IgnacioDM/Llama-2-7b-hf-bf16

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kLicense:llama2Architecture:Transformer Open Weights Cold

IgnacioDM/Llama-2-7b-hf-bf16 is a 7 billion parameter Llama 2 model, converted to the Hugging Face format while retaining its original bf16 precision. This model maintains the training precision of the base Llama 2 architecture, offering performance nearly identical to its fp16 counterparts. It is suitable for applications requiring the specific bf16 precision of the Llama 2 7B model.

Loading preview...

Overview

IgnacioDM/Llama-2-7b-hf-bf16 is a specialized conversion of the Meta Llama 2 7B model. Unlike the standard Hugging Face conversion (meta-llama/Llama-2-7b-hf) which typically converts to fp16 precision, this version explicitly preserves the original bf16 (bfloat16) precision that the model was trained on.

Key Characteristics

  • Architecture: Llama 2 family.
  • Parameter Count: 7 billion parameters.
  • Precision: Retains original bf16 precision, distinguishing it from fp16 conversions.
  • Performance: Benchmarking indicates performance is almost identical to the fp16 version of Llama 2 7B.
  • Context Length: Supports a context length of 4096 tokens.

Use Cases

This model is particularly relevant for developers and researchers who specifically require the bf16 precision of the Llama 2 7B model within the Hugging Face ecosystem. It can be used for a wide range of natural language processing tasks where the Llama 2 7B model is applicable, with the added benefit of maintaining its original training precision.