Kirill Rybachuk

Depthwise separable convolutions are notoriously hard to implement efficiently. In particular, `SeparableConv2D` in Keras (as well as group convolutions in Keras and PyTorch) leads to almost no speedup on CPU,...

an option to download only added files for a given dataset version

2

So far, when calling `get_local_copy` method on a `clearml.Dataset` object, you would download all the files from the dataset AND all its parents recursively, or create soft links for all...

Updating the code / Improving wikis about MongoTrials and parallel Hyperopt

6

I was struggling 2 days in vain trying to parallelize Hyperopt. One of the problems I ran into is described [in this issue](https://github.com/hyperopt/hyperopt/issues/241). It concerns deprecated pymongo version that Hyperopt...