gaborvecsei / federated-learning-mini-framework

Federated Learning mini-framework with Keras

Hi there 👋

• Website: gaborvecsei.com
• 📫 How to reach me:
  • @gawesomebe
  • @gaborvecsei
• Products
  • image-gen.com
• ⚡ Fun fact(s):
  • I love chocolate 🍫

federated-learning-mini-framework's Issues

Parallel model training on separate CPU cores

Tensorboard logging for the global model

Saving and loading the global model weights

Parallel client model training problem

Hi mate, thank you for your work. I am new in Federated Learning, and try to deploy your FL framework using my own network. But I met an error which is "TypeError: Cannot interpret feed_dict key as Tensor: Tensor is not an element of this graph."

I searched online and tried two method adding "tf.reset_default_graph()" or "K.clear_session()" with no luck.

Finally, in this issue, they said that this issue occuers when using threading.

Any help would be appreciate.

Environment:
OS: Ubuntu18.04
python: 3.6.8
tensorflow: 1.13.1
keras: 2.2.4

Global model update factor when updating weights

Tensorboard logging for the clients

Parallel client model training on GPUs

where is the actual fed_learn function?

I mean that where is the actual fed_learn function is written, I can't find it anywhere
Everywhere the function is being imported but not defined
Any help??

Clients should send the difference of their weights

Now, they send the weights to the server but this should be only the difference