Somshubra Majumdar

@VahidooX can you change this line to read from the model_defaults (https://github.com/NVIDIA/NeMo/blob/main/examples/asr/conf/lstm/lstm_transducer_bpe.yaml#L117). @sarasesar did you train the model without modification to the config ? I just tried it locally for...

Hmm that's a problem. We'll look into what's going on.

@VahidooX can you take a look at the LSTM encoder based RNNT code ? The decoder rakes a new arg but checkpoint loading is affected c

@borisfom please review pr

Hey @Borda Thanks for the update ! I reviewed the Nemo config file in that pr, and mostly it seems great. The one thing popped up was the pytorch version...

One thing - Nemo has a monthly release cycle - and generally releases at the last day of the month (or the closest working day). There are exceptions but that's...

On the previous discussion of Numba support - would be beneficial if a conda env with Numba 0.54/0.55 were available to test a few things in Nemo such as RNNT...

Sounds good, I'll try setting those up on the ecosystem repo over the next week or two.

Sorry for delay, major PRs needed to be merged for release. Added two feature requests for numba and CI test against container, https://github.com/PyTorchLightning/ecosystem-ci/issues/38 and https://github.com/PyTorchLightning/ecosystem-ci/issues/39. The configs for nemo seem...

Final lr is approximately after 1/ gamma update steps have occurred. At this point, the clipping bounds are somewhat tight and cause the actual lr to fall close to the...