Divij/llama-3.2-3b-sft-llama-star
Divij/llama-3.2-3b-sft-llama-star is a 3.2 billion parameter language model, fine-tuned from the Meta Llama-3.2-3B-Instruct architecture. This model was trained using Supervised Fine-Tuning (SFT) with TRL, offering a 32768 token context length. It is designed for general text generation tasks, leveraging its instruction-tuned base for conversational and prompt-based applications.
Loading preview...
Overview
Divij/llama-3.2-3b-sft-llama-star is a 3.2 billion parameter language model built upon the Meta Llama-3.2-3B-Instruct architecture. It has been specifically fine-tuned using Supervised Fine-Tuning (SFT) with the TRL library, indicating an optimization for following instructions and generating coherent text based on prompts. This model supports a substantial context length of 32768 tokens, allowing it to process and generate longer sequences of text while maintaining context.
Key Capabilities
- Instruction Following: As an SFT model, it is designed to understand and respond to user instructions effectively.
- Text Generation: Capable of generating human-like text for various applications.
- Extended Context: Benefits from a 32768 token context window, suitable for tasks requiring longer input or output.
Training Details
The model's training involved Supervised Fine-Tuning (SFT) using the TRL library. This method typically involves training on a dataset of instruction-response pairs to enhance the model's ability to follow directions and produce relevant outputs. The specific versions of frameworks used during training include TRL 0.27.0, Transformers 4.57.6, Pytorch 2.10.0, Datasets 4.5.0, and Tokenizers 0.22.2.
Good For
- General-purpose text generation based on prompts.
- Applications requiring a model with a good balance of size and context handling.
- Conversational AI and instruction-based tasks where the base Llama-3.2-3B-Instruct capabilities are enhanced through SFT.