Downloads · 30 days
135
9% of all-time downloads
chillies/vistral-legal-chat-q4
vistral-legal-chat-q4 is a machine learning model from chillies. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for transformers. The card lists the license as apache-2.0.
[](https://huggingface.co/username/Vistral-legal-chat)
Downloads · 30 days
135
9% of all-time downloads
All-time downloads
1.4K
Public
Repo size
4.4 GB
Likes
1
Public
Click a slice to open those files.
.gguf4.4 GB · 100%
From the Hugging Face model README
Vistral-legal-chat is a fine-tuned version of Vistral, enhanced using QLoRA on 31,000 question-answer pairs about Vietnamese law. This model serves as an expert legal advisor, capable of providing detailed answers and legal consultations for questions related to Vietnamese law.
To use this model, you will need to install the following dependencies:
pip install transformers
pip install torch # or tensorflow depending on your preference
Here is how you can load and use the model in your code:
from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("username/Vistral-legal-chat")
model = AutoModelForCausalLM.from_pretrained("username/Vistral-legal-chat")
# Example usage
chat_template = """
<<SYS>>
Bạn là một chuyên viên tư vấn pháp luật Việt Nam. Bạn có nhiều năm kinh nghiệm và kiến thức chuyên sâu. Bạn sẽ cung cấp câu trả lời về pháp luật, tư vấn luật pháp cho các câu hỏi của User.
<</SYS>>
## user:
Tạm trú là gì?
## assistant:
"""
inputs = tokenizer(chat_template, return_tensors="pt")
outputs = model.generate(**inputs)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)
Provide example code for performing inference with your model:
# Example inference
user_question = "Tạm trú là gì?"
chat_template = f"""
<<SYS>>
Bạn là một chuyên viên tư vấn pháp luật Việt Nam. Bạn có nhiều năm kinh nghiệm và kiến thức chuyên sâu. Bạn sẽ cung cấp câu trả lời về pháp luật, tư vấn luật pháp cho các câu hỏi của User.
<</SYS>>
## user:
{user_question}
## assistant:
"""
inputs = tokenizer(chat_template, return_tensors="pt")
outputs = model.generate(**inputs)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)
If your model can be trained further, provide instructions for training:
# Example training code
from transformers import Trainer, TrainingArguments
training_args = TrainingArguments(
output_dir="./results",
evaluation_strategy="epoch",
per_device_train_batch_size=8,
per_device_eval_batch_size=8,
num_train_epochs=3,
weight_decay=0.01,
)
trainer = Trainer(
model=model,
args=training_args,
train_dataset=train_dataset,
eval_dataset=eval_dataset,
)
trainer.train()
The model was fine-tuned on a dataset of 31,000 question-answer pairs related to Vietnamese law. This dataset encompasses a wide range of legal topics to ensure comprehensive legal advice.
The model was fine-tuned using the QLoRA technique, optimizing for legal language understanding and response accuracy. Training was conducted on [describe hardware, e.g., GPUs, TPUs] over [number of epochs] epochs with [any relevant hyperparameters].
The model was evaluated using the following metrics:
The performance of Vistral-legal-chat was benchmarked against other legal advice models, demonstrating superior accuracy and comprehensiveness in the Vietnamese legal domain.
While Vistral-legal-chat is highly accurate, it may have limitations in the following areas:
We welcome contributions! Please see our contributing guidelines for more information on how to contribute to this project.
This model is licensed under the MIT License.
We would like to thank the contributors and the creators of the datasets used for training this model.
### Tips for Completing the Template
1. **Replace placeholders** (like `username`, `training data`, `evaluation metrics`) with your actual data.
2. **Include any additional information** specific to your model or training process.
3. **Keep the document updated** as the model evolves or more information becomes available.