Juniplayground/Mist_LLaMA-2-7B-1024_V3

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kPublished:Aug 18, 2023Architecture:Transformer Cold

Juniplayground/Mist_LLaMA-2-7B-1024_V3 is a 7 billion parameter pretrained generative text model developed by Meta, based on the Llama 2 architecture. This model is a Hugging Face Transformers conversion of the Llama 2 7B base model, designed for general natural language generation tasks. It features a 4096-token context length and was trained on 2 trillion tokens of publicly available online data. The Llama 2 family of models generally outperforms other open-source chat models on various benchmarks.

Loading preview...

Model Overview

Juniplayground/Mist_LLaMA-2-7B-1024_V3 is a 7 billion parameter pretrained model from Meta's Llama 2 family, converted for the Hugging Face Transformers format. Llama 2 models are auto-regressive language models utilizing an optimized transformer architecture. This specific model has a 4096-token context length and was trained on 2 trillion tokens of publicly available data, with a data cutoff of September 2022.

Key Characteristics

  • Architecture: Llama 2, an optimized transformer architecture.
  • Parameter Size: 7 billion parameters.
  • Context Length: 4096 tokens.
  • Training Data: Pretrained on 2 trillion tokens from publicly available online data.
  • Performance: The Llama 2 7B model shows improved performance over Llama 1 7B across various academic benchmarks, including Code (16.8 vs 14.1), Commonsense Reasoning (63.9 vs 60.8), and MMLU (45.3 vs 35.1).

Intended Use Cases

  • Research and Commercial Use: Primarily intended for use in English-speaking contexts.
  • Natural Language Generation: Suitable for adaptation to a variety of natural language generation tasks, as it is a pretrained base model rather than a fine-tuned chat model.

Limitations

  • Language: Primarily tested and intended for use in English.
  • Safety: As with all LLMs, it may produce inaccurate, biased, or objectionable responses, requiring developers to perform safety testing and tuning for specific applications.