From 2ea13d502f9815e3c4da7fb99336ccf4bb664e86 Mon Sep 17 00:00:00 2001 From: Ch1y0q Date: Thu, 26 Sep 2024 13:51:37 +0800 Subject: [PATCH] Add minicpm3 gpu example (#12114) * add minicpm3 gpu example * update GPU example * update --------- Co-authored-by: Huang, Xinshengzi --- README.md | 1 + README.zh-CN.md | 1 + .../GPU/HuggingFace/LLM/minicpm3/README.md | 146 ++++++++++++++++++ .../GPU/HuggingFace/LLM/minicpm3/generate.py | 82 ++++++++++ 4 files changed, 230 insertions(+) create mode 100644 python/llm/example/GPU/HuggingFace/LLM/minicpm3/README.md create mode 100644 python/llm/example/GPU/HuggingFace/LLM/minicpm3/generate.py diff --git a/README.md b/README.md index 504e1e7b2ab..f95f7e3ae18 100644 --- a/README.md +++ b/README.md @@ -322,6 +322,7 @@ Over 50 models have been optimized/verified on `ipex-llm`, including *LLaMA/LLaM | Command-R/cohere | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/cohere) | [link](python/llm/example/GPU/HuggingFace/LLM/cohere) | | CodeGeeX2 | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/codegeex2) | [link](python/llm/example/GPU/HuggingFace/LLM/codegeex2) | | MiniCPM | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/minicpm) | [link](python/llm/example/GPU/HuggingFace/LLM/minicpm) | +| MiniCPM3 | | [link](python/llm/example/GPU/HuggingFace/LLM/minicpm3) | | MiniCPM-V | | [link](python/llm/example/GPU/HuggingFace/Multimodal/MiniCPM-V) | | MiniCPM-V-2 | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/minicpm-v-2) | [link](python/llm/example/GPU/HuggingFace/Multimodal/MiniCPM-V-2) | | MiniCPM-Llama3-V-2_5 | | [link](python/llm/example/GPU/HuggingFace/Multimodal/MiniCPM-Llama3-V-2_5) | diff --git a/README.zh-CN.md b/README.zh-CN.md index 778d968136a..4a0c6ed213e 100644 --- a/README.zh-CN.md +++ b/README.zh-CN.md @@ -321,6 +321,7 @@ See the demo of running [*Text-Generation-WebUI*](https://ipex-llm.readthedocs.i | Command-R/cohere | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/cohere) | [link](python/llm/example/GPU/HuggingFace/LLM/cohere) | | CodeGeeX2 | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/codegeex2) | [link](python/llm/example/GPU/HuggingFace/LLM/codegeex2) | | MiniCPM | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/minicpm) | [link](python/llm/example/GPU/HuggingFace/LLM/minicpm) | +| MiniCPM3 | | [link](python/llm/example/GPU/HuggingFace/LLM/minicpm3) | | MiniCPM-V | | [link](python/llm/example/GPU/HuggingFace/Multimodal/MiniCPM-V) | | MiniCPM-V-2 | | [link](python/llm/example/GPU/HuggingFace/Multimodal/MiniCPM-V-2) | | MiniCPM-Llama3-V-2_5 | | [link](python/llm/example/GPU/HuggingFace/Multimodal/MiniCPM-Llama3-V-2_5) | diff --git a/python/llm/example/GPU/HuggingFace/LLM/minicpm3/README.md b/python/llm/example/GPU/HuggingFace/LLM/minicpm3/README.md new file mode 100644 index 00000000000..0f008dd0649 --- /dev/null +++ b/python/llm/example/GPU/HuggingFace/LLM/minicpm3/README.md @@ -0,0 +1,146 @@ +# MiniCPM3 +In this directory, you will find examples on how you could apply IPEX-LLM INT4 optimizations on MiniCPM3 models on [Intel GPUs](../../../README.md). For illustration purposes, we utilize the [openbmb/MiniCPM3-4B](https://huggingface.co/openbmb/MiniCPM3-4B) as a reference MiniCPM3 model. + +## 0. Requirements +To run these examples with IPEX-LLM on Intel GPUs, we have some recommended requirements for your machine, please refer to [here](../../../README.md#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 MiniCPM3 model to predict the next N tokens using `generate()` API, with IPEX-LLM INT4 optimizations on Intel GPUs. +### 1. Install +#### 1.1 Installation on Linux +We suggest using conda to manage environment: +```bash +conda create -n llm python=3.11 +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/ + +pip install jsonschema datamodel_code_generator +``` + +#### 1.2 Installation on Windows +We suggest using conda to manage environment: +```bash +conda create -n llm python=3.11 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/ + +pip install jsonschema datamodel_code_generator +``` + +### 2. Configures OneAPI environment variables for Linux + +> [!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. + +```bash +source /opt/intel/oneapi/setvars.sh +``` + +### 3. Runtime Configurations +For optimal performance, it is recommended to set several environment variables. Please check out the suggestions based on your device. +#### 3.1 Configurations for Linux +
