Update README.md
Browse filesDescription and base model link
README.md
CHANGED
|
@@ -3,6 +3,8 @@ inference: false
|
|
| 3 |
license: apache-2.0
|
| 4 |
language:
|
| 5 |
- en
|
|
|
|
|
|
|
| 6 |
---
|
| 7 |
<!-- header start -->
|
| 8 |
<!-- 200823 -->
|
|
@@ -22,7 +24,10 @@ Created by: [upstage](https://huggingface.co/upstage)
|
|
| 22 |
[8bpw h8](https://huggingface.co/cgus/Solar-10.7B-SLERP-exl2/tree/8bpw-h8)
|
| 23 |
|
| 24 |
## Quantization notes
|
| 25 |
-
|
|
|
|
|
|
|
|
|
|
| 26 |
30.01.2024 Replaced old quants with newer ones that were made with exllamav2-0.0.12, calibrated with default exllamav2 dataset.
|
| 27 |
This model uses a lot less memory compared to Llama2-13B models and it *almost* fits 12GB VRAM even at 8bpw.
|
| 28 |
But with 8-bit cache it even uses just a tiny bit less than 12GB VRAM.
|
|
|
|
| 3 |
license: apache-2.0
|
| 4 |
language:
|
| 5 |
- en
|
| 6 |
+
base_model:
|
| 7 |
+
- jan-hq/Solar-10.7B-SLERP
|
| 8 |
---
|
| 9 |
<!-- header start -->
|
| 10 |
<!-- 200823 -->
|
|
|
|
| 24 |
[8bpw h8](https://huggingface.co/cgus/Solar-10.7B-SLERP-exl2/tree/8bpw-h8)
|
| 25 |
|
| 26 |
## Quantization notes
|
| 27 |
+
In my experience, it's one of the best models within 13B range for Slavic languages.
|
| 28 |
+
And overall it has very good language skills. I haven't encountered cases when the model randomly switches language unlike most models I tried.
|
| 29 |
+
It also capable to understand instructions in other languages, a pretty rare feat.
|
| 30 |
+
|
| 31 |
30.01.2024 Replaced old quants with newer ones that were made with exllamav2-0.0.12, calibrated with default exllamav2 dataset.
|
| 32 |
This model uses a lot less memory compared to Llama2-13B models and it *almost* fits 12GB VRAM even at 8bpw.
|
| 33 |
But with 8-bit cache it even uses just a tiny bit less than 12GB VRAM.
|