Tested to 15872 tokens!
GGUF - nocudaexe/Neural-Dark-Waifu-V0.2
The first attempt at Neural-Dark-Waifu started showing strange behaviour at 4-8k context, so I re-merged, and did more testing, this one is exhibiting the correct behaviours
An attempt to merge mlabonne/AlphaMonarch-7B excellent chat ability, with a bunch of ERP thrown at it to bring down it's guardrails, as well as Test157t/Kunocchini-7b-128k-test to increase context
license: apache-2.0 tags: - merge - mergekit - lazymergekit - nocudaexe/Dark-Waifu-7b - nocudaexe/Infinite-Waifu
DarkNeural
DarkNeural is a merge of the following models using mergekit:
🧩 Configuration
```yamlmodels: # No parameters necessary for base model
- model: nocudaexe/Dark-Waifu-7b parameters: density: 0.33 weight: 0.4
- model: nocudaexe/Infinite-Waifu parameters: density: 0.38 weight: 0.3 merge_method: Slerp
parameters: int8_mask: true dtype: bfloat16 ```