lmsys/vicuna-7b-delta-v1.1
The lmsys/vicuna-7b-delta-v1.1 is a 7 billion parameter auto-regressive language model developed by LMSYS, fine-tuned from LLaMA. This delta model is designed as a chat assistant, trained on approximately 70,000 user-shared conversations from ShareGPT. It specializes in conversational AI and is intended for research and hobbyist use in natural language processing and chatbots.
Loading preview...
Vicuna 7B Delta v1.1 Overview
Vicuna 7B Delta v1.1, developed by LMSYS, is a 7 billion parameter auto-regressive language model fine-tuned from the LLaMA architecture. This specific release is a "delta model," meaning it cannot be used directly but must be applied on top of original LLaMA weights to obtain the full Vicuna model. It is primarily designed as a chat assistant, leveraging supervised instruction fine-tuning on a dataset of around 70,000 user-shared conversations collected from ShareGPT.com.
Key Capabilities
- Chat Assistant: Optimized for generating human-like responses in conversational contexts.
- Instruction Following: Benefits from supervised instruction fine-tuning for better adherence to prompts.
- Research Tool: Intended for researchers and hobbyists exploring large language models and chatbot development.
Important Notes
- Delta Weights: This model requires application onto base LLaMA weights; it is not a standalone runnable model.
- Non-commercial License: Licensed for non-commercial use, primarily for research purposes.
- Evaluation: Performance is evaluated using standard benchmarks, human preference, and LLM-as-a-judge methods, with details available in its associated paper and leaderboard.
When to Use This Model
This model is suitable for:
- Researchers and developers interested in building or experimenting with conversational AI.
- Hobbyists looking to explore fine-tuned LLaMA-based chat models.
- Projects requiring a chat assistant foundation, provided the delta weights are correctly applied to LLaMA.