metadata

base_model:
  - CultriX/Qwen2.5-14B-Hyperionv5
  - arcee-ai/Virtuoso-Small-v2
  - sometimesanotion/Lamarck-14B-v0.7-rc4
  - sthenno-com/miscii-14b-1225
  - CultriX/Enhanced-TIES-Base-v1
  - CultriX/Qwen2.5-14B-Hyperionv3
library_name: transformers
tags:
  - mergekit
  - merge
license: apache-2.0
language:
  - en
pipeline_tag: text-generation

merge

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

Merge Details

Merge Method

This model was merged using the SLERP merge method.

Models Merged

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:

name: UltimateMerge-14B-v1-slerp
merge_method: slerp
base_model: CultriX/Enhanced-TIES-Base-v1
tokenizer_source: base
dtype: float32
out_dtype: bfloat16
parameters:
  normalize: true
  rescale: false
  int8_mask: true
parameters:
  t:
    - value: 0.3 # Small value for refined tuning.
slices:
  - sources:
      - model: CultriX/Enhanced-TIES-Base-v1 #Base model.
        layer_range: [ 0, 8 ]
      - model: arcee-ai/Virtuoso-Small-v2 #The arcee model focuses on IFEval so it is the best as the output layer
        layer_range: [ 0, 8 ]
  - sources:
      - model: CultriX/Enhanced-TIES-Base-v1
        layer_range: [ 8, 16 ]
      - model: sometimesanotion/Lamarck-14B-v0.7-rc4 #Strong average score.
        layer_range: [ 8, 16 ]
  - sources:
      - model: CultriX/Enhanced-TIES-Base-v1
        layer_range: [ 16, 24 ]
      - model: CultriX/Qwen2.5-14B-Hyperionv3 #Improve Reasoning benchmarks.
        layer_range: [ 16, 24 ]
  - sources:
      - model: CultriX/Enhanced-TIES-Base-v1
        layer_range: [ 24, 32 ]
      - model: CultriX/Qwen2.5-14B-Hyperionv5 # General Benchmarks.
        layer_range: [ 24, 32 ]
  - sources:
      - model: CultriX/Enhanced-TIES-Base-v1
        layer_range: [ 32, 40 ]
      - model: sthenno-com/miscii-14b-1225 #IFEval and BBH
        layer_range: [ 32, 40 ]
  - sources:
      - model: CultriX/Enhanced-TIES-Base-v1
        layer_range: [ 40, 48 ]
      - model: arcee-ai/Virtuoso-Small-v2 #Output layer
        layer_range: [ 40, 48 ]