qwp4w3hyb's picture
Create README.md
c3c5d8b verified
|
raw
history blame
1.74 kB
metadata
license: other
license_name: llama-3
license_link: https://llama.meta.com/llama3/license/
pipeline_tag: text-generation
base_model: Replete-AI/Llama-3-11.5B-Instruct-V2
tags:
  - facebook
  - meta
  - pytorch
  - llama
  - llama-3
  - instruct
  - finetune
  - frankenmerge
  - merge
  - gguf
  - imatrix
  - importance matrix
model-index:
  - name: Yi-1.5-34B-Chat-16K-iMat-GGUF
    results: []

Quant Infos

  • quants done with an importance matrix for improved quantization loss
  • ggufs & imatrix generated from bf16 for "optimal" accuracy loss
  • Wide coverage of different gguf quant types from Q_8_0 down to IQ1_S
  • Quantized with llama.cpp commit fabf30b4c4fca32e116009527180c252919ca922 (master as of 2024-05-20)
  • Imatrix generated with this multi-purpose dataset.
    ./imatrix -c 512 -m $model_name-f16.gguf -f $llama_cpp_path/groups_merged.txt -o $out_path/imat-f16-gmerged.dat
    

Original Model Card:

Llama-3-11.5B-v2

Thank you to Meta for the weights for Meta-Llama-3-8B

image/png

This is an upscaling of the Meta-Llama-3-8B Ai using techniques created for chargoddard/mistral-11b-slimorca. This Ai model has been upscaled from 8b parameters to 11.5b parameters without any continuous pretraining or fine-tuning.

Unlike version 1 this model has no issues at fp16 or any quantizations.

The model that was used to create this one is linked below:

https://huggingface.co/meta-llama/Meta-Llama-3-8B