https://github.com/percevalw/rich-logger

Table logger using Rich
https://github.com/percevalw/rich-logger

Last synced: 3 months ago
JSON representation

Table logger using Rich

• Host: GitHub
• URL: https://github.com/percevalw/rich-logger
• Owner: percevalw
• License: bsd-3-clause
• Created: 2021-02-05T14:18:06.000Z (about 4 years ago)
• Default Branch: master
• Last Pushed: 2023-07-31T19:41:00.000Z (over 1 year ago)
• Last Synced: 2024-11-01T19:52:32.759Z (4 months ago)
• Language: Python
• Homepage:
• Size: 505 KB
• Stars: 8
• Watchers: 3
• Forks: 2
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
![tests](https://github.com/percevalw/rich-logger/actions/workflows/tests.yml/badge.svg)

[![pypi](https://badge.fury.io/py/rich-logger.svg)](https://pypi.org/project/rich-logger)

# rich-logger

Table logger using Rich, aimed at Pytorch Lightning logging

## Features

- display your training logs with pretty [rich](https://github.com/willmcgugan/rich) tables

- describe your fields with `goal` ("higher_is_better" or "lower_is_better"), `format` and `name`

- a field descriptor can be matched with any regex

- a field name can be computed as a regex substitution

- works in Jupyter notebooks as well as in a command line

- integrates easily with [Pytorch Lightning](https://github.com/PyTorchLightning/pytorch-lightning)

## Demo

```python

from rich_logger import RichTablePrinter

import time

import random

from tqdm import trange

logger_fields = {

    "step": {},

    "(.*)_precision": {

        "goal": "higher_is_better",

        "format": "{:.4f}",

        "name": r"\1_p",

    },

    "(.*)_recall": {

        "goal": "higher_is_better",

        "format": "{:.4f}",

        "name": r"\1_r",

    },

    "duration": {"format": "{:.1f}", "name": "dur(s)"},

    ".*": True,  # Any other field must be logged at the end

}

def optimization():

    printer = RichTablePrinter(key="step", fields=logger_fields)

    printer.hijack_tqdm()

    t = time.time()

    for i in trange(10):

        time.sleep(random.random() / 3)

        printer.log(

            {

                "step": i,

                "task_precision": i / 10.0 if i < 5 else 0.5 - (i - 5) / 10.0,

            }

        )

        time.sleep(random.random() / 3)

        printer.log(

            {

                "step": i,

                "task_recall": 0.0 if i < 3 else (i - 3) / 10.0,

                "duration": time.time() - t,

            }

        )

        printer.log({"test": i})

        t = time.time()

        for j in trange(5):

            time.sleep(random.random() / 10)

    printer.finalize()

optimization()

```

![Demo](demo.gif)

## Use it with PytorchLightning

```python

from rich_logger import RichTableLogger

trainer = pl.Trainer(..., logger=[RichTableLogger(key="epoch", fields=logger_fields)])

```