Hugging Face Undi95/Meta-Llama-3.1-8B-Claude Overview
This model is an 8 billion parameter instruction-tuned variant of Meta's Llama 3.1 architecture, developed by Undi95. Its primary differentiator is the extensive fine-tuning on 9 million tokens of high-quality conversational data generated by Anthropic's Claude Opus and Sonnet models.
Key Training Details
- Base Model: Meta-Llama-3.1-8B
- Fine-tuning Data: A curated collection of Claude Opus/Sonnet tokens from datasets including
Norquinal/claude_multiround_chat_30k,kalomaze/Opus_Instruct_3k,mahiatlinux/Claude3-Opus-Instruct-ShareGPT-14k,kalomaze/Opus_Instruct_25k,meseca/opus-instruct-9k,Gryphe/Sonnet3.5-SlimOrcaDedupCleaned, andGryphe/Opus-WritingPrompts. - Training Duration: 2 epochs, completed in 6 hours on 8x H100 NVL GPUs.
Emulated Persona
The model is configured to adopt the persona of Claude, as defined by the provided system prompts for both Claude Opus 20240306 and Claude Sonnet 3 20240306. This includes specific instructions regarding knowledge base currency (last updated August 2023), response conciseness for simple queries, thoroughness for complex questions, assistance with diverse viewpoints, and adherence to ethical guidelines.
Use Cases
This model is particularly well-suited for applications requiring:
- Sophisticated conversational AI: Emulating the detailed and nuanced responses characteristic of Claude models.
- Content generation: Producing high-quality text for writing, analysis, and creative tasks.
- Instruction following: Excelling in tasks where precise and context-aware responses are crucial, leveraging the instruction-tuned nature of the Claude datasets.
- Coding assistance: Utilizing markdown for code formatting as per the Claude system prompt.