Knowledge Distillation Python Packages

neural-compressor

SOTA low-bit LLM quantization (INT8/FP8/MXFP8/INT4/MXFP4/NVFP4) & sparsity; leading model compression techniques on PyTorch, TensorFlow, and ONNX Runtime

freesolo-flash-dev

LoRA post-training for open-weight models: SFT, GRPO, and on-policy distillation. Describe a run in TOML; Flash allocates a GPU, trains, streams checkpoints, and serves the adapter.

suod

(MLSys' 21) An Acceleration System for Large-scare Unsupervised Heterogeneous Outlier Detection (Anomaly Detection)

ssondo

Official SSONDO implementation and trained model's weights (ICASSP 2026)

jevstiller

Distill a repeated Jev classification task into a local model, on the fly — same answers, your hardware.

freesolo-flash

LoRA post-training for open-weight models: SFT, GRPO, and on-policy distillation. Describe a run in TOML; Flash allocates a GPU, trains, streams checkpoints, and serves the adapter.

shrewd

Turn repeated LLM classification calls into a small, fast, local model for one fixed task.

miniverl

verl for a single consumer GPU. PPO, GRPO and on-policy distillation on NVIDIA GPUs.

paddleclas

A treasure chest for visual classification and recognition powered by PaddlePaddle

neural-compressor-pt

SOTA low-bit LLM quantization (INT8/FP8/MXFP8/INT4/MXFP4/NVFP4) & sparsity; leading model compression techniques on PyTorch, TensorFlow, and ONNX Runtime

thekaveh-nnx

Lightweight PyTorch toolkit for training, fine-tuning, and exporting modern neural nets. FFN, GNN, decoder-only LM, diffusion, JEPA, MoE, PEFT, quantization, pruning, surgery, ONNX and Hugging Face Hub interop, plus experimental GGUF/Ollama bundle export.

mmrazor

OpenMMLab Model Compression Toolbox and Benchmark.

neural-compressor-tf

SOTA low-bit LLM quantization (INT8/FP8/MXFP8/INT4/MXFP4/NVFP4) & sparsity; leading model compression techniques on PyTorch, TensorFlow, and ONNX Runtime

torchdistill

A coding-free framework built on PyTorch for reproducible deep learning studies. PyTorch Ecosystem. 🏆26 knowledge distillation methods presented at TPAMI, CVPR, ICLR, ECCV, NeurIPS, ICCV, AAAI, etc are implemented so far. 🎁 Trained models, training logs and configurations are available for ensuring the reproducibiliy and benchmark.

resonance-layer

Emotional intelligence layer for AI - detects emotion, tracks wellbeing, and gives any LLM the context to respond to how people actually feel.

neural-compressor-jax

SOTA low-bit LLM quantization (INT8/FP8/MXFP8/INT4/MXFP4/NVFP4) & sparsity; leading model compression techniques on PyTorch, TensorFlow, and ONNX Runtime

optipfair

Structured pruning, knowledge distillation, and fairness analysis for Large Language Models.

growmos

🪐 Growmos: A new approach to agentic AI—a living system expanding its knowledge like the universe, compounding its capabilities, and getting more work done with fewer tokens.

tokenknows-mcp

Distill AI coding sessions (Claude Code / Codex / Cursor) into weekly reports, ADRs, incident reviews and a knowledge graph — local-first, evidence-linked.

silverspoon-kd

A general-purpose toolkit for knowledge distillation of transformer models

taskdistill

Distil an expensive LLM API call on a narrow task into a small local model: capture traffic, curate, LoRA fine-tune with MLX on Apple Silicon, evaluate against the teacher, and serve an OpenAI-compatible cascade that escalates low-confidence requests.

sqlite-predict

Zero-shot forecasting, anomaly detection, and tabular prediction as SQLite/libSQL functions. Distill your model (or a licensed foundation model) into a tiny student that lives inside your database. Pure C, no dependencies, runs anywhere.

kd-lib

A Pytorch Knowledge Distillation library for benchmarking and extending works in the domains of Knowledge Distillation, Pruning, and Quantization.

distill-anything

Distill any AI model into one you own — a teacher (Claude/GPT/HF/Ollama) generates your dataset, a student trains on its logits or its words, a judge scores it blind, a benchmark prices it. Runs on a MacBook.