The PKU-Alignment/alpaca-8b-reproduced-llama-3 is an 8 billion parameter instruction-following language model developed by the PKU-Alignment Team. It is a reproduced version of the Stanford Alpaca model, fine-tuned from the Llama 3 foundation model. This model specializes in following instructions and generating responses based on user prompts, utilizing a transformer architecture. It is designed for general instruction-following tasks, offering a non-commercial license.
Loading preview...
Model Overview
This model, PKU-Alignment/alpaca-8b-reproduced-llama-3, is an 8 billion parameter instruction-following language model developed by the PKU-Alignment Team. It is a reproduction of the original Stanford Alpaca model, but notably fine-tuned from the Llama 3 foundation model (specifically meta-llama/Meta-Llama-3-8B) rather than Llama 1. The training utilized the PKU-Alignment/safe-rlhf library, which employs DeepSpeed as its training backend.
Key Characteristics
- Foundation Model: Built upon the Llama 3 8B base model.
- Instruction Following: Designed to accurately follow user instructions and generate relevant responses.
- Implementation Differences: Features a different conversation template and training backend (DeepSpeed) compared to the original Stanford Alpaca.
- License: Distributed under a non-commercial license.
Use Cases
This model is suitable for various instruction-following applications, including generating text based on prompts, answering questions, and engaging in conversational tasks where adherence to specific instructions is crucial. Developers can interact with the model using the provided safe_rlhf CLI or through the Hugging Face Transformers library.