Tatiana Likhomanenko

You need to comment this if block too https://github.com/facebookresearch/wav2letter/blob/v0.2/inference/inference/decoder/Decoder.cpp#L74 because right now it creates the trie and here if trie is non none https://github.com/facebookresearch/wav2letter/blob/v0.2/inference/inference/decoder/Decoder.cpp#L97 it will run lexicon-based decoding.

Before you said that you created the dummy lexicon. So first please test with hack when you provide this mapping in the kenlm. I suggested two ways of doing this:...

Seems post processing of transcription is not fully correct (like you have still _ and not merged tokens) with lexicon free cc @xuqiantong

Could you send your train config? How many GPUs do you use to run? What is the dataset size?

your learning rate is too small for sgd (from my experience), try larger. About the corruption - try to run on, say, 1000 samples, probably the problem in the validation...

@viig99 the gamma is not used here (by default it is 1), so no lr decaying happening here.

@Bernardo-Favoreto what exactly error do you have? Could you post here your log?

You have several options here: - use the same token set and just extend lexicon file with spelling for your words. Here you need to apply sentencepiece model to your...

For force alignment and audio split you also could have a look here https://github.com/facebookresearch/wav2letter/tree/master/tools at voice activity detection (it is similar to librilight one we used). Regarding your questions: >...

You can start from our sota/2019 models to see how with the best pre-trained models you can transfer and then switch to inference model if you need it in real...