bamec66557/Qwen-2.5-14B-MINUS

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:

• sometimesanotion/Qwen2.5-14B-Vimarckoso-v3
• djuna/Q2.5-Veltha-14B

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: djuna/Q2.5-Veltha-14B
  - model: sometimesanotion/Qwen2.5-14B-Vimarckoso-v3
merge_method: slerp
base_model: djuna/Q2.5-Veltha-14B
dtype: bfloat16
parameters:
  t: [0, 0.2, 0.8, 0.1, 0] # Skewed towards the second model

regularization:
  - method: gradient_penalty
    scale: 0.07 # Increased for stronger regularization
  - method: weight_clipping
    clip_range: [-0.2, 0.2] # Widened range
  - method: random_noise
    scale: 0.005 # Reduced to avoid excessive noise
  - method: attention_dropout
    scale: 0.03 # Increased dropout

postprocessing:
  - operation: entropy_regularization
    scale: 0.07 # Increased for stronger effect
  - operation: non_linear_scaling
    parameters:
      function: gelu # Changed to GELU for potentially better performance
  - operation: sharpening
    intensity: 0.7 # Increased sharpening
  - operation: gaussian_smoothing
    sigma: 0.2 # Reduced smoothing for more detail
  - operation: normalize
  - operation: dynamic_scaling
    scale_range: [0.97, 1.03] # Slightly wider range
  - operation: smoothing
    parameters:
      adaptive: true
      range: [0.97, 1.03] # Slightly wider range
      kernel_size: 5 # Increased kernel size for more aggressive smoothing