Edit model card

I received a request to upload the FP16 version of my first MergeMonster 13b merge, which had the sole focus of reducing overused ChatGPT roleplaying phrases. Many ministrations were murdered throughout this process.

GGUF quantizations can be found in my main MergeMonster repo.

Merge Composition

Unsurprisingly, not much remained of MythoMax. I'm (obviously!) very fond of it, but it's not perfect.

Reminder: MergeMonster works with hundreds of pieces scattered throughout the final merge. If a model takes a large part of the composition this does not mean it is merged in a manner that might be logical to humans.

Base model: MythoMax-L2-13b

------ FINAL MERGE COMPOSITION ------
LLaMA2-13B-Tiefighter: 0.27
Emerhyst-13B: 0.26
Orca-2-13b: 0.24
Stheno-1.8-L2-13B: 0.09
MythoMax-L2-13b: 0.09
Athena-v4: 0.04

Prompt Format

As always, I suggest sticking to the Alpaca format.

### Instruction:
Your instruction or question here.

### Response:
Downloads last month
9
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.