Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,32 @@
|
|
1 |
---
|
2 |
license: llama2
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: llama2
|
3 |
+
tags:
|
4 |
+
- not-for-all-audiences
|
5 |
---
|
6 |
+
5 bits/bpw quantization of [Venus-103b-v1.1](https://huggingface.co/nsfwthrowitaway69/Venus-103b-v1.1) to be used on exllamav2.
|
7 |
+
|
8 |
+
Calibration dataset was a cleaned Pippa dataset (https://huggingface.co/datasets/royallab/PIPPA-cleaned), same as used as on the original model card.
|
9 |
+
|
10 |
+
You can use the measurement.json from there to do your own quant sizes
|
11 |
+
|
12 |
+
# Original model card
|
13 |
+
|
14 |
+
# Venus 103b - version 1.1
|
15 |
+
|
16 |
+
![image/png](https://cdn-uploads.huggingface.co/production/uploads/655febd724e0d359c1f21096/BSKlxWQSbh-liU8kGz4fF.png)
|
17 |
+
|
18 |
+
## Model Details
|
19 |
+
|
20 |
+
- A result of interleaving layers of [Sao10K/Euryale-1.3-L2-70B](https://huggingface.co/Sao10K/Euryale-1.3-L2-70B), [migtissera/SynthIA-70B-v1.5](https://huggingface.co/migtissera/SynthIA-70B-v1.5), and [Xwin-LM/Xwin-LM-70B-V0.1](https://huggingface.co/Xwin-LM/Xwin-LM-70B-V0.1) using [mergekit](https://github.com/cg123/mergekit).
|
21 |
+
- The resulting model has 120 layers and 103 billion parameters.
|
22 |
+
- See mergekit-config.yml for details on the merge method used.
|
23 |
+
- See the `exl2-*` branches for exllama2 quantizations. The 5.65 bpw quant should fit in 80GB VRAM, and the 3.35 bpw quant should fit in 48GB VRAM.
|
24 |
+
- Inspired by [Goliath-120b](https://huggingface.co/alpindale/goliath-120b)
|
25 |
+
|
26 |
+
**Warning: This model will produce NSFW content!**
|
27 |
+
|
28 |
+
## Results
|
29 |
+
|
30 |
+
1. Seems to be more "talkative" than Venus-103b-v1.0 (i.e characters speakmore often in roleplays)
|
31 |
+
2. Sometimes struggles to pay attention to small details in the scenes
|
32 |
+
3. Prose seems pretty creative and more logical than Venus-120b-v1.0
|