Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/adamvvu/tf_compactprogbar
A simple and compact one-line progress bar for TensorFlow 2 Keras.
https://github.com/adamvvu/tf_compactprogbar
keras keras-callback tensorflow tensorflow-addons tensorflow2
Last synced: 1 day ago
JSON representation
A simple and compact one-line progress bar for TensorFlow 2 Keras.
- Host: GitHub
- URL: https://github.com/adamvvu/tf_compactprogbar
- Owner: adamvvu
- License: mit
- Created: 2024-10-21T06:19:18.000Z (4 months ago)
- Default Branch: master
- Last Pushed: 2024-12-12T00:18:38.000Z (about 2 months ago)
- Last Synced: 2024-12-12T01:22:02.064Z (about 2 months ago)
- Topics: keras, keras-callback, tensorflow, tensorflow-addons, tensorflow2
- Language: Python
- Homepage:
- Size: 128 KB
- Stars: 1
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
[](https://github.com/adamvvu/tf_compactprogbar/actions/workflows/tf_compactprogbar_tests.yml)
[](https://pypi.org/project/tf-compactprogbar/)
[](https://pypi.org/project/tf-compactprogbar/)
[](https://github.com/adamvvu/tf_compactprogbar/blob/master/LICENSE)
[](https://mybinder.org/v2/gh/adamvvu/tf_compactprogbar/HEAD?labpath=examples%2Fexample.ipynb)
A simple and compact one-line progress bar for TensorFlow 2 Keras.
---
## TensorFlow Compact Progress Bar
Existing ways of monitoring training of TensorFlow Keras models either display nothing (`verbose=0`), one progress bar per epoch (`verbose=1`), or prints one line of metrics per epoch (`verbose=2`). When training for thousands of epochs, this often leads to bloated log files or crashed/sluggish Jupyter environments when working interactively.
This library provides a compact progress bar that simply displays the overall training progress by epoch. There are also a few small additional features for convenience, such as excluding certain metrics to avoid excessive clutter.
**Notebook mode**
**Console mode**
### Getting Started
Install from PyPi:
`$ pip install tf-compactprogbar`
#### Dependencies
- `tensorflow >= 2` *(TF 1 likely works, but untested)*
- `tqdm`
- `python >= 3.7`
- `ipywidgets` *(Optional)*
- `jupyterlab_widgets` *(Optional)*
For nice looking Jupyter/IPython progress bars, make sure you have `ipywidgets` and `jupyterlab_widgets` if you are on Jupyter Lab.
#### Usage
To use it, disable the built-in logging (`verbose=0`) and pass it in as a Callback:
```
from tf_compactprogbar import CompactProgressBar
progBar = CompactProgressBar()
history = model.fit(X_train, Y_train,
epochs=200,
batch_size=100,
verbose=0,
validation_data = (X_test, Y_test),
callbacks=[progBar])
```
### Documentation
```
# Call signature
CompactProgressBar(show_best=True, best_as_max=[], exclude=[], notebook='auto', epochs=None)
Args:
- show_best (bool) Display best metrics. Default: True
- best_as_max (list) Metrics which should be maximized (see note)
- exclude (list) Metrics which should be excluded from display
- notebook (str/bool) Whether to use IPython/Jupyter widget or console. Default: 'auto'
- epochs (int) Optional total number of epochs. Default is inferred from `.fit`.
Note: When using `show_best`, by default the "best" metric is the minimum. Pass
in the metric name to `best_as_max` to change this behavior.
```
If there are any issues in Jupyter, please see the [tqdm Issues](https://github.com/tqdm/tqdm/issues) page for help or disable notebook mode with `notebook=False`.