W&B Integration

[ayulockin]

What this PR is about?

Integration of this repo with Weights and Biases(W&B). W&B allows to track machine learning experiments, visualize metrics, and share results. You can check it out here.

Motivations

• This repository is really good for understanding the how various GAN flavors can be implemented. Thank you for this awesome resource.
• To visualize the metrics one had to write his/her own code.
• Not very convenient to do experiments involving hyperparameter tuning.
• Not compatible with latest keras or tf.keras versions/ecosystem.

Contributions:

• I tried instrumenting few scrips with W&B.
• Make it work in tf.keras ecosystem. It can be changed back to keras ecosystem easily.

Demonstration

Here is the notebook to demonstrate the use case: https://colab.research.google.com/drive/1u2gxV8tn42Vb_tmlGKa7Ww3xnEVcrxLB

(PS: You will need to signup with W&B. It's free and one can sign up using his/her GitHub account.)

Results

• Check out this run page for aae.py here.
• Check out this run page for acgan.py here.
• Check out this run page for bgan.py here.
• Check out this run page for bigan.py here.
• Check out this run page for ccgan.py here.
• Check out this run page for cgan.py here.
• Check out this run page for cogan.py here.
• Check out the added report in Additional section for dcgan.py

Additional

Here is a report to support the brilliance of your work with W&B. The instrumentation that I did will allow the users of this repo to experiment with few configurations like different latentdim. Link: https://app.wandb.ai/ayush-thakur/keras-gan/reports/Towards-Deep-Generative-Modeling-with-W%26B--Vmlldzo4MDI4Mw