NobodyExistsOnTheInternet/Yi-34B-GiftedConvo-merged

TEXT GENERATIONConcurrency Cost:2Model Size:34BQuant:FP8Ctx Length:32kPublished:Nov 8, 2023License:mitArchitecture:Transformer0.0K Open Weights Cold

The Yi-34B-GiftedConvo-merged model by NobodyExistsOnTheInternet is a 34 billion parameter language model fine-tuned for conversational AI. It leverages a diverse dataset including long evolved conversations, PRM800k data, and scientific data from CamelAI. This model is specifically optimized for generating high-quality, instruction-following responses in a Vicuna 1.1 chat format, making it suitable for advanced dialogue systems.

Loading preview...

Yi-34B-GiftedConvo-merged: Advanced Conversational AI

This model, developed by NobodyExistsOnTheInternet, is a 34 billion parameter language model specifically fine-tuned for high-quality conversational interactions. It stands out due to its unique training methodology and diverse dataset composition, aiming to produce more nuanced and instruction-following responses.

Key Capabilities & Training

  • Extensive Instruction Tuning: Trained on over 20,000 instruct-generated examples, primarily sourced from GPT-4 or human annotations, ensuring strong adherence to instructions.
  • Diverse Dataset: Incorporates a rich dataset featuring:
    • 1,000 long, evolved conversations based on the LIMA methodology, enhancing its ability to maintain coherent and extended dialogues.
    • A subsection of the correct PRM800k data, contributing to its reasoning and problem-solving capabilities.
    • A subsection of CamelAI's Physics and Chemistry data, providing specialized knowledge in scientific domains.
  • Efficient Fine-tuning: The model was fine-tuned using QLoRA and Axolotl, indicating an efficient training process that maintains performance while optimizing resource usage.
  • Vicuna 1.1 Format: Adheres to the Vicuna 1.1 chat format, making it compatible with established conversational interfaces and ensuring predictable interaction patterns.

Good For

  • Advanced Chatbots: Ideal for applications requiring sophisticated, instruction-following conversational agents.
  • Dialogue Systems: Suitable for building systems that need to engage in long, coherent, and context-aware conversations.
  • Specialized Q&A: Its inclusion of scientific data suggests potential for accurate responses in physics and chemistry-related queries.