Commit
•
d7160cb
1
Parent(s):
c4568e5
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,31 @@
|
|
| 1 |
---
|
| 2 |
license: llama2
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: llama2
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
tags:
|
| 6 |
+
- not-for-all-audiences
|
| 7 |
---
|
| 8 |
+
|
| 9 |
+
# Venus 103b - version 1.2
|
| 10 |
+
|
| 11 |
+
|
| 12 |
+
|
| 13 |
+
## Model Details
|
| 14 |
+
|
| 15 |
+
- A result of interleaving layers of [Sao10K/Euryale-1.3-L2-70B](https://huggingface.co/Sao10K/Euryale-1.3-L2-70B) and [GOAT-AI/GOAT-70B-Storytelling](https://huggingface.co/GOAT-AI/GOAT-70B-Storytelling)
|
| 16 |
+
- The resulting model has 120 layers and 103 billion parameters.
|
| 17 |
+
- See mergekit-config.yml for details on the merge method used.
|
| 18 |
+
- See the `exl2-*` branches for exllama2 quantizations. The 5.65 bpw quant should fit in 80GB VRAM, and the 3.35/3.0 bpw quants should fit in 48GB VRAM.
|
| 19 |
+
- Inspired by [Goliath-120b](https://huggingface.co/alpindale/goliath-120b)
|
| 20 |
+
|
| 21 |
+
**Warning: This model will produce NSFW content!**
|
| 22 |
+
|
| 23 |
+
## Results
|
| 24 |
+
|
| 25 |
+
1. In my limited testing, I've found this model to be the most creative of the 103b merges I've made so far.
|
| 26 |
+
2. Seems to tolerate higher temperatures than the previous Venus models
|
| 27 |
+
3. Doesn't seem to suffer from any censorship issues
|
| 28 |
+
4. Does not follow instructions as well as v1.1, but still does a bit better than v1.0
|
| 29 |
+
5. Has some issues with formatting sometimes (i.e not closing asterisks or quotes)
|
| 30 |
+
|
| 31 |
+
Note that these are obviously just my personal observations, everyone will have their own unique experience based on their settings and specific scenarios they're using the model for.
|