Update README.md
Browse files
README.md
CHANGED
|
@@ -14,11 +14,17 @@ tags:
|
|
| 14 |
|
| 15 |
> There's no 'I' in 'brain damage'
|
| 16 |
|
| 17 |
-
![]()
|
| 18 |
|
| 19 |
### Overview
|
| 20 |
|
| 21 |
-
I'
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
|
| 23 |
### Quants
|
| 24 |
|
|
@@ -35,7 +41,7 @@ tokenizer_source: base
|
|
| 35 |
merge_method: della_linear
|
| 36 |
parameters:
|
| 37 |
density: 0.5
|
| 38 |
-
epsilon: 0.4 #
|
| 39 |
lambda: 1.1
|
| 40 |
base_model: allura-org/Qwen2.5-32b-RP-Ink
|
| 41 |
models:
|
|
|
|
| 14 |
|
| 15 |
> There's no 'I' in 'brain damage'
|
| 16 |
|
| 17 |
+
|
| 18 |
|
| 19 |
### Overview
|
| 20 |
|
| 21 |
+
An attempt to make QwentileSwap write better by merging it with RP-Ink. And DeepSeek, because why not. However, I screwed up the first merge step by accidentally setting an extremely high epsilon value. Step2 wasn't planned, but due to a wonky tensor size mismatch error, I couldn't merge Step1 into QwentileSwap using sce, so I just threw in some random model. And that did, in fact, solve the issue.
|
| 22 |
+
|
| 23 |
+
The result? Well, it's usable, I guess. The slop is reduced, more details are brought up, but said details sometimes get messed up. It's fixed by a few swipes and there's a chance that it's caused by my sampler settings, but uhh I'll just leave them as they are.
|
| 24 |
+
|
| 25 |
+
Prompt format: ChatML
|
| 26 |
+
|
| 27 |
+
Settings: [This kinda works but I'm weird](https://files.catbox.moe/hh551f.json)
|
| 28 |
|
| 29 |
### Quants
|
| 30 |
|
|
|
|
| 41 |
merge_method: della_linear
|
| 42 |
parameters:
|
| 43 |
density: 0.5
|
| 44 |
+
epsilon: 0.4 #was supposed to be 0.04
|
| 45 |
lambda: 1.1
|
| 46 |
base_model: allura-org/Qwen2.5-32b-RP-Ink
|
| 47 |
models:
|