135 lines
3.0 KiB
Markdown
135 lines
3.0 KiB
Markdown
|
|
---
|
||
|
|
base_model:
|
||
|
|
- TheDrummer/Magidonia-24B-v4.3
|
||
|
|
- Ateron/Sketch-Cydonia
|
||
|
|
- OddTheGreat/Rotor_24B_V.1
|
||
|
|
- DarkArtsForge/Magistaroth-24B-v1.1
|
||
|
|
- MrRikyz/Rei-Pulse-24B
|
||
|
|
- sophosympatheia/Magistry-24B-v1.0
|
||
|
|
- TheDrummer/Cydonia-24B-v4.3
|
||
|
|
library_name: transformers
|
||
|
|
tags:
|
||
|
|
- mergekit
|
||
|
|
- merge
|
||
|
|
- RP
|
||
|
|
- NSFW
|
||
|
|
- roleplay
|
||
|
|
---
|
||
|
|
<div style="width: 100%; text-align: center; margin-bottom: 20px;">
|
||
|
|
<h1 style="font-size: 2.5em; font-weight: bold; color: #4A90E2;">FusionPulse-24B</h1>
|
||
|
|
<hr style="border: 0; height: 1px; background: linear-gradient(to right, transparent, #4A90E2, transparent); margin: 20px 0;">
|
||
|
|
</div>
|
||
|
|
|
||
|
|
## 🌟 Overview
|
||
|
|
|
||
|
|
**FusionPulse-24B** is a merge built on top of `Magidonia-24B-v4.3`.
|
||
|
|
|
||
|
|
The model uses the **TIES** method
|
||
|
|
|
||
|
|
### 🧩 Models Merged
|
||
|
|
|
||
|
|
This model results from a merge between:
|
||
|
|
|
||
|
|
* **TheDrummer/Magidonia-24B-v4.3** (Base)
|
||
|
|
* **Ateron/Sketch-Cydonia**
|
||
|
|
* **OddTheGreat/Rotor_24B_V.1**
|
||
|
|
* **DarkArtsForge/Magistaroth-24B-v1.1**
|
||
|
|
* **MrRikyz/Rei-Pulse-24B**
|
||
|
|
* **sophosympatheia/Magistry-24B-v1.0**
|
||
|
|
* **TheDrummer/Cydonia-24B-v4.3**
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## 🛠️ Merge Details
|
||
|
|
|
||
|
|
### Method: TIES
|
||
|
|
The merge was performed using `mergekit` with the following parameters:
|
||
|
|
- **Base Model:** TheDrummer/Magidonia-24B-v4.3
|
||
|
|
- **Dtype:** float32
|
||
|
|
- **ODtype** BFloat16
|
||
|
|
- **tokenizer source** base
|
||
|
|
|
||
|
|
### ⚙️ Configuration
|
||
|
|
<details>
|
||
|
|
<summary><b>View Full Mergekit YAML</b></summary>
|
||
|
|
|
||
|
|
```yaml
|
||
|
|
base_model: TheDrummer/Magidonia-24B-v4.3
|
||
|
|
dtype: float32
|
||
|
|
merge_method: ties
|
||
|
|
modules:
|
||
|
|
default:
|
||
|
|
slices:
|
||
|
|
- sources:
|
||
|
|
- layer_range: [0, 40]
|
||
|
|
model: Ateron/Sketch-Cydonia
|
||
|
|
parameters:
|
||
|
|
density: 0.55
|
||
|
|
weight: 0.18
|
||
|
|
- layer_range: [0, 40]
|
||
|
|
model: OddTheGreat/Rotor_24B_V.1
|
||
|
|
parameters:
|
||
|
|
density: 0.65
|
||
|
|
weight: 0.22
|
||
|
|
- layer_range: [0, 40]
|
||
|
|
model: DarkArtsForge/Magistaroth-24B-v1.1
|
||
|
|
parameters:
|
||
|
|
density: 0.7
|
||
|
|
weight: 0.27
|
||
|
|
- layer_range: [0, 40]
|
||
|
|
model: MrRikyz/Rei-Pulse-24B
|
||
|
|
parameters:
|
||
|
|
density: 0.6
|
||
|
|
weight: 0.19
|
||
|
|
- layer_range: [0, 40]
|
||
|
|
model: sophosympatheia/Magistry-24B-v1.0
|
||
|
|
parameters:
|
||
|
|
density: 0.44
|
||
|
|
weight: 0.23
|
||
|
|
- layer_range: [0, 40]
|
||
|
|
model: TheDrummer/Cydonia-24B-v4.3
|
||
|
|
parameters:
|
||
|
|
density: 0.25
|
||
|
|
weight: 0.18
|
||
|
|
- layer_range: [0, 40]
|
||
|
|
model: TheDrummer/Magidonia-24B-v4.3
|
||
|
|
|
||
|
|
base_model_alpha: 0.85
|
||
|
|
|
||
|
|
ties:
|
||
|
|
merge_strategy: sum
|
||
|
|
normalize: true
|
||
|
|
sparsity: 0.17
|
||
|
|
rescale: true
|
||
|
|
|
||
|
|
layer_wise:
|
||
|
|
- filter: "layers.0-8.*"
|
||
|
|
scale: 0.75
|
||
|
|
- filter: "layers.9-20.*"
|
||
|
|
scale: 1.05
|
||
|
|
- filter: "layers.21-31.*"
|
||
|
|
scale: 1.15
|
||
|
|
|
||
|
|
tensor_factors:
|
||
|
|
attention: 1.1
|
||
|
|
mlp: 1.2
|
||
|
|
|
||
|
|
post:
|
||
|
|
normalize: true
|
||
|
|
clamp: 2.5
|
||
|
|
|
||
|
|
out_dtype: bfloat16
|
||
|
|
tokenizer:
|
||
|
|
source: base
|
||
|
|
```
|
||
|
|
</details>
|
||
|
|
|
||
|
|
# ✨ Acknowledgements
|
||
|
|
|
||
|
|
Thanks to the authors of the original models for their incredible work:
|
||
|
|
|
||
|
|
- Ateron for `Sketch-Cydonia`
|
||
|
|
- OddTheGreat for `Rotor_24B_V.1`
|
||
|
|
- DarkArtsForge for `Magistaroth-24B-v1.1`
|
||
|
|
- sophosympatheia for `Magistry-24B-v1.0`
|
||
|
|
- TheDrummer for `Cydonia-24B-v4.3` and `Magidonia-24B-v4.3`
|