Commit
•
e1591f2
1
Parent(s):
36796e5
Update README.md
Browse files
README.md
CHANGED
|
@@ -26,10 +26,12 @@ PROUDLY PRESENTS
|
|
| 26 |
|
| 27 |
## Neophanis-8x7B-iMat-GGUF
|
| 28 |
|
|
|
|
|
|
|
| 29 |
Quantized from fp16 with love.
|
| 30 |
* Quantizations made possible using mixtral-8x7b-instruct-v0.1.imatrix file from [this](https://huggingface.co/datasets/ikawrakow/imatrix-from-wiki-train) repo (special thanks to [ikawrakow](https://huggingface.co/ikawrakow) again)
|
| 31 |
-
|
| 32 |
-
For a brief rundown of iMatrix quant performance please see this [PR](https://github.com/ggerganov/llama.cpp/pull/5747)
|
| 33 |
|
| 34 |
<i>All quants are verified working prior to uploading to repo for your safety and convenience. </i>
|
| 35 |
|
|
|
|
| 26 |
|
| 27 |
## Neophanis-8x7B-iMat-GGUF
|
| 28 |
|
| 29 |
+
<b>The Good, The Bad, And The Ugly iMats edition</b>
|
| 30 |
+
|
| 31 |
Quantized from fp16 with love.
|
| 32 |
* Quantizations made possible using mixtral-8x7b-instruct-v0.1.imatrix file from [this](https://huggingface.co/datasets/ikawrakow/imatrix-from-wiki-train) repo (special thanks to [ikawrakow](https://huggingface.co/ikawrakow) again)
|
| 33 |
+
* An analysis was run on mixtral-8x7b.imatrix that showed worse KL-Divergence than mixtral-8x7b-instruct-v0.1, hence the latter was used for the imatrixes instead
|
| 34 |
+
* For a brief rundown of iMatrix quant performance please see this [PR](https://github.com/ggerganov/llama.cpp/pull/5747)
|
| 35 |
|
| 36 |
<i>All quants are verified working prior to uploading to repo for your safety and convenience. </i>
|
| 37 |
|