Yann Dupis

Hey @justin1121, with your recent work on TF2.0, how does the API look like now if we want to add two private tensors? Does it look like this? ``` @tf.function...

Gotcha, thanks! that's helpful. what would be the equival in plaintext TF2.0? ``` @tf.function def func(): x = tf.constant([1, 2, 3, 4]) y = tf.constant([5, 6, 7, 8]) res =...

Thank you @burhanusman for raising this issue. In general `fit` needs to be improved to perfectly match tf.keras That's correct, currently we are not batching the data in `fit`. It...

@mortendahl - happy to explore other alternatives. In general, I think it's important that's easy to use and we can support arbitrary Keras models. So maybe we just need to...

> Can you provide some more details about this also? Not 100% sure what you mean. In tf federated, they have this concept of [tff.simulation.datasets.emnist.load_data](https://www.tensorflow.org/federated/tutorials/federated_learning_for_image_classification). It constructs a `tf.data.Dataset` object...

I saw this a couple of times. Not sure what's the problem. However it's always the same test which is failing. And only occasionally. If it's a graph/session cross-over issue...

`predict` is blocked by #590. In order to get the following flow we will haver to use K.get_session() inside predict: ``` model = tfe.keras.models.Sequential() model.add(tfe.keras.layers.Dense(5, use_bias=False, input_shape=[5])) model.set_weights([np.ones([5, 5])]) y...

@jvmancuso I think we can close this issue?

Hey @shreyansh26, thanks for raising this issue. Sorry for the delay. I have started to investigate. I think I know where is the problem. Under the hood, when we do...

ok the second problem is that the Batchnorm layer as the following weigths: ``` model.layers[2].weights [, , , ] ``` However when instantiating the Batchnorm TFE layer we are not...