helen ngo

If we don't care about matching exactly the way that the Hub does it, we can do it in pure python with pyyaml: ``` with open('README.md') as f_yaml: x =...

Oh hmmm. I thought emojis would have been filtered out with the function I built for filtering out bad strings, but it seems they're getting handled differently than other strings....

Relatedly, the download manager will fail again when downloading to S3 for the same reason (parent dir doesn't exist because it's an empty dir) because of this line: https://github.com/tensorflow/datasets/blob/1ad5b6c8cfab11eee27031c2aac22e0beeefb0a7/tensorflow_datasets/core/download/download_manager.py#L518 I...

@davisyoshida Thanks for doing this! I've been looking at implementing the same thing; will test out yours.

If you're looking to do gradient checkpointing in graph mode I suggest the implementation tf-slim here, which I've extracted and successfully tested on tf-nightly in graph mode on TPU: https://github.com/google-research/tf-slim/blob/a62dc893de5e46e6f2e9ec24a74b2abce026307a/tf_slim/layers/rev_block_lib.py

+1 to this question, I'm doing something similar with `tf-nightly==2.2.0.dev20200119` with custom layers inherited from Keras layers and the docs could use some clarification on "keep a reference to the...

All my outstanding questions/comments re: cudagraphs are resolved. I will defer formal approval on behalf of @NVIDIA/inference in case anyone else wants to weigh in before merge.