Tatiana Likhomanenko

Still for me your example is not clear. Do you have an idea of interfaces/functions in c++ and python which you consider?

@dmzubr What is the problem then just to write simple wrapper over https://github.com/facebookresearch/wav2letter/wiki/Inference-Run-Examples "How we can I use this from Python"? Models will be anyway in binary C++ format, so...

@dmzubr thanks for clarification! I agree on the point of lowering an entry threshold to use framework. However, there should be the balance between simplicity for people to use and...

@dmzubr Thanks for your comments and sharing thoughts and experience! Yep, let's create another issue, so other could provide their input / API suggestions on this too, so we could...

Hey @light42! For transitions it is expected some list, it could be empty. Mostly this parameter corresponds to the ASG trained models where you have trained transition matrix between each...

Yes, you need to add into lexicon mapping `{laugh} {laugh}` so that it will map these words into the token itself and not using anything else. In list file you...

Hi @AndroYD84, > Is this model as accurate as your new [SOTA 2019](https://github.com/facebookresearch/wav2letter/tree/master/recipes/models/sota/2019) model? I jumped to the conclusion that I was still using an older model because I got...

@AndroYD84, To compare with ESPnet (if you don't care about speed for now) you should use model [Transformer S2S on Librivox](https://dl.fbaipublicfiles.com/wav2letter/sota/2019/librivox/models/am/am_transformer_s2s_librivox_dev_other.bin) without any beam-search decoder or with [ngram](https://github.com/facebookresearch/wav2letter/blob/master/recipes/models/sota/2019/librivox/decode_transformer_s2s_ngram_other.cfg) (convlm will...

@AndroYD84, could you send the final WER for comparison on your dataset ESPnet, transformer and tds-inference models? Also I noticed that we haven't some words in the lexicon while ESPnet...

> Is it also possible to use pre-trained models in `recipes` using **simple_streaming_asr_example** instead of the model in S3 bucket mentioned [here](https://github.com/facebookresearch/wav2letter/issues/485#issuecomment-576768533)? @avidov @vineelpratap?