EDFlow - Evaluation Driven workFlow

A small framework for training and evaluating tensorflow models by Mimo Tilbich.

Table of Contents

1. Setup
2. Workflow
3. Example
4. Other
   1. Parameters
   2. Known Issues
   3. Compatibility
5. Contributions
6. LICENSE
7. Authors

Setup

Clone this repository:

git clone https://github.com/pesser/edflow.git
cd edflow

We provide different conda environments in the folder environments:

• edflow_tf_cu9.yaml: Use if you have CUDA>=9 available and want to use tensorflow.
• edflow_pt_cu9.yaml: Use if you have CUDA>=9 available and want to use pytorch.
• edflow_cpu: Use if you don't have a CUDA>=9 GPU available.

Choose an appropriate environment and execute

conda env create -f environments/<env>.yaml
conda activate <env>
pip install -e .

where <env> is one of the yaml files described above.

Workflow

For more information, look into our documentation.

Example

Tensorflow

cd examples
edflow -t mnist_tf/train.yaml -n hello_tensorflow

Pytorch

cd examples
edflow -t mnist_pytorch/mnist_config.yaml -n hello_pytorch

Other

Parameters

• --config path/to/config

  yaml file with all information see [Workflow][#Workflow]

• --checkpoint path/to/checkpoint to restore

• --noeval only run training

• --retrain reset global step to zero

Known Issues

Compatibility

Contributions

LICENSE

Authors

Mimo Tilbich

Contributors

Patrick Esser 💻 🤔 ✅

Johannes Haux 💻 📖 🤔

rromb ✅

arwehei 📖 🚇

Sandro Braun 💻 💡 ⚠️

Conrad Sachweh 📖 ⚠️

Ritvik Marwaha 💡

Thanks goes to these wonderful people (emoji key):

This project follows the all-contributors specification. Contributions of any kind welcome! source/source_files/edflow