Wu, Ke

If your client ids are sufficiently randomized (e.g. datasets under fedjax.datasets are; we appended a random id to the original client id when creating these datasets), you can use FederatedData.slice...

Sorry, I misread your original question. A ClientDataset can be sliced using the [] operator (only slicing is supported): ``` _, client_dataset = next(fedjax.datasets.emnist.load_data()[0].clients()) client_dataset[:len(cd)//10] ``` This creates a new...

In most cases you can use `preprocess_cilent` for this. Check out https://fedjax.readthedocs.io/en/latest/notebooks/dataset_tutorial.html#preprocessing-at-the-client-level. We can append to the current preprocessing another function that slices the partially processed examples as the final...

> Hi, has any work been done for this issue? Is there still a need for it? > > More generally, what is the state of this repo? Is it...

Hi @marcociccone! I am not very familiar with these two datasets. By "images have different shapes" do you mean images in these datasets are not already transformed into a uniform...

Sorry about the confusion, I didn't know the datasets were this big (should have read the READMEs more carefully). Could you help me run some quick stats on gldv2? That...

The latest usable copy on archive.org appears to https://web.archive.org/web/20180421011035/http://research.baidu.com:80/bringing-hpc-techniques-deep-learning/

Sorry for the late response. Somehow the email notification slipped through all team members' inboxes. I cannot reproduce the problem on TensorFlow 2.5.3 installed from pip. Could you tell us...

@eustlb `make fixup` is not yet passing due to `hop_length` not being used in modelling code.

Unfortunately this will break decoding with LM. Perhaps we should just hardcode hop_length to a fixed value for now in LasrFeatureExtractor?