Skip to content

Latest commit

 

History

History
463 lines (359 loc) · 21.6 KB

README-en.md

File metadata and controls

463 lines (359 loc) · 21.6 KB
Raw

MiniCPM: Unveiling the Potential of End-side Large Language Models

中文 | English

Technical Blog | Multi-modal Model OmniLMM | CPM-C 100B Model Trial | Join our discord and wechat

MiniCPM is an End-Side LLM developed by ModelBest Inc. and TsinghuaNLP, with only 2.4B parameters excluding embeddings.

  • MiniCPM has very close performance compared with Mistral-7B on open-sourced general benchmarks with better ability on Chinese, Mathmetics and Coding after SFT. The overall performance exceeds Llama2-13B, MPT-30B, Falcon-40B, etc.
  • After DPO, MiniCPM outperforms Llama2-70B-Chat, Vicuna-33B, Mistral-7B-Instruct-v0.1, Zephyr-7B-alpha, etc. on MTBench.
  • MiniCPM-V, based on MiniCPM-2B, achieves the best overall performance among multimodel models of the same scale, surpassing existing multimodal large models built on Phi-2 and achieving performance comparable to or even better than 9.6B Qwen-VL-Chat on some tasks.
  • MiniCPM can be deployed and infer on smartphones, and the speed of streaming output is relatively higher than human verbal speed. MiniCPM-V has also successfully deployed multi-modal models on smartphones.
  • The cost of developing based on MiniCPM is low. Parameter efficient finetuning can be conducted with a single 1080/2080 GPU and full parameter finetuning can be conducted with a 3090/4090 GPU.

We release all model parameters for research and limited commercial use. In future, we will also release all the checkpoint during training and most public training data for research on model mechanism.

  • SFT and DPO version based on MiniCPM-2B and human preference: MiniCPM-2B-SFT/DPO
  • The multi-modal model MiniCPM-V based on MiniCPM-2B, which outperforms models with similar size, i.e., Phi-2
  • The INT4 quantized version MiniCPM-2B-SFT/DPO-Int4 based on MiniCPM-2B-SFT/DPO
  • Mobile phone application based on MLC-LLM and LLMFarm. Both language model and multimodel model can conduct inference on smartphones.

Limitations

  • Due to limitations in model size, the model may experience hallucinatory issues. As DPO model tend to generate longer response, hallucinations are more likely to occur. We will also continue to iterate and improve the MiniCPM model.
  • To ensure the generality of the model for academic research purposes, we have not subject it to any identity-specific training. Meanwhile, as we use ShareGPT open-source corpus as part of the training data, the model may output identity-related information similar to the GPT series models.
  • Due to the limitation of model size, the output of the model is greatly influenced by prompts, which may result in inconsistent results from multiple attempts.
  • Due to limited model capacity, the model's knowledge recall may not be accurate. In the future, we will combine the RAG method to enhance the model's knowledge retention ability.

Quick Links

Downloading

Quick Start

vLLM

  • Install vLLM supporting MiniCPM.
    • MiniCPM adopts the MUP program, which introduces some extra scaling operations to make the training process stable. And the MUP structure is a little different from the structure used by Llama and other LLMs.
    • vLLM 0.2.2 is adapted to MiniCPM in the folder inference. More vLLM versions will be supported in the future.
pip install inference/vllm
  • Transfer Huggingface Transformers repo to vLLM-MiniCPM repo, where <hf_repo_path>, <vllmcpm_repo_path> are local paths.
python inference/convert_hf_to_vllmcpm.py --load <hf_repo_path> --save <vllmcpm_repo_path>
  • Examples
cd inference/vllm/examples/infer_cpm
python inference.py --model_path <vllmcpm_repo_path> --prompt_path prompts/prompt_final.txt

