text
text copied to clipboard
Text utilities, including beam search decoding, tokenizing, and more, built for use in Flashlight.
### Question I notice that CTC and ASG criterion are supported in the LexiconDecoderOptions, does it support RNN-T? how can I perform decoding for RNN-T models using flashlight? Thank you...
I am trying to decode a fine-tuned ASR model,fine-tuned using the [vakyansh](https://github.com/Open-Speech-EkStep/vakyansh-wav2vec2-experimentation) toolkit using the [this](https://github.com/Open-Speech-EkStep/vakyansh-wav2vec2-experimentation/blob/main/utils/inference/infer.py) script and the following command line:- python ../../utils/inference/infer.py ${data_path} --task audio_pretraining \ --nbest 1...
### Bug Description Hello, I'm trying to package this as a python package for NixOS, and I'm experiencing trouble due to the mixed cmake / `setup.py` layout. When I'm trying...
### Bug Description I tried building ASR systems on a very common standard task (LibriSpeech-100h) using the [torchaudio ctc decoder](https://pytorch.org/audio/stable/generated/torchaudio.models.decoder.ctc_decoder.html#torchaudio.models.decoder.ctc_decoder). This decoder uses the flashlight/text library as decoding backend. While...
Hello, I am trying to migrate my ASR model from OpenSeq2Seq decoder to Flashlight. Currently, I am using [Nemo Conformer large](https://github.com/NVIDIA/NeMo/blob/main/examples/asr/conf/conformer/conformer_ctc_bpe.yaml) as the acoustic model, which is tuned on my...
### Summary Adds diverse beam search from [this paper](https://arxiv.org/pdf/1610.02424.pdf). #### Helpful Diagrams Figure 1: Example of how diversity factor (Hamming) is applied at timestep *t* Figure 2: General algo ###...
### Feature Description Currently, the dictionary cannot handle duplicate entries. It would be interesting if this would be supported. Possibly a flag that allows one to 'allow' multiples would be...