Commit
•
8167bc3
1
Parent(s):
9914062
Update README.md
Browse files
README.md
CHANGED
|
@@ -7,10 +7,19 @@ tags:
|
|
| 7 |
---
|
| 8 |
|
| 9 |
Experimental RP-oriented MoE, the idea was to get a model that would be equal to or better than the Mixtral 8x7B and it's finetunes in RP/ERP tasks.
|
|
|
|
| 10 |
|
| 11 |
### Llama 3 ChaoticSoliloquy-v1.5-4x8B
|
| 12 |
```
|
| 13 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 14 |
```
|
| 15 |
|
| 16 |
|
|
|
|
| 7 |
---
|
| 8 |
|
| 9 |
Experimental RP-oriented MoE, the idea was to get a model that would be equal to or better than the Mixtral 8x7B and it's finetunes in RP/ERP tasks.
|
| 10 |
+
Im not sure but it should be better than the [first version](https://huggingface.co/xxx777xxxASD/ChaoticSoliloquy-4x8B)
|
| 11 |
|
| 12 |
### Llama 3 ChaoticSoliloquy-v1.5-4x8B
|
| 13 |
```
|
| 14 |
+
base_model: NeverSleep_Llama-3-Lumimaid-8B-v0.1
|
| 15 |
+
gate_mode: random
|
| 16 |
+
dtype: bfloat16
|
| 17 |
+
experts_per_token: 2
|
| 18 |
+
experts:
|
| 19 |
+
- source_model: ChaoticNeutrals_Poppy_Porpoise-v0.7-L3-8B
|
| 20 |
+
- source_model: NeverSleep_Llama-3-Lumimaid-8B-v0.1
|
| 21 |
+
- source_model: openlynn_Llama-3-Soliloquy-8B
|
| 22 |
+
- source_model: Sao10K_L3-Solana-8B-v1
|
| 23 |
```
|
| 24 |
|
| 25 |
|