merge_linear_normbalanced

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

Merge Details

Merge Method

This model was merged using the Linear merge method.

Models Merged

The following models were included in the merge:

• sraj/CMB_MARK_CX_LRD
• sraj/CMB_FWEdu_V2_FastTxt_CX_LRD
• sraj/CMB_WX_SYN_CX_LRD

Configuration

The following YAML configuration was used to produce this model:

# Norm-balanced: weights set inversely proportional to avg task vector L2 norm
# A ≈ 16, F ≈ 43, S ≈ 16 → weights ≈ 1/16, 1/43, 1/16 → normalized ≈ 2.7, 1.0, 2.7
models:
  - model: sraj/CMB_MARK_CX_LRD
    parameters:
      weight: 2.7
  - model: sraj/CMB_FWEdu_V2_FastTxt_CX_LRD
    parameters:
      weight: 1.0
  - model: sraj/CMB_WX_SYN_CX_LRD
    parameters:
      weight: 2.7
merge_method: linear
parameters:
  normalize: true
dtype: bfloat16