Panchovix's picture
Update README.md
a125036
|
raw
history blame
1.59 kB
---
license: llama2
tags:
- not-for-all-audiences
---
5 bits/bpw quantization of [Venus-103b-v1.1](https://huggingface.co/nsfwthrowitaway69/Venus-103b-v1.1) to be used on exllamav2.
Calibration dataset was a cleaned Pippa dataset (https://huggingface.co/datasets/royallab/PIPPA-cleaned), same as used as on the original model card.
You can use the measurement.json from there to do your own quant sizes
# Original model card
# Venus 103b - version 1.1
![image/png](https://cdn-uploads.huggingface.co/production/uploads/655febd724e0d359c1f21096/BSKlxWQSbh-liU8kGz4fF.png)
## Model Details
- 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).
- The resulting model has 120 layers and 103 billion parameters.
- See mergekit-config.yml for details on the merge method used.
- 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.
- Inspired by [Goliath-120b](https://huggingface.co/alpindale/goliath-120b)
**Warning: This model will produce NSFW content!**
## Results
1. Seems to be more "talkative" than Venus-103b-v1.0 (i.e characters speakmore often in roleplays)
2. Sometimes struggles to pay attention to small details in the scenes
3. Prose seems pretty creative and more logical than Venus-120b-v1.0