Declarative data routing for PyTorch modules.
Open the hood on neural networks. Component-level model surgery, analysis, and composition.
BayLing-MLingual: 2500-way multilingual QA across 50 languages without LLM retraining
Experiments in learned communication, model composition, and compact neural protocols
Composable probabilistic models in Rust and Triton, with independently trained sources.
Python research framework for modular Mixture-of-Experts (MoE), LoRA adapters, and routed specialization of frozen language models with exact base-model fallback.
To associate your repository with the model-composition topic, visit your repo's landing page and select "manage topics."