Add crdt-merge — CRDT-guaranteed model merging with 25 strategies

[mgillr]

Add crdt-merge

crdt-merge is a Python library that brings CRDT (Conflict-free Replicated Data Type) guarantees to data merging — covering DataFrames, JSON, ML model weights, and distributed agent state.

Key features:

• Mathematical guarantees: Every merge is commutative, associative, and idempotent
• 25 merge strategies including SLERP, TIES, DARE, Fisher merge, and task arithmetic
• LoRA adapter merging with rank harmonization
• Federated learning support (FedAvg, FedProx)
• 2,600+ tests passing
• pip installable: pip install crdt-merge

Thank you for considering this addition!

[EnnengYang]

Thank you for your contribution! I have added this excellent library to the following subsections:

https://github.com/EnnengYang/Awesome-Model-Merging-Methods-Theories-Applications?tab=readme-ov-file#benchmarkevaluation

[mgillr]

@EnnengYang Thank you for including crdt-merge in the Benchmark/Evaluation section. Would it be possible to update the Remark to better reflect the project's core contribution? The current description doesn't distinguish it from existing toolkits like MergeKit.
Suggested Remark: "CRDT-based distributed model merging with formal convergence guarantees. 25 strategies (SLERP, TIES, DARE, Fisher, evolutionary). Two-layer OR-Set architecture enabling conflict-free multi-node merge."
The full technical architecture document is here: https://github.com/mgillr/crdt-merge/blob/main/docs/CRDT_ARCHITECTURE.md

[EnnengYang]

The "Remark" has been updated as per your suggestion. Thank you for your contribution.