Pedro Porto Buarque de Gusmão

I believe we do not have a `get_properties()` implementation yet.

Hello @MakGulati, When sending back the model parameters to the server, you might want to change the [ tensor_type: str](https://github.com/adap/flower/blob/1bc9649204821a204e25436034c955f5fc54c126/src/py/flwr/common/typing.py#L59) of your `Parameters` before returning them in `fit`. You could...

Hi @MakGulati , Sorry, I means using the `tensor_type` as the tag: `parameters.tensor_type='model_a'` after having created the `parameters` object

Hi @MakGulati , We are currently looking in to how to allow for a Strategy to have multiple Parameters. Will get back to you soon.

The original problem is when we try to assign `Dict[str, Scalar]` to `ClientMessages` in general as seen below. Apparently there `typing.Mapping` is how protobuf translates `map` messages and it is...

Hi @sandracl72 , Thanks for this. I'm guessing you have already tried with `FedAvg` and it works, right? Could you run a few tests for me, please? In your [FedYogi](https://github.com/adap/flower/blob/main/src/py/flwr/server/strategy/fedyogi.py),...

Hi @sandracl72, thanks for this. From the log file I don't see a NaN on the prints, but only in; `File "/workspace/flower/src/py/flwr/server/strategy/fedavg.py", line 178, in evaluate`. Could you try and...

Hi @sandracl72 , I might be wrong, but could it be that evaluation is performing something weird? From the log I see `INFO flower 2022-02-16 08:34:53,155 | fedavg.py:182 | Number...

@danieljanes this might be related to BatchNorm. Do you remember if we had any example with aggregation and batch norm?

This is being investigated, but for now, I'd recommend: - Changing swapping `BatchNorm`s for `GroupNorm` layers. - Make sure `grpcio==1.43` in `poetry.lock`. If not, run `poetry add grpcio==1.43`. This is...