+ +For Intel Arc™ A-Series Graphics and Intel Data Center GPU Flex Series + +```bash +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 + +```bash +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 by `conda install -c conda-forge -y gperftools=2.10`. +
+ +
+ +For Intel iGPU + +```bash +export SYCL_CACHE_PERSISTENT=1 +export BIGDL_LLM_XMX_DISABLED=1 +``` + +
+ +#### 3.2 Configurations for Windows +
+ +For Intel iGPU + +```cmd +set SYCL_CACHE_PERSISTENT=1 +set BIGDL_LLM_XMX_DISABLED=1 +``` + +
+ +
+ +For Intel Arc™ A-Series Graphics + +```cmd +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. +### 4. Running examples + +``` +python ./generate.py --prompt 'What is AI?' +``` + +Arguments info: +- `--repo-id-or-model-path REPO_ID_OR_MODEL_PATH`: argument defining the huggingface repo id for the MiniCPM3 model (e.g. `openbmb/MiniCPM3-4B`) to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'openbmb/MiniCPM3-4B'`. +- `--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`. + +#### Sample Output +#### [openbmb/MiniCPM3-4B](https://huggingface.co/openbmb/MiniCPM3-4B) +```log +Inference time: xxxx s +-------------------- Prompt -------------------- +<|im_start|>user +AI是什么?<|im_end|> +<|im_start|>assistant + +-------------------- Output -------------------- +<|im_start|> user +AI是什么?<|im_end|> +<|im_start|> assistant +AI，即人工智能（Artificial Intelligence），是指由人类创造的、能够模拟人类智能的相关理论和实践的一门新兴技术。它使计算机 或其他 +``` + +```log +Inference time: xxxx s +-------------------- Prompt -------------------- +<|im_start|>user +What is AI?<|im_end|> +<|im_start|>assistant + +-------------------- Output -------------------- +<|im_start|> user +What is AI?<|im_end|> +<|im_start|> assistant +AI, or Artificial Intelligence, is a field of computer science that emphasizes the creation of intelligent machines capable of performing tasks that typically require human intelligence. These tasks include +``` diff --git a/python/llm/example/GPU/HuggingFace/LLM/minicpm3/generate.py b/python/llm/example/GPU/HuggingFace/LLM/minicpm3/generate.py new file mode 100644 index 00000000000..2a8ebed5b9e --- /dev/null +++ b/python/llm/example/GPU/HuggingFace/LLM/minicpm3/generate.py @@ -0,0 +1,82 @@ +# +# 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 ipex_llm.transformers import AutoModelForCausalLM +from transformers import AutoTokenizer + + +if __name__ == '__main__': + parser = argparse.ArgumentParser(description='Predict Tokens using `generate()` API for MiniCPM3 model') + parser.add_argument('--repo-id-or-model-path', type=str, default="openbmb/MiniCPM3-4B", + help='The huggingface repo id for the MiniCPM3 model 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 + # When running LLMs on Intel iGPUs for Windows users, we recommend setting `cpu_embedding=True` in the from_pretrained function. + # This will allow the memory-intensive embedding layer to utilize the CPU instead of iGPU. + model = AutoModelForCausalLM.from_pretrained(model_path, + load_in_4bit=True, + trust_remote_code=True, + optimize_model=True, + use_cache=True) + + model = model.half().to('xpu') + + # Load tokenizer + tokenizer = AutoTokenizer.from_pretrained(model_path, + trust_remote_code=True) + + # Generate predicted tokens + with torch.inference_mode(): + # here the prompt formatting refers to: https://huggingface.co/openbmb/MiniCPM3-4B#inference-with-transformers + chat = [ + { "role": "user", "content": args.prompt }, + ] + + prompt = tokenizer.apply_chat_template(chat, tokenize=False, add_generation_prompt=True) + input_ids = tokenizer.encode(prompt, return_tensors="pt").to('xpu') + + # ipex_llm model needs a warmup, then inference time can be accurate + output = model.generate(input_ids, + do_sample=False, + max_new_tokens=args.n_predict) + # start inference + st = time.time() + + output = model.generate(input_ids, + do_sample=False, + max_new_tokens=args.n_predict) + torch.xpu.synchronize() + end = time.time() + output_str = tokenizer.decode(output[0], skip_special_tokens=False) + + print(f'Inference time: {end-st} s') + print('-'*20, 'Prompt', '-'*20) + print(prompt) + print('-'*20, 'Output', '-'*20) + print(output_str)