model-surgery

Open the hood on neural networks. Component-level model surgery, analysis, and composition.

A local AI model training, distillation, model surgery, GGUF conversion, and Ollama export workbench. 本地 AI 模型训练、蒸馏、模型手术与导出工具箱。

Deterministic tensor patch plans, patch application, and tensor path traversal for neural network model surgery.

End-to-end audio-native SLM fine-tuning & model surgery framework (Qwen3-Omni backbone) for Indic telephony speech-to-speech.

Visual model merging, layer surgery, and weight interpolation studio for local LLMs

Turn any HF LLM into an attention+recurrent hybrid in minutes on one consumer GPU: per-layer surgery, teacher distillation, executable-exam gating.

Nx SVD/SVF factorization primitives for neural network model surgery and TRINITY artifact export.

🧠🔪 Digital lobotomy for local LLMs: cut layers, swap words, inject moods — and compare the patient with a healthy donor. Web UI, runs 100% locally.

Public EPI research scaffold for energy-aware AI evaluation, measurement discipline, and claim boundaries.

SafeTensors parsing, validation, bounded slicing, checksums, deterministic writing, and row-chunk helpers for Elixir and Nx.

Bidirectional synthesis of ultrametric dyadic sparse attention (Llama Surgery) and lateralized neuromorphic test-time adaptation (Brain-AI). Lean 4 verified (0 sorry). 27/400 canonical ARC-AGI-1 & surpasses dense F1 on LongBench QASPER ($n=200$) with zero new weights.