# 666DZY666/micronet

micronet, a model compression and deploy lib. compression: 1、quantization: quantization-aware-training(QAT), High-Bit(>2b)(DoReFa/Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inference)、Low-Bit(≤2b)/Ternary and Binary(TWN/BNN/XNOR-Net); post-training-quantization(PTQ), 8-bit(tensorrt); 2、 pruning: normal、regular and group convolutional channel pruning; 3、 group convolution structure; 4、batch-normalization fuse for quantization. deploy: tensorrt, fp32/fp16/int8(ptq-calibration)、op-adapt(upsample)、dynamic_shape

Repository: https://github.com/666DZY666/micronet
Canonical: https://ross.abutalabs.com/products/micronet
Language: Python
License: MIT
License Family: permissive
Topics: quantization, pruning, dorefa, twn, bnn, xnor-net, pytorch, model-compression, group-convolution, network-slimming, neuromorphic-computing, convolutional-networks, network-in-network, integer-arithmetic-only, quantization-aware-training, post-training-quantization, tensorrt, onnx, tensorrt-int8-python, batch-normalization-fuse
Last push: 2025-05-06T16:33:33+00:00

## Health v2 (maintenance only)
Score: 41/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 20, release rhythm 35, longevity 100
- inputs: {"age_days": 2464, "days_push": 484, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2266, forks 470 (observed 2026-08-28T04:06:32.747320+00:00)

## What it is
micronet is a Python library for deep neural network model compression and deployment built on PyTorch. It provides quantization (QAT, PTQ, high-bit and low-bit/ternary/binary schemes like DoReFa, TWN, BNN, XNOR-Net), convolutional channel pruning, batch-normalization fusion, and TensorRT deployment with fp32/fp16/int8 support.

## Use cases
- quantize a pytorch model to low-bit or binary weights for edge deployment
- compress a resnet with channel pruning before deploying to mobile
- convert a trained model to int8 with tensorrt post-training calibration
- fuse batch normalization layers into convolution weights for quantized inference
- deploy a model with tensorrt using dynamic shapes and fp16
- implement DoReFa or XNOR-Net style quantization-aware training

## When to choose
- you need both quantization and pruning in one pytorch library
- you want low-bit (ternary/binary) quantization methods like TWN, BNN, or XNOR-Net
- your deployment target is NVIDIA GPUs via TensorRT with int8 calibration
- you need batch-normalization fusion tailored to quantized models

## When to avoid
- you need compression for transformers or LLMs rather than convolutional networks
- you target non-TensorRT runtimes like ONNX Runtime or Core ML exclusively
- you want a maintained production framework with broad community support rather than a research-oriented codebase

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, llm-training
- domain: deep-learning, machine-learning, gpu-computing
- platform: python
- tags: model-compression, quantization, pruning, quantization-aware-training, post-training-quantization, tensorrt, onnx, dorefa, binary-neural-networks, network-slimming, bn-fuse, pytorch, gpu

## Member repositories
- 666DZY666/micronet (main) score 41

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:32.747320+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:42:22.008454+00:00, confidence not recorded.
  - readme: https://github.com/666DZY666/micronet (fetched 2026-08-28T04:06:32.747320+00:00, sha 8f5b5af14999)
  - registry_pypi: https://pypi.org/pypi/micronet/json (fetched 2026-08-29T10:22:31.074412+00:00, sha 0300b461257c)
- Data as of 2026-08-30T08:39:29.467469+00:00.