mergekit-community/mergekit-dare_ties-lmociuf

merge

This is a merge of pre-trained language models created using mergekit.

Merge Details

Merge Method

This model was merged using the DARE TIES merge method using mistralai/Mistral-Small-24B-Base-2501 as a base.

Models Merged

The following models were included in the merge:

• ReadyArt/Forgotten-Safeword-24B-3.6
• huihui-ai/Mistral-Small-24B-Instruct-2501-abliterated
• TheDrummer/Cydonia-24B-v2.1
• PocketDoc/Dans-PersonalityEngine-V1.2.0-24b

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: mistralai/Mistral-Small-24B-Base-2501
    # No parameters necessary for the base model
  - model: huihui-ai/Mistral-Small-24B-Instruct-2501-abliterated
    parameters:
      density: 0.5  # Retaining 50% of this model's parameters
      weight: 0.1   # Lower influence
  - model: TheDrummer/Cydonia-24B-v2.1  # Highest influence
    parameters:
      density: 0.9  # Retaining 90% of this model's parameters
      weight: 0.4   # Highest influence
  - model: PocketDoc/Dans-PersonalityEngine-V1.2.0-24b  # Second highest influence
    parameters:
      density: 0.7  # Retaining 70% of this model's parameters
      weight: 0.3   # Second highest influence
  - model: ReadyArt/Forgotten-Safeword-24B-3.6
    parameters:
      density: 0.6  # Retaining 60% of this model's parameters
      weight: 0.2   # Moderate influence

merge_method: dare_ties
base_model: mistralai/Mistral-Small-24B-Base-2501
parameters:
  normalize: true  # Normalizes parameter scaling for consistency
  int8_mask: true  # Optimizes for memory-efficient int8 operations
dtype: bfloat16  # Maintains computations in bfloat16 for performance efficiency