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:
slices:
- sources:
- model: bamec66557/mergekit-model_stock-zdaysvi
layer_range: [0, 20]
- model: bamec66557/MISCHIEVOUS-12B-Mix_0.4v
layer_range: [0, 20]
parameters:
t:
- value: 0.8
- sources:
- model: bamec66557/mergekit-model_stock-zdaysvi
layer_range: [20, 40]
- model: bamec66557/MISCHIEVOUS-12B-Mix_0.4v
layer_range: [20, 40]
parameters:
t:
- value: 1.0
- filter: self_attn
value: [0.8, 0.9, 1.0, 1.1, 1.2]
merge_method: slerp # Preserve merge method
base_model: bamec66557/mergekit-model_stock-zdaysvi # Base model
dtype: bfloat16 # Data types for fast merges
# Additional options
regularization:
- method: weight_clipping
clip_range: [-0.1, 0.1]
postprocessing:
- operation: gaussian_smoothing
sigma: 1.5 # Gaussian smoothing intensity
- operation: smoothing
parameters:
adaptive: true
range: [0.8, 1.2] # Adaptively adjust
kernel_size: 5 # Smoothing larger ranges with increased kernel size
- operation: normalize # Normalise after merge
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 26.05 |
| IFEval (0-Shot) | 63.65 |
| BBH (3-Shot) | 30.51 |
| MATH Lvl 5 (4-Shot) | 11.25 |
| GPQA (0-shot) | 8.50 |
| MuSR (0-shot) | 12.44 |
| MMLU-PRO (5-shot) | 29.96 |