TheBloke/Llama-2-13B-fp16

TEXT GENERATIONConcurrency Cost:1Model Size:13BQuant:FP8Ctx Length:4kPublished:Jul 18, 2023Architecture:Transformer0.1K Cold

TheBloke/Llama-2-13B-fp16 is a 13 billion parameter Llama 2 model developed by Meta, converted to fp16 format for Hugging Face Transformers. This pretrained generative text model, with a 4k context length, is suitable for a variety of natural language generation tasks. It is part of the Llama 2 family, which was trained on 2 trillion tokens of publicly available online data.

Loading preview...

Llama 2 13B fp16 by TheBloke

This model is an fp16 conversion of Meta's Llama 2 13B, a 13 billion parameter pretrained generative text model. It was converted using the Hugging Face Transformers library from the original PTH files provided by Meta. The Llama 2 family of models was developed by Meta and trained on 2 trillion tokens of publicly available online data, with a pretraining data cutoff of September 2022.

Key Capabilities

  • Generative Text: Capable of generating human-like text for various natural language tasks.
  • Optimized Transformer Architecture: Utilizes an auto-regressive language model with an optimized transformer architecture.
  • Context Length: Supports a context length of 4k tokens.
  • Commercial and Research Use: Intended for both commercial and research applications in English.

Good For

  • Natural Language Generation: Adapting for a wide range of text generation tasks.
  • Further Conversions: Suitable as a base for further model conversions or fine-tuning.
  • GPU Inference: Designed for efficient inference on GPUs in its fp16 format.