|
--- |
|
base_model: Undi95/Meta-Llama-3.1-8B-Claude |
|
library_name: transformers |
|
quantized_by: InferenceIllusionist |
|
tags: |
|
- iMat |
|
- gguf |
|
- llama3 |
|
--- |
|
<img src="https://i.imgur.com/P68dXux.png" width="400"/> |
|
|
|
# Meta-Llama-3.1-8B-Claude-iMat-GGUF |
|
>[!TIP] |
|
><h2>7/28 Update:</h2> |
|
> |
|
>* Reconverted using llama.cpp [b3479](https://github.com/ggerganov/llama.cpp/releases?page=1), adds llama 3.1 rope scaling factors to llama conversion and inference, improving results for context windows above 8192 |
|
>* Importance matrix re-calculated with updated fp16 gguf |
|
>* If using Kobold.cpp make sure you are on [v1.71.1](https://github.com/LostRuins/koboldcpp/releases/tag/v1.71.1) or later to take advantage of rope scaling |
|
|
|
|
|
Quantized from Meta-Llama-3.1-8B-Claude fp16 |
|
* Weighted quantizations were creating using fp16 GGUF and groups_merged.txt in 88 chunks and n_ctx=512 |
|
* Static fp16 will also be included in repo |
|
* For a brief rundown of iMatrix quant performance please see this [PR](https://github.com/ggerganov/llama.cpp/pull/5747) |
|
* <i>All quants are verified working prior to uploading to repo for your safety and convenience</i> |
|
|
|
<b>KL-Divergence Reference Chart</b> |
|
(Click on image to view in full size) |
|
[<img src="https://i.imgur.com/mV0nYdA.png" width="920"/>](https://i.imgur.com/mV0nYdA.png) |
|
|
|
|
|
Original model card can be found [here](https://huggingface.co/Undi95/Meta-Llama-3.1-8B-Claude) |
