Deris-SSS-7B / README.md
Virt-io's picture
Update README.md
5366f4a verified
---
library_name: transformers
tags:
- mergekit
- merge
- mistral
---
<img src="https://huggingface.co/Virt-io/Deris-SSS-7B/resolve/main/README.png">
---
# Tips
SillyTavern presets in [presets folder](https://huggingface.co/Virt-io/Deris-SSS-7B/tree/main/presets). (Thanks to [Lewdiculous](https://huggingface.co/Lewdiculous) for ideas.)
Model has formatting issues when using asterisks. Recommended to use novel like formatting (only use quotes)
System prompt can be improved, help welcomed.
Model seems to take characters too seriously, if you find it too stubborn regenerate reply or edit it. It should comply after.(Seems to be fixed now, appears to have been a bad system prompt. oops)
---
# Models Used
[ChaoticNeutrals/Eris_Floramix_DPO_7B](https://huggingface.co/ChaoticNeutrals/Eris_Floramix_DPO_7B)
[ResplendentAI/Datura_7B](https://huggingface.co/ResplendentAI/Datura_7B)
[FuseAI/FuseChat-7B-VaRM](https://huggingface.co/FuseAI/FuseChat-7B-VaRM)
[Weyaxi/Einstein-v4-7B](https://huggingface.co/Weyaxi/Einstein-v4-7B)
[ogno-monarch-jaskier-merge-7b-OH-PREF-DPO](https://huggingface.co/eren23/ogno-monarch-jaskier-merge-7b-OH-PREF-DPO)
---
# Deris-SSS
This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
## Merge Details
### Merge Method
This model was merged using the SLERP merge method.
### Models Merged
The following models were included in the merge:
* Deris-v2
* SmartyPants-step2
### Configuration
The following YAML configuration was used to produce this model:
### Deris-SSS
Final merge combine the smart models with the unhinged ones
```yaml
slices:
- sources:
- model: ./Mergekit/Deris-v2
layer_range: [0, 32]
- model: ./Mergekit/SmartyPants-step2
layer_range: [0, 32]
merge_method: slerp
base_model: ./Mergekit/Deris-v2
parameters:
t:
- filter: self_attn
value: [0, 0.5, 0.3, 0.7, 1]
- filter: mlp
value: [1, 0.5, 0.7, 0.3, 0]
- value: 0.5
dtype: float16
```
### Deris-v2
Combine a bit of Datura_7B into Eris_Floramix_DPO_7B
Reason Datura is extremely unhinged :) more so than Eris
```
slices:
- sources:
- model: ChaoticNeutrals/Eris_Floramix_DPO_7B
layer_range: [0, 32]
- model: ResplendentAI/Datura_7B
layer_range: [0, 32]
merge_method: slerp
base_model: ChaoticNeutrals/Eris_Floramix_DPO_7B
parameters:
t:
- filter: self_attn
value: [0, 0.20, 0.15, 0.25, 0.35]
- filter: mlp
value: [0.35, 0.20, 0.25, 0.15, 0]
- value: 0.20
dtype: float16
```
### SmartyPants-step1
Combine OMJ into Einstein
Reason Einstein looks interesting and OMJ was a high ranking model
```
slices:
- sources:
- model: Weyaxi/Einstein-v4-7B
layer_range: [0, 32]
- model: eren23/ogno-monarch-jaskier-merge-7b-OH-PREF-DPO
layer_range: [0, 32]
merge_method: slerp
base_model: Weyaxi/Einstein-v4-7B
parameters:
t:
- filter: self_attn
value: [0, 0.45, 0.3, 0.55, 0.65]
- filter: mlp
value: [0.65, 0.45, 0.55, 0.3, 0]
- value: 0.45
dtype: float16
```
### SmartyPants-step2
Combine Smarty pants into FuseChat-VaRM
Reason IDK I just like FuseChat-VaRM
```
slices:
- sources:
- model: FuseAI/FuseChat-7B-VaRM
layer_range: [0, 32]
- model: ./Mergekit/SmartyPants-step1
layer_range: [0, 32]
merge_method: slerp
base_model: FuseAI/FuseChat-7B-VaRM
parameters:
t:
- filter: self_attn
value: [0, 0.45, 0.3, 0.55, 0.65]
- filter: mlp
value: [0.65, 0.45, 0.55, 0.3, 0]
- value: 0.45
dtype: float16
```