sovrasov/flops-counter.pytorch
A PyTorch library that counts multiply-add operations, parameters, and per-layer costs in deep neural networks.
Not currently ranked — collecting fresh signals.
star history
ptflops is a computational complexity analysis tool for PyTorch models. It supports two backends: a legacy pytorch backend for CNN layer analysis and an aten backend that covers transformer architectures and aten operations. The library can compute FLOPs, parameter counts, and print per-module statistics, making it useful for model benchmarking, optimization, and comparing architectures during research and development.
Frequently asked
- What is sovrasov/flops-counter.pytorch?
- A PyTorch library that counts multiply-add operations, parameters, and per-layer costs in deep neural networks.
- Is flops-counter.pytorch open source?
- Yes — sovrasov/flops-counter.pytorch is open source, released under the MIT license.
- What language is flops-counter.pytorch written in?
- sovrasov/flops-counter.pytorch is primarily written in Python.
- How popular is flops-counter.pytorch?
- sovrasov/flops-counter.pytorch has 3k stars on GitHub.
- Where can I find flops-counter.pytorch?
- sovrasov/flops-counter.pytorch is on GitHub at https://github.com/sovrasov/flops-counter.pytorch.