Huggingface

MiniCPM-2B
  • Install transformers>=4.36.0 and accelerate，run the following python code.
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
torch.manual_seed(0)

path = 'openbmb/MiniCPM-2B-dpo-bf16'
tokenizer = AutoTokenizer.from_pretrained(path)
model = AutoModelForCausalLM.from_pretrained(path, torch_dtype=torch.bfloat16, device_map='cuda', trust_remote_code=True)

responds, history = model.chat(tokenizer, "Which city is the capital of China?", temperature=0.8, top_p=0.8)
print(responds)
  • Examples
The capital city of China is Beijing. Beijing is not only the political center of China but also a cultural and economic hub. It is known for its rich history and numerous landmarks, such as the Great Wall, the Forbidden City, and the Temple of Heaven. The city is also home to the National Stadium, also known as the "Bird's Nest," and the National Aquatics Center, or "Water Cube." Beijing is a significant city in China, with a population of over 21 million people.
MiniCPM-V
import torch
from PIL import Image
from transformers import AutoModel, AutoTokenizer

model = AutoModel.from_pretrained('openbmb/MiniCPM-V/', trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained('openbmb/MiniCPM-V', trust_remote_code=True)
model.eval().cuda()

image = Image.open('xx.jpg').convert('RGB')
question = 'What is in the image?'
msgs = [{'role': 'user', 'content': question}]

res, context, _ = model.chat(
    image=image,
    msgs=msgs,
    context=None,
    tokenizer=tokenizer,
    sampling=True,
    temperature=0.7
)
print(res)

Evaluation results

Evaluation Settings

  • Since it is difficult to standardize the evaluation of LLMs and there is no public prompt and test code for a large number of evaluations, we can only try our best to make it suitable for all types of models in terms of specific evaluation methods.
  • Overall, we use a unified prompt input for testing, and adjust the input according to the corresponding template for each model.
  • The evaluation scripts and prompts have been open-sourced in our Github repository, and we welcome more developers to continuously improve our evaluation methods.
    • For the text evaluation part, we use our open source large model capability evaluation framework UltraEval. The following is the open source model reproduction process:
      • install UltraEval 
        git clone https://github.com/OpenBMB/UltraEval.git
cd UltraEval
pip install -e .
      • Download the relevant data and unzip it for processing 
        wget -O RawData.zip "https://cloud.tsinghua.edu.cn/f/71b5232264ae4833a4d0/?dl=1"
unzip RawData.zip
python data_process.py
      • Execute evaluation scripts (templates are provided and can be customized) 
        bash run_eval.sh

Deployment mode

  • Because MiniCPM uses the structure of Mup, which is slightly different from existing models in terms of specific computations, we have based the implementation of our model on the vllm=0.2.2 version.
  • For non-MiniCPM models, we directly sampled the latest version of vllm=0.2.7 for inference.

Evaluation method

  • For the QA task (multiple-choice task), we chose to test in two ways:
    • PPL: The options are used as a continuation of the question generation and the answer selection is based on the PPL of each option;
    • The second is to generate the answer options directly.
  • For different models, the results obtained by these two approaches vary widely. the results on both MiniCPM models are closer, while models such as Mistral-7B-v0.1 perform better on PPL and worse on direct generation.
  • In the specific evaluation, we take the higher score of the two evaluation methods as the final result, so as to ensure the fairness of the comparison (* in the following table indicates the PPL).

