Titan-123b

Titan consists of Behemoth 1.2, Magnum v4, Tess 3, and Gigaberg Mistral Large. Using Monstral v2 as inspiration, I merged (SLERP) each of the models with Behemoth, one by one. Then merged (model stock) the 3 resulting models together to create Titan.

I mainly use LLMs for creative writing and storytelling. This one turned out pretty decently for that purpose. Comparing it to Monstral, it has it's similarities but feels like a different flavor. It still has some tendencies to take stories in particular directions, but I'm still enjoying playing with it.

Not a ground breaking merge by any means, but nice to mix things up with. (I'm just some random guy who experiments merging models for fun)

Works well with Mistral and Metharme templates.

Personal preference for settings (needs more testing)

Temp: .99

Top_p: .9 - .95

Typical_p: 1

Min_p: 0 (needs more testing)

Recommend using and experimenting with DRY and XTC

Merge Details

Merge Method

This model was merged using the Model Stock merge method using /workspace/cache/models--TheDrummer--Behemoth-123B-v1.2/snapshots/51354019a02b742aa5a73fe16800225ff719c46d as a base.

Models Merged

The following models were included in the merge:

  • /workspace/Titan
  • /workspace/cache/models--bruhzair--Behemoth-Magnum-v4-SLERP-123b/snapshots/605e4c7b0ea17bd8afb6fcddefbd40235329c2b7
  • /workspace/cache/models--bruhzair--Behemoth-Gigaberg-SLERP-0.2/snapshots/69098286a26291009468fe22ccbe8ba659debfb0

Configuration

The following YAML configuration was used to produce this model:

base_model: /workspace/cache/models--TheDrummer--Behemoth-123B-v1.2/snapshots/51354019a02b742aa5a73fe16800225ff719c46d
dtype: float16
merge_method: model_stock
slices:
- sources:
  - layer_range: [0, 88]
    model: /workspace/Titan
  - layer_range: [0, 88]
    model: /workspace/cache/models--bruhzair--Behemoth-Gigaberg-SLERP-0.2/snapshots/69098286a26291009468fe22ccbe8ba659debfb0
  - layer_range: [0, 88]
    model: /workspace/cache/models--bruhzair--Behemoth-Magnum-v4-SLERP-123b/snapshots/605e4c7b0ea17bd8afb6fcddefbd40235329c2b7
  - layer_range: [0, 88]
    model: /workspace/cache/models--TheDrummer--Behemoth-123B-v1.2/snapshots/51354019a02b742aa5a73fe16800225ff719c46d
Downloads last month
28
Safetensors
Model size
123B params
Tensor type
FP16
·
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.

Model tree for bruhzair/Titan-123b-0.1

Quantizations
2 models