metadata

datasets:
  - Yukang/LongAlpaca-16k-length
library_name: transformers
pipeline_tag: text-generation
base_model: mattshumer/Llama-3-8B-16K

Llama-3-8B-16K-GGUF

This is quantized version of mattshumer/Llama-3-8B-16K created using llama.cpp

Model Description

This is an extended (16K) context version of LLaMA 3 8B (base, not instruct). Trained for five hours on 8x A6000 GPUs, using the Yukang/LongAlpaca-16k-length dataset.

rope_theta was set to 1000000.0. Trained with Axolotl.