Commit
•
ba8e2e7
1
Parent(s):
ec495f0
Update README.md
Browse files
README.md
CHANGED
|
@@ -7,7 +7,7 @@ language:
|
|
| 7 |
library_name: transformers
|
| 8 |
pipeline_tag: text-generation
|
| 9 |
---
|
| 10 |
-
NousResearch/Nous-Capybara-34B and migtissera/Tess-M-Creative-v1.0 ties merged with mergekit, then quantized with exllamav2 on 200 rows (400K tokens) on a long Vicuna format chat, a sci fi story and a fantasy story.
|
| 11 |
|
| 12 |
Quantized to 4bpw, enough for **~47K context on a 24GB GPU.**
|
| 13 |
|
|
|
|
| 7 |
library_name: transformers
|
| 8 |
pipeline_tag: text-generation
|
| 9 |
---
|
| 10 |
+
NousResearch/Nous-Capybara-34B and migtissera/Tess-M-Creative-v1.0 ties merged with mergekit, then quantized with exllamav2 on 200 rows (400K tokens) on a long Vicuna format chat, a sci fi story and a fantasy story. This should yield better chat performance than the default wikitext quantization.
|
| 11 |
|
| 12 |
Quantized to 4bpw, enough for **~47K context on a 24GB GPU.**
|
| 13 |
|