Text evaluation

Model Average Score Average Score in English Average Score in Chinese C-Eval CMMLU MMLU HumanEval MBPP GSM8K MATH BBH ARC-E ARC-C HellaSwag
Llama2-7B 35.40 36.21 31.765 32.42 31.11 44.32 12.2 27.17 13.57 1.8 33.23 75.25 42.75 75.62*
Qwen-7B 49.46 47.19 59.655 58.96 60.35 57.65 17.07 42.15 41.24 5.34 37.75 83.42 64.76 75.32*
Deepseek-7B 39.96 39.15 43.635 42.82 44.45 47.82 20.12 41.45 15.85 1.53 33.38 74.58* 42.15* 75.45*
Mistral-7B 48.97 49.96 44.54 46.12 42.96 62.69 27.44 45.2 33.13 5.0 41.06 83.92 70.73 80.43*
Llama2-13B 41.48 42.44 37.19 37.32 37.06 54.71 17.07 32.55 21.15 2.25 37.92 78.87* 58.19 79.23*
MPT-30B 38.17 39.82 30.715 29.34 32.09 46.56 21.95 35.36 10.31 1.56 38.22 78.66* 46.08* 79.72*
Falcon-40B 43.62 44.21 40.93 40.29 41.57 53.53 24.39 36.53 22.44 1.92 36.24 81.94* 57.68 83.26*
MiniCPM-2B 52.33 52.6 51.1 51.13 51.07 53.46 50.00 47.31 53.83 10.24 36.87 85.44 68.00 68.25
Model Average Score Average Score in English Average Score in Chinese C-Eval CMMLU MMLU HumanEval MBPP GSM8K MATH BBH ARC-E ARC-C HellaSwag
TinyLlama-1.1B 25.36 25.55 24.525 25.02 24.03 24.3 6.71 19.91 2.27 0.74 28.78 60.77* 28.15* 58.33*
Qwen-1.8B 34.72 31.87 47.565 49.81 45.32 43.37 7.93 17.8 19.26 2.42 29.07 63.97* 43.69 59.28*
Gemini Nano-3B - - - - - - - 27.2(report) 22.8(report) - 42.4(report) - - -
StableLM-Zephyr-3B 43.46 46.31 30.615 30.34 30.89 45.9 35.37 31.85 52.54 12.49 37.68 73.78 55.38 71.87*
Phi-2-2B 48.84 54.41 23.775 23.37 24.18 52.66 47.56 55.04 57.16 3.5 43.39 86.11 71.25 73.07*
MiniCPM-2B 52.33 52.6 51.1 51.13 51.07 53.46 50.00 47.31 53.83 10.24 36.87 85.44 68.00 68.25
Model Average Score Average Score in English Average Score in Chinese C-Eval CMMLU MMLU HumanEval MBPP GSM8K MATH BBH ARC-E ARC-C HellaSwag
ChatGLM2-6B 37.98 35.17 50.63 52.05 49.21 45.77 10.37 9.38 22.74 5.96 32.6 74.45 56.82 58.48*
Mistral-7B-Instruct-v0.1 44.36 45.89 37.51 38.06 36.96 53.56 29.27 39.34 28.73 3.48 39.52 81.61 63.99 73.47*
Mistral-7B-Instruct-v0.2 50.91 52.83 42.235 42.55 41.92 60.51 36.59 48.95 40.49 4.95 39.81 86.28 73.38 84.55*
Qwen-7B-Chat 44.93 42.05 57.9 58.57 57.23 56.03 15.85 40.52 42.23 8.3 37.34 64.44* 39.25* 74.52*
Yi-6B-Chat 50.46 45.89 70.995 70.88 71.11 62.95 14.02 28.34 36.54 3.88 37.43 84.89 70.39 74.6*
Baichuan2-7B-Chat 44.68 42.74 53.39 53.28 53.5 53 21.34 32.32 25.25 6.32 37.46 79.63 60.15 69.23*
Deepseek-7B-chat 49.34 49.56 48.335 46.95 49.72 51.67 40.85 48.48 48.52 4.26 35.7 76.85 63.05 76.68*
Llama2-7B-Chat 38.16 39.17 33.59 34.54 32.64 47.64 14.02 27.4 21.15 2.08 35.54 74.28 54.78 75.65*
MiniCPM-2B 52.33 52.6 51.1 51.13 51.07 53.46 50.00 47.31 53.83 10.24 36.87 85.44 68.00 68.25

Multimodal evaluation

Model Size MME MMB dev (en) MMB dev (zh) MMMU val CMMMU val
LLaVA-Phi 3B 1335 59.8 - - -
MobileVLM 3B 1289 59.6 - - -
Imp-v1 3B 1434 66.5 - - -
Qwen-VL-Chat 9.6B 1487 60.6 56.7 35.9 30.7
CogVLM 17.4B 1438 63.7 53.8 32.1 -
MiniCPM-V（3B) 3B 1452 67.3 61.9 34.7 32.1

DPO evaluation

Model MT-bench
GPT-4-turbo 9.32
GPT-3.5-turbo 8.39
Mistral-8*7b-Instruct-v0.1 8.30
Claude-2.1 8.18
Zephyr-7B-beta 7.34
MiniCPM-2B 7.25
Vicuna-33B 7.12
Zephyr-7B-alpha 6.88
LLaMA-2-70B-chat 6.86
Mistral-7B-Instruct-v0.1 6.84
MPT-34B-instruct 6.39

