https://github.com/mtli/nntime

Timing utilities for deep learning modules in PyTorch :stopwatch:
https://github.com/mtli/nntime

Last synced: about 2 months ago
JSON representation

Timing utilities for deep learning modules in PyTorch :stopwatch:

• Host: GitHub
• URL: https://github.com/mtli/nntime
• Owner: mtli
• License: mit
• Created: 2021-01-25T07:38:40.000Z (over 3 years ago)
• Default Branch: main
• Last Pushed: 2022-02-23T05:57:10.000Z (over 2 years ago)
• Last Synced: 2024-04-21T03:43:22.831Z (5 months ago)
• Language: Python
• Homepage:
• Size: 88.9 KB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# nntime - Simple Utilities for Timing Artificial Neural Networks

Want to time some of your deep learning code, but find general-purpose profilers hard to work with? Then you are here at the right place!



- Offers an easy way to manually mark and time only relevant code

- Auto exports all timings to a table (CSV file)

- Timer markers can optionally include semantic call depths (not actual but uninterpretable ones) and the output will be organized by these call depths

- Works with CPU/GPU

- Introduces negligible overhead

- (New in 0.1.1) Recursively times all submodules with automatic depth markers

Check out the `examples` folder for usage.

## Installation

[![PyPI version](https://badge.fury.io/py/nntime.svg)](https://badge.fury.io/py/nntime)

Install [PyTorch](https://pytorch.org/) and run:

```

pip install nntime

```