BA-Zephyria-39b / README.md
Steelskull's picture
Update README.md
88d3ae5 verified
|
raw
history blame
3.34 kB
metadata
base_model:
  - unsloth/Mistral-Small-Instruct-2409
library_name: transformers
tags:
  - mergekit
  - merge

BA-Zephyria-39b [EXPERIMENTAL]

Model Information

Base Model: unsloth/Mistral-Small-Instruct-2409

Strategy: Balanced Approach

Total Layers: 55

Duplication Start: Layer 19 (34.5% of model)

Duplicated Layers: 23 (41.8% of model)

Unique Final Layers: 14 (25.5% of model)

Model Characteristics

  • Combines benefits of early and mid duplication strategies
  • Balanced between unique initial layers, duplicated middle layers, and unique final layers
  • Versatile approach suitable for a wide range of tasks
  • Provides substantial unique layers at the end for task-specific adaptations

Configuration Visualization


[    Unique    ][    Duplicated    ][    Unique    ]
0 ----------- 18 19 ------------ 41 42 ---------- 54
     34.5%           41.8%            23.7%