Chatglm3 6b Csc Chinese Lora

🚀 Chinese Spelling Correction LoRA Model

A LoRA model for Chinese spelling correction based on ChatGLM3-6B

shibing624/chatglm3-6b-csc-chinese-lora evaluates test data:

The overall performance of shibing624/chatglm3-6b-csc-chinese-lora on CSC test:

On the CSC test set, the generated results have a high accuracy rate for error correction. Since it is based on the THUDM/chatglm3-6b model, the results can often bring surprises. It can not only correct errors but also has the functions of sentence polishing and rewriting.

🚀 Quick Start

✨ Features

• High accuracy in Chinese spelling correction on the CSC test set.
• Based on the THUDM/chatglm3-6b model, capable of sentence polishing and rewriting.

📦 Installation

Install the necessary packages:

pip install -U pycorrector

💻 Usage Examples

Basic Usage

This project is open - sourced in the pycorrector project: pycorrector. It supports both the native ChatGLM model and the LoRA fine - tuned model. You can call it through the following commands:

from pycorrector import GptCorrector
model = GptCorrector("THUDM/chatglm3-6b", "chatglm", peft_name="shibing624/chatglm3-6b-csc-chinese-lora")
r = model.correct_batch(["少先队员因该为老人让坐。"])
print(r) # ['少先队员应该为老人让座。']

Advanced Usage

Without pycorrector, you can use the model like this:

First, you pass your input through the transformer model, then you get the generated sentence.

Install the package:

pip install transformers 

import os

import torch
from peft import PeftModel
from transformers import AutoTokenizer, AutoModel

os.environ["KMP_DUPLICATE_LIB_OK"] = "TRUE"
tokenizer = AutoTokenizer.from_pretrained("THUDM/chatglm3-6b", trust_remote_code=True)
model = AutoModel.from_pretrained("THUDM/chatglm3-6b", trust_remote_code=True).half().cuda()
model = PeftModel.from_pretrained(model, "shibing624/chatglm3-6b-csc-chinese-lora")

sents = ['对下面文本纠错\n\n少先队员因该为老人让坐。',
         '对下面文本纠错\n\n下个星期，我跟我朋唷打算去法国玩儿。']


def get_prompt(user_query):
    vicuna_prompt = "A chat between a curious user and an artificial intelligence assistant. " \
                    "The assistant gives helpful, detailed, and polite answers to the user's questions. " \
                    "USER: {query} ASSISTANT:"
    return vicuna_prompt.format(query=user_query)


for s in sents:
    q = get_prompt(s)
    input_ids = tokenizer(q).input_ids
    generation_kwargs = dict(max_new_tokens=128, do_sample=True, temperature=0.8)
    outputs = model.generate(input_ids=torch.as_tensor([input_ids]).to('cuda:0'), **generation_kwargs)
    output_tensor = outputs[0][len(input_ids):]
    response = tokenizer.decode(output_tensor, skip_special_tokens=True)
    print(response)

Output:

少先队员应该为老人让座。
下个星期，我跟我朋友打算去法国玩儿。

🔧 Technical Details

Model File Structure

The model files are organized as follows:

chatglm3-6b-csc-chinese-lora
    ├── adapter_config.json
    └── adapter_model.bin

Training Parameters

Training Datasets

The training set includes the following data:

• Chinese Spelling Correction Dataset: https://huggingface.co/datasets/shibing624/CSC
• Chinese Grammar Correction Dataset: https://github.com/shibing624/pycorrector/tree/llm/examples/data/grammar
• General GPT4 Q&A Dataset: https://huggingface.co/datasets/shibing624/sharegpt_gpt4

If you need to train a text correction model, please refer to https://github.com/shibing624/pycorrector

📄 License

This project is licensed under the apache - 2.0 license.

📚 Documentation

Citation

@software{pycorrector,
  author = {Ming Xu},
  title = {pycorrector: Text Error Correction Tool},
  year = {2023},
  url = {https://github.com/shibing624/pycorrector},
}