ChaoticVision

This is a highly experimental merge of two Mistral Llava models with the intent of splitting out a mmproj projector file with a more robust captioning capability. I do not know if this model is functional, and will not be testing it as a language model, so use at your own risk.

Merge Details

Merge Method

This model was merged using the SLERP merge method.

Models Merged

The following models were included in the merge:

• jeiku/llavamistral1.6configedit
• jeiku/noushermesvisionalphaconfigedit

Configuration

The following YAML configuration was used to produce this model:

slices:
  - sources:
      - model: jeiku/llavamistral1.6configedit
        layer_range: [0, 32]
      - model: jeiku/noushermesvisionalphaconfigedit
        layer_range: [0, 32]
merge_method: slerp
base_model: jeiku/llavamistral1.6configedit
parameters:
  t:
    - filter: self_attn
      value: [0.5, 0.5, 0.5, 0.5, 0.5]
    - filter: mlp
      value: [0.5, 0.5, 0.5, 0.5, 0.5]
    - value: 0.5
dtype: bfloat16