISTRAL_PROMPT_FORMAT.format(prompt=args.prompt) + input_ids = tokenizer.encode(prompt, return_tensors="pt").to('xpu') + # ipex model needs a warmup, then inference time can be accurate + output = model.generate(input_ids, + max_new_tokens=args.n_predict) + + # start inference + st = time.time() + # if your selected model is capable of utilizing previous key/value attentions + # to enhance decoding speed, but has `"use_cache": false` in its model config, + # it is important to set `use_cache=True` explicitly in the `generate` function + # to obtain optimal performance with BigDL-LLM INT4 optimizations + output = model.generate(input_ids, + max_new_tokens=args.n_predict) + torch.xpu.synchronize() + end = time.time() + output = output.cpu() + output_str = tokenizer.decode(output[0], skip_special_tokens=True) + print(f'Inference time: {end-st} s') + print('-'*20, 'Output', '-'*20) + print(output_str) diff --git a/python/llm/example/GPU/PyTorch-Models/Model/.keep b/python/llm/example/GPU/PyTorch-Models/Model/.keep deleted file mode 100644 index e69de29b..00000000 diff --git a/python/llm/example/GPU/PyTorch-Models/Model/README.md b/python/llm/example/GPU/PyTorch-Models/Model/README.md new file mode 100644 index 00000000..4dd7c311 --- /dev/null +++ b/python/llm/example/GPU/PyTorch-Models/Model/README.md @@ -0,0 +1,31 @@ +# BigDL-LLM INT4 Optimization for Large Language Model on Intel GPUs +You can use `optimize_model` API to accelerate general PyTorch models on Intel GPUs. This directory contains example scripts to help you quickly get started using BigDL-LLM to run some popular open-source models in the community. Each model has its own dedicated folder, where you can find detailed instructions on how to install and run it. + +## Verified models +| Model | Example | +|----------------|----------------------------------------------------------| +| Mistral | [link](mistral) | + +## Verified Hardware Platforms + +- Intel Arc™ A-Series Graphics +- Intel Data Center GPU Flex Series +- Intel Data Center GPU Max Series + +## Recommended Requirements +To apply Intel GPU acceleration, there’re several steps for tools installation and environment preparation. + +Step 1, only Linux system is supported now, Ubuntu 22.04 is prefered. + +Step 2, please refer to our [driver installation](https://dgpu-docs.intel.com/driver/installation.html) for general purpose GPU capabilities. +> **Note**: IPEX 2.0.110+xpu requires Intel GPU Driver version is [Stable 647.21](https://dgpu-docs.intel.com/releases/stable_647_21_20230714.html). + +Step 3, you also need to download and install [Intel® oneAPI Base Toolkit](https://www.intel.com/content/www/us/en/developer/tools/oneapi/base-toolkit-download.html). OneMKL and DPC++ compiler are needed, others are optional. +> **Note**: IPEX 2.0.110+xpu requires Intel® oneAPI Base Toolkit's version >= 2023.2.0. + +## Best Known Configuration on Linux +For better performance, it is recommended to set environment variables on Linux: +```bash +export USE_XETLA=OFF +export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1 +``` diff --git a/python/llm/example/GPU/PyTorch-Models/Model/mistral/README.md b/python/llm/example/GPU/PyTorch-Models/Model/mistral/README.md new file mode 100644 index 00000000..54488863 --- /dev/null +++ b/python/llm/example/GPU/PyTorch-Models/Model/mistral/README.md @@ -0,0 +1,64 @@ +# Mistral +In this directory, you will find examples on how you could use BigDL-LLM `optimize_model` API to accelerate Mistral models. For illustration purposes, we utilize the [mistralai/Mistral-7B-Instruct-v0.1](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1) and [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) as reference Mistral models. + +## Requirements +To run these examples with BigDL-LLM on Intel GPUs, we have some recommended requirements for your machine, please refer to [here](../README.md#recommended-requirements) for more information. + +## Example: Predict Tokens using `generate()` API +In the example [generate.py](./generate.py), we show a basic use case for a Mistral model to predict the next N tokens using `generate()` API, with BigDL-LLM INT4 optimizations on Intel GPUs. +### 1. Install +We suggest using conda to manage the Python environment. For more information about conda installation, please refer to [here](https://docs.conda.io/en/latest/miniconda.html#). + +After installing conda, create a Python environment for BigDL-LLM: +```bash +conda create -n llm python=3.9 # recommend to use Python 3.9 +conda