HuggingFaceH4/mistral-7b-sft-beta

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:8kPublished:Oct 26, 2023License:mitArchitecture:Transformer0.0K Open Weights Warm

HuggingFaceH4/mistral-7b-sft-beta is a 7 billion parameter GPT-like model developed by HuggingFaceH4, fine-tuned from mistralai/Mistral-7B-v0.1. It is primarily English-language and was fine-tuned on the HuggingFaceH4/ultrachat_200k dataset, consisting of synthetic dialogues. This model serves as the Supervised Fine-Tuning (SFT) base used to train Zephyr-7B-β with Direct Preference Optimization, making it suitable for conversational AI applications.

Loading preview...

Model Overview

HuggingFaceH4/mistral-7b-sft-beta is a 7 billion parameter language model, fine-tuned by HuggingFaceH4 from the original mistralai/Mistral-7B-v0.1 base model. This model was specifically trained using Supervised Fine-Tuning (SFT) on the HuggingFaceH4/ultrachat_200k dataset, which comprises a diverse collection of synthetic dialogues generated by ChatGPT. It is primarily designed for English language tasks.

Key Characteristics

  • Base Model: Fine-tuned from Mistral-7B-v0.1, inheriting its architectural strengths.
  • Training Data: Utilizes the UltraChat dataset, focusing on conversational data to enhance dialogue capabilities.
  • Purpose: This SFT model served as the foundational step for training the Zephyr-7B-β model through Direct Preference Optimization (DPO).
  • License: Released under the MIT License, allowing for broad use and distribution.

Intended Uses

This model is well-suited for applications requiring a conversational understanding and generation, particularly as a base for further alignment or instruction-tuning. Its training on synthetic dialogues makes it a strong candidate for chatbot development and interactive AI systems. Developers can leverage its fine-tuned conversational abilities for various natural language processing tasks.

Popular Sampler Settings

Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.

temperature
top_p
top_k
frequency_penalty
presence_penalty
repetition_penalty
min_p