Azure99/blossom-v2-llama2-7b

Overview

Azure99/blossom-v2-llama2-7b is a 7 billion parameter conversational language model developed by Azure99. It is built upon the Llama-2-7b pre-trained model and has undergone instruction fine-tuning using a mixed dataset comprising Blossom, Wizard, and Dolphin data.

Key Capabilities

• Conversational AI: Specifically fine-tuned for dialogue generation, supporting both single-turn and multi-turn conversations.
• Instruction Following: Trained on a diverse set of instruction datasets, enhancing its ability to follow user prompts.
• Two-Stage Training: The model was trained in two distinct stages: an initial stage with 120K Wizard and 180K Dolphin single-turn instruction datasets, followed by a second stage using 60K Blossom chat and 2K Blossom math multi-turn dialogue datasets.

Good For

• English Dialogue Systems: Ideal for applications requiring conversational AI in English, given its Llama-2 base and training data.
• General Purpose Chatbots: Suitable for creating chatbots that can engage in extended, coherent multi-turn interactions.

Limitations

• Limited Chinese Knowledge: Due to the Llama-2-7b base model's characteristics, its Chinese language proficiency is noted as deficient. For Chinese-specific applications, the developers recommend using blossom-v2-baichuan-7b.