Wanted to create something similar to Monstral v2 for story writing. Instead of using migtissera/Tess-3-Mistral-Large-2-123B, I used nbeerbower/Gigaberg-Mistral-Large-123B. I wanted to include Tess into the merge, but struggled to slerp it with Behemoth, for whatever reason. This was experimental and hasn't been tested much.
This is a merge of pre-trained language models created using
mergekit.
This model was merged using the
Model Stock merge method using /workspace/cache/models--TheDrummer--Behemoth-123B-v1.2/snapshots/51354019a02b742aa5a73fe16800225ff719c46d as a base.
1base_model: /workspace/cache/models--TheDrummer--Behemoth-123B-v1.2/snapshots/51354019a02b742aa5a73fe16800225ff719c46d
2dtype: float16
3merge_method: model_stock
4slices:
5- sources:
6 - layer_range: [0, 88]
7 model: /workspace/models
8 - layer_range: [0, 88]
9 model: /workspace/bmag
10 - layer_range: [0, 88]
11 model: /workspace/cache/models--TheDrummer--Behemoth-123B-v1.2/snapshots/51354019a02b742aa5a73fe16800225ff719c46d