activate llm + +# below command will install intel_extension_for_pytorch==2.0.110+xpu as default +# you can install specific ipex/torch version for your need +pip install --pre --upgrade bigdl-llm[xpu] -f https://developer.intel.com/ipex-whl-stable-xpu + +# Refer to https://huggingface.co/mistralai/Mistral-7B-v0.1#troubleshooting, please make sure you are using a stable version of Transformers, 4.34.0 or newer. +pip install transformers==4.34.0 +``` + +### 2. Configures OneAPI environment variables +```bash +source /opt/intel/oneapi/setvars.sh +``` + +### 3. Run + +For optimal performance on Arc, it is recommended to set several environment variables. + +```bash +export USE_XETLA=OFF +export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1 +``` + +```bash +python ./generate.py --prompt 'What is AI?' +``` + +In the example, several arguments can be passed to satisfy your requirements: + +- `--repo-id-or-model-path REPO_ID_OR_MODEL_PATH`: argument defining the huggingface repo id for the Mistral model (e.g. `mistralai/Mistral-7B-Instruct-v0.1` and `mistralai/Mistral-7B-v0.1`) to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'mistralai/Mistral-7B-Instruct-v0.1'`. +- `--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`. + +#### 2.3 Sample Output +#### [mistralai/Mistral-7B-Instruct-v0.1](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1) +```log +Inference time: xxxx s +-------------------- Output -------------------- +[INST] What is AI? [/INST] AI stands for Artificial Intelligence. It is a branch of computer science that focuses on the development of intelligent machines that work, react, and even think like humans +``` + +#### [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) +```log +Inference time: xxxx s +-------------------- Output -------------------- +[INST] What is AI? [/INST] + +[INST] Artificial Intelligence (AI) is a branch of computer science that deals with the simulation of intelligent behavior in computers. It is a broad +``` diff --git a/python/llm/example/GPU/PyTorch-Models/Model/mistral/generate.py b/python/llm/example/GPU/PyTorch-Models/Model/mistral/generate.py new file mode 100644 index 00000000..6d9a9c1e --- /dev/null +++ b/python/llm/example/GPU/PyTorch-Models/Model/mistral/generate.py @@ -0,0 +1,74 @@ +# +# Copyright 2016 The BigDL Authors. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# + +import torch +import intel_extension_for_pytorch as ipex +import time +import argparse + +from transformers import AutoModelForCausalLM, AutoTokenizer +from bigdl.llm import optimize_model + +# you could tune the prompt based on your own model, +# here the prompt tuning refers to https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1#instruction-format +MISTRAL_PROMPT_FORMAT = """[INST] {prompt} [/INST]""" + +if __name__ == '__main__': + parser = argparse.ArgumentParser(description='Predict Tokens using `generate()` API for Mistral model') + parser.add_argument('--repo-id-or-model-path', type=str, default="mistralai/Mistral-7B-Instruct-v0.1", + help='The huggingface repo id for the Mistral (e.g. `mistralai/Mistral-7B-Instruct-v0.1` and `mistralai/Mistral-7B-v0.1`) to be downloaded' + ', or the path to the huggingface checkpoint folder') + parser.add_argument('--prompt', type=str, default="What is AI?", + help='Prompt to infer') + parser.add_argument('--n-predict', type=int, default=32, + help='Max tokens to predict') + + args = parser.parse_args() + model_path = args.repo_id_or_model_path + + # Load model + model = AutoModelForCausalLM.from_pretrained(model_path, + trust_remote_code=True, + torch_dtype='auto', + low_cpu_mem_usage=True) + + # With only one line to enable BigDL-LLM optimization on model + model = optimize_model(model) + + model = model.to('xpu') + + # Load tokenizer + tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True) + + # Generate predicted tokens + with torch.inference_mode(): + prompt = MISTRAL_PROMPT_FORMAT.format(prompt=args.prompt) + input_ids = tokenizer.encode(prompt, return_tensors="pt").to('xpu') + # ipex model needs a warmup, then inference time can be accurate + output = model.generate(input_ids, + max_new_tokens=args.n_predict) + + # start inference + st = time.time() + output = model.generate(input_ids, + max_new_tokens=args.n_predict) + torch.xpu.synchronize() + end = time.time() + output = output.cpu() + output_str = tokenizer.decode(output[0], skip_special_tokens=True) + print(f'Inference time: {end-st} s') + print('-'*20, 'Output', '-'*20) + print(output_str)