CultriX/Qwen2.5-14B-MultiCultyv2

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 CultriX/SeQwence-14Bv1 as a base.

Models Merged

The following models were included in the merge:

• allknowingroger/QwenSlerp6-14B
• sometimesanotion/Lamarck-14B-v0.3
• allknowingroger/QwenStock3-14B
• CultriX/Qwen2.5-14B-Wernicke
• CultriX/SeQwence-14B-EvolMerge
• VAGOsolutions/SauerkrautLM-v2-14b-DPO

Configuration

The following YAML configuration was used to produce this model:

base_model: CultriX/SeQwence-14Bv1
merge_method: dare_ties
dtype: bfloat16
parameters:
  normalize: true
  int8_mask: true
models:
  - model: CultriX/SeQwence-14Bv1
    parameters:
      weight: 0.25
      density: 0.6
  - model: allknowingroger/QwenSlerp6-14B
    parameters:
      weight: 0.2
      density: 0.6
  - model: VAGOsolutions/SauerkrautLM-v2-14b-DPO
    parameters:
      weight: 0.2
      density: 0.6
  - model: CultriX/SeQwence-14B-EvolMerge
    parameters:
      weight: 0.12
      density: 0.5
  - model: sometimesanotion/Lamarck-14B-v0.3
    parameters:
      weight: 0.1
      density: 0.5
  - model: allknowingroger/QwenStock3-14B
    parameters:
      weight: 0.08
      density: 0.5
  - model: CultriX/Qwen2.5-14B-Wernicke
    parameters:
      weight: 0.05
      density: 0.6
adaptive_merge_parameters:
  task_weights:
    IFEval: 1.3
    BBH: 1.2
    MATH: 1.1
    GPQA: 1.2
    MUSR: 1.2
    MMLU_PRO: 1.2
  smoothing_factor: 0.1
gradient_clipping: 1