666DZY666/micronet
A Python library for compressing neural networks via quantization and pruning, and deploying them on TensorRT with FP32/FP16/INT8 support.
Not currently ranked — collecting fresh signals.
star history
The library provides model compression techniques including quantization-aware training for high-bit and low-bit/ternary/binary schemes, post-training quantization for 8-bit inference, and structured pruning strategies. It also supports batch normalization fusion for quantization efficiency. For deployment, it integrates with TensorRT offering FP32, FP16, and INT8 calibration, operation adaptation, and dynamic shape support.
Frequently asked
- What is 666DZY666/micronet?
- A Python library for compressing neural networks via quantization and pruning, and deploying them on TensorRT with FP32/FP16/INT8 support.
- Is micronet open source?
- Yes — 666DZY666/micronet is open source, released under the MIT license.
- What language is micronet written in?
- 666DZY666/micronet is primarily written in Python.
- How popular is micronet?
- 666DZY666/micronet has 2.3k stars on GitHub.
- Where can I find micronet?
- 666DZY666/micronet is on GitHub at https://github.com/666DZY666/micronet.