|
--- |
|
language: |
|
- en |
|
--- |
|
Quantizations for [Sao10K/Stheno-1.8-L2-13B](https://huggingface.co/Sao10K/Stheno-1.8-L2-13B) in the [EXL2 format](https://github.com/turboderp/exllamav2) |
|
|
|
Quant|Mem 4k|Mem 4k 8b |
|
-----|------|--------- |
|
[4k_h8_b8](https://huggingface.co/Beinsezii/Stheno-1.8-L2-13B-EXL2/tree/4k_h8_b8)|17.2GB|15.7GB |
|
[4k_h6_b5](https://huggingface.co/Beinsezii/Stheno-1.8-L2-13B-EXL2/tree/4k_h6_b5)|12.6GB|11.0GB |
|
|
|
Breaking down the names: |
|
- **4k** is calibrated with 4096 length as opposed to the default 2048 |
|
- **h8** is a header depth of 8 bits |
|
- **b8** is a model weight average of 8.0 bits |
|
|
|
All quantizations were calibrated with [wikitext-2](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/train) unless otherwise specified |
|
|
|
MEM estimates are performed with an extremely long chatlog in [oobabooga webui](https://github.com/oobabooga/text-generation-webui) on a 7900 XTX using [nvtop](https://github.com/Syllo/nvtop) to monitor **pytorch usage only**. Systems with lots of extra background processes may use more. Additionally, NVIDIA based systems with [flash attention 2](https://github.com/Dao-AILab/flash-attention) **will use less VRAM** than otherwise estimated. |
|
|
|
The measurement files are provided in the main branch so you can [make your own quants](https://github.com/turboderp/exllamav2/blob/master/doc/convert.md) at other bit depths without going through the 2-3 hours of measuring. |
|
|