All Tools
M
Fine-tuningFreeOpen Source
MERGOO
Merge fine-tuned LLM experts into one powerful model
LGPL-3.0
ABOUT
Fine-tuning separate models for different domains creates a fragmentation problem — teams maintain multiple LLM checkpoints for coding, writing, math, and domain-specific tasks, duplicating infrastructure and degrading per-task performance. Mergoo combines multiple fine-tuned expert models into a single merged model using techniques like weight averaging and task-vector composition. The result preserves each expert's strengths without requiring additional training data or compute.
INSTALL
pip install mergooINTEGRATION GUIDE
1. Merge a code-specialized model with a writing-specialized model to get unified coding-and-documentation performance
2. Combine domain-specific fine-tunes (legal, medical, finance) into one versatile enterprise model
3. Average multiple RL-fine-tuned checkpoints to produce a more robust instruction-following model
4. Compose task vectors from LoRA adapters to enable multi-task inference without separate deployments
5. Merge a math reasoning expert with a general chat expert for an assistant that handles both logic and conversation
TAGS
pythonfine-tuningllmmodel-mergingensembletraining