Parth Thakkar

You could have a server that pushes the logs to the bots front end. I'd maybe not put it in the same python process though. On Wed, Dec 26, 2018,...

I have a possibly dumb question -- is there a benefit of doing this? I thought Triton already supports dynamic batching, so why would proxy code need to handle this?

@fdegier thanks for your [PR](https://github.com/thakkarparth007/fauxpilot/pull/1) that merges python_backend code on top of #73 I made a few small changes on top of those (except your last commit since my changes...

Update: Fixed the test and merged with main. For running the test, you can just run `pytest tests` from the root directory.

Yup, that's exactly the case. from vocab.json: "()":3419 "(": 7

Perhaps if you pre-compute an index of the vocabulary, of the form (char -> [toks_containing_this_char]), and/or bigrams ("c1,c2" -> [toks containing c1c2]), this might be efficient enough. Unigram index has...

Hmm, I wasn't aware that we could omit the field name if there was only one field. Also, the golang/go#12854 PR would make it much simpler, but I still don't...

@aajtodd how the current implementation works is by generating a field for the one-of, rather than the children of one-of. For example: ```protobuf message Foo { oneof bar { uint32...

This particular problem can be solved by replacing the following line in `seq2seq.training.utils.py`: ```cell_args = set(inspect.getargspec(cell_class.__init__).args[1:])``` (found in cell_from_spec) method with the following line: ```cell_args = set(list(inspect.signature(cell_class.__init__).parameters)[1:])``` But of course...

Closing this as @Frederisk finds it working. If still an issue, please reopen.