mistralai/Mistral-7B-v0.1

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kPublished:Sep 20, 2023License:apache-2.0Architecture:Transformer4.1K Open Weights Warm

Mistral-7B-v0.1 is a 7 billion parameter pretrained generative text model developed by Mistral AI. This transformer model incorporates Grouped-Query Attention, Sliding-Window Attention, and a Byte-fallback BPE tokenizer. It demonstrates performance superior to Llama 2 13B across all tested benchmarks, making it suitable for various general-purpose text generation tasks where efficiency and strong performance are critical.

Loading preview...

Overview

Mistral-7B-v0.1 is a 7 billion parameter pretrained generative text model developed by Mistral AI. It is a transformer-based architecture that introduces several key innovations to enhance performance and efficiency, notably Grouped-Query Attention (GQA) and Sliding-Window Attention (SWA). The model also utilizes a Byte-fallback BPE tokenizer.

Key Capabilities & Performance

  • Strong Performance: Mistral-7B-v0.1 has been shown to outperform larger models like Llama 2 13B on all benchmarks tested by its developers, indicating a highly efficient and capable architecture for its size.
  • Architectural Innovations: The integration of GQA and SWA allows for more efficient processing and improved context handling, contributing to its superior performance.

Considerations for Use

  • Base Model: Mistral-7B-v0.1 is a pretrained base model, meaning it does not include built-in moderation mechanisms. Users should implement their own safety measures when deploying applications based on this model.
  • Software Compatibility: To ensure proper functionality and avoid KeyError or NotImplementedError messages, it is recommended to use a stable version of the Hugging Face Transformers library, specifically 4.34.0 or newer.

For more in-depth technical details, refer to the official paper and the release blog post.

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