Deployment on mobile phones

Tutorial

  • After INT4 quantization, MiniCPM only occupies 2GB of space, meeting the requirements of inference on end devices.
  • We have made different adaptations for different operating systems.
  • Note: The current open-source framework is still improving its support for mobile phones, and not all chips and operating system versions can successfully run MLC-LLM or LLMFarm.
  • Android, Harmony OS
    • Adapt based on open-source framework MLC-LLM.
    • Adapted for text model MiniCPM, and multimodel model MiniCPM-V.
    • Support MiniCPM-2B-SFT-INT4、MiniCPM-2B-DPO-INT4、MiniCPM-V.
    • Compile and Installation Guide
  • iOS
    • Adapt based on open-source framework LLMFarm.
    • Adapted for text model MiniCPM.
    • Support MiniCPM-2B-SFT-INT4、MiniCPM-2B-DPO-INT4.
    • Compile and Installation Guide

Performance

  • We did not conduct in-depth optimization and system testing on the mobile inference model, only verifying the feasibility of MiniCPM using mobile phone chips for inference.
  • There have been no previous attempts to deploy multimodal models on mobile phones. We have verified the feasibility of deploying MiniCPM-V on mobile phones based on MLC-LLM this time, and it can input and output normally. However, there also exist a problem of long image processing time, which needs further optimization :)
  • We welcome more developers to continuously improve the inference performance of LLMs on mobile phones and update the test results below.
Mobile Phones OS Processor Memory（GB） Inference Throughput（token/s）
OPPO Find N3 Android 13 snapdragon 8 Gen2 12 6.5
Samsung S23 Ultra Android 14 snapdragon 8 Gen2 12 6.4
Meizu M182Q Android 11 snapdragon 888Plus 8 3.7
Xiaomi 12 Pro Android 13 snapdragon 8 Gen1 8+3 3.7
Xiaomi Redmi K40 Android 11 snapdragon 870 8 3.5
Oneplus LE 2100 Android 13 snapdragon 870 12 3.5
Oneplus HD1900 Android 11 snapdragon 865 8 3.2
Oneplus HD1900 Android 11 snapdragon 855 8 3.0
Oneplus HD1905 Android 10 snapdragon 855 8 3.0
Oneplus HD1900 Android 11 snapdragon 855 8 3.0
Xiaomi MI 8 Android 9 snapdragon 845 6 2.3
Huawei Nova 11SE Harmony 4.0.0 snapdragon 778 12 1.9
Xiaomi MIX 2 Android 9 snapdragon 835 6 1.3
iPhone 15 Pro iOS 17.2.1 A16 8 18.0
iPhone 15 iOS 17.2.1 A16 6 15.0
iPhone 12 Pro iOS 16.5.1 A14 6 5.8
iPhone 12 iOS 17.2.1 A14 4 5.8
iPhone 11 iOS 16.6 A13 4 4.6

multimodel demo

Demo & API

Web-demo based on Gradio

Using the following command can launch the gradio-based demo.

# generation powered by vllm
python demo/vllm_based_demo.py --model_path <vllmcpm_repo_path>
# generation powered by huggingface
python demo/hf_based_demo.py --model_path <hf_repo_path>

Fine-tuning

  • Parameter-efficient Tuning

  • Full-parameter Tuning

    • Using BMTrain，as well as checkpointing and ZeRO-3 (zero redundancy optimizer)，we can tune all parameters of MiniCPM using one piece of NVIDIA GeForce GTX 3090/4090.
    • This code will be available soon.

Show Cases

Text Generation

内容创作-case1

内容创作-case2

Code Generation

代码生成-case1

Reasoning

数理逻辑-case1

数理逻辑-case2

Translation

文本翻译-case1

Instruction Following

指令跟随-case1

Special characters

指令跟随-case1

指令跟随-case2

LICENSE

Model LICENSE

  • This repository is released under the Apache-2.0 License.
  • The usage of MiniCPM model weights must strictly follow the General Model License (GML).
  • The models and weights of MiniCPM are completely free for academic research.
  • If you intend to utilize the model for commercial purposes, please reach out to [email protected] to obtain the certificate of authorization.

Statement

  • As a language model, MiniCPM generates content by learning from a vast amount of text.
  • However, it does not possess the ability to comprehend or express personal opinions or value judgments.
  • Any content generated by MiniCPM does not represent the viewpoints or positions of the model developers.
  • Therefore, when using content generated by MiniCPM, users should take full responsibility for evaluating and verifying it on their own.

Citation

@inproceedings{minicpm2024,
	title={MiniCPM：Unveiling the Potential of End-side Large Language Models},
	booktitle={OpenBMB Blog},
	year={2024}
}