You can deploy this model and start earning money today!

nepalsatish/
llama-3-finetuned-8b

Name: llama-3-finetuned-8b
Availability: InStock

safetensors

unsloth

Views

No views yet

LLaMA-3 Fine-tuned Model

This repository contains a fine-tuned version of the LLaMA-3 (8B parameters) language model, designed for instruction-based tasks in the German language. The model has been trained using the Alpaca prompting technique to generate contextually relevant responses.

Model Description
Installation
Usage
Training
Evaluation
Contributing
License

Model Description

The llama-3-8b-bnb-4bit model has been fine-tuned on a dataset containing German instructions and inputs. The model employs the PEFT (Parameter Efficient Fine-Tuning) technique using LoRA (Low-Rank Adaptation) to enhance its performance efficiently.

Installation

To use the model, you need to have Python 3.7 or later and the required libraries installed. You can install the necessary packages with the following commands:

bash

pip install -qqq "unsloth[colab-new] @ git+https://github.com/unslothai/unsloth.git"
pip install -qqq --no-deps xformers trl peft accelerate bitsandbytes triton
Make sure to check the compatibility of xformers with your PyTorch version.

Usage
You can use the fine-tuned model for inference as follows:


from transformers import AutoTokenizer
from unsloth import FastLanguageModel

# Load the model and tokenizer
model, tokenizer = FastLanguageModel.from_pretrained("nepalsatish/llama-3-finetuned-8b")

# Prepare input
input_prompt = """
### Instruction: Bitte erläutern Sie
### Input: Earth
### Response:
"""
inputs = tokenizer(input_prompt, return_tensors="pt").to("cuda")

# Generate response
outputs = model.generate(**inputs, max_new_tokens=64, use_cache=True)
response = tokenizer.batch_decode(outputs)

print(response)
Training
The model has been trained using the CustomSFTTrainer class, which captures the training loss. The training data is in JSON format, and the prompts are formatted according to the Alpaca instruction-response structure.

To train the model, use the provided training script. Adjust the hyperparameters in the TrainingArguments section as needed.

Evaluation
Training loss is recorded and stored in training_loss.csv. A plot of the training loss over time is saved as training_loss_plot.png. You can analyze these files to monitor the model's performance during training.

Contributing
Contributions are welcome! Please feel free to open issues, submit pull requests, or suggest improvements to the model or documentation.

License
This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments
Thanks to the Hugging Face team and the community for providing the tools and resources needed for training and deploying language models.

This README will serve as a comprehensive guide for users who want to understand, use, or contribute to your fine-tuned LLaMA-3 model.

nepalsatish/llama-3-finetuned-8b

LLaMA-3 Fine-tuned Model

Table of Contents

Model Description

Installation

nepalsatish/
llama-3-finetuned-8b