Somshubra Majumdar

There's a keras implementation here - https://github.com/titu1994/keras-adabound It also provides weights for ResNet 34 on CIFAR. Note, final acc isn't the same as paper because I didn't apply random crop...

It is not currently possible to partially update the Tokenizer - Sentencepiece itself does not support it. It does support adding extra tokens when the tokenizer is first built but...

Cool hack of the underlying library of Sentencepiece. Guess I can correct my statement to "there is no sanctioned, official way offered by Sentencepiece to add tokens to a prebuilt...

Seems import test is failing due to some syntax error

@erastorgueva-nv could you respond with some info here

Sorry but we can't do busy waiting as your pr requires. It has a measurable impact on all domain training. Please think of another solutions

4410 is localized to one model, and it's upto that model owner to choose busy waiting. However, #4544 affects core, effectively all Nemo models, and we will not accept busy...

As far as I remember, it's to standardize the use of export path for all models so that users don't have to write their own one off scripts to load...

Needs formatting of code + some docstrings. Other stuff looks quite good.

Forgot to mention, lets add a directory for stateless transducers in conf, with subdirectory for Conformer. So that people can more easily find and train such models c