-
Hi, I'm trying to train my Kaldi with corpus mini-librispeech.
I run kaldi on the latest version Kaldi Docker.
I start the docker with command
```
nvidia-docker run -it kaldiasr/kaldi:gpu-latest…
-
In OpenBLAS 0.3.21, on mips64, the Debian package fails to build because there are test failures in the utests for `dnrm2_tiny` and `dnrm2_inf`:
```
TEST 32/37 dnrm2:dnrm2_tiny [FAIL]
ERR: test_d…
-
Hello!
I would like to train your tuda-de model with the current common voice dataset 7.0 with 965 hrs of German text. Can I bring it in the kaldi formats (scp,...), split it in train, test, dev and…
-
Hello,
I am trying to do a domain adaptation for an existing model in a STT context by only providing a text corpus (no speech data). All the tutorials I can find teach how to create a new model fr…
-
Start off with `nemo:22.01` container, install `riva_quickstart_v2.0.0` (`riva_api` and `nemo2riva`), download `stt_en_conformer_ctc_large_1.6.0` from ngc and unzip it.
Running `nemo2riva --out st…
-
I tried to compile kaldi for windows using the guide [windows/INSTALL.md](https://github.com/roscisz/kaldi-mpi/blob/master/windows/INSTALL.md).
At the last step I ran into the issue: `LINK : fatal er…
-
# 🚀 Feature request
Kaldi is a popular open source speech recognition system that manages audio in a simple but unique format. It would be great to have some addition to the datasets object that wo…
-
Vosk, an open-source realtime ASR system based on Kaldi, has Ukrainian models available:
https://alphacephei.com/vosk/models
-
Hi, I've been trying to transform back the input of the encoder (attention location-aware mechanism) to a wav file using the function convert_fbank_to_wav.py. Just to be sure that I understand how it …
-
Hi, thank you for your work.
I import your code as ctc-crf package into my DL model.
In my training code, I calculate loss with
`loss = CRFLoss(logits, labels, inputLenBatch, input_lengths)`
I…