Jan Vainer

Compression fails on non-contigous feats

2

Hi, I am getting a stride error when running compression on non-contiguous tensors: ``` lilcom: compression error: last stride should be 1, got 253 ``` The error message is a...

Mixed precision and warp-ctc

1

Hi, I am trying to use the CTCLoss from this repo with mixed precision from apex, but I am getting `RuntimeError: CUDA error: an illegal memory access was encountered`. Without...

online beam search inference

8

Are there any plans to support online inference with beam search? Usecase: Audio comes in n milisecond chunks and gets processed by an accoustic model that outputs a log_prob frame...

Past hypothesis rescoring based on future context with beam search

4

When I run online inference, it seems like once a part of a sentence is printed to the stdout, the part of the sentence never changes regardless of the following...

Bug: shift augmentation sometimes produces silent tensor

2

Hi, I was playing with the augmentations in this repo and I am sometimes getting silent audio results with the Shift augmentation. It happens shorter audio files with duration around...

Would be nice to be able to do ``` result = cuts1 - cuts2 ``` where the result is a subset of cuts1 without cuts with ids from cuts2. Likewise...

Allow argument dependencies

6

**Is your feature request related to a problem? Please describe.** I am wondering if its possible to tie some argument values together. For example, when I train an Encoder-Decoder architecture,...

Audio quality improvements

6

Hi, awesome contribution for TTS community :) I am wondering, did you manage to train a model that would have higher audio quality than the pretrained checkpoint provided with this...

How long does a training from scratch take?

2

I was wondering how long would it take to train the model from scratch? And on hoe many GPUs would it be? I could have 2 V100s available. How long...