Daniel Galvez
Daniel Galvez
Exact message: /install/miniconda3/envs/100k-hours-lingvo-3/lib/python3.7/importlib/_bootstrap.py:219: RuntimeWarning: numpy.ufunc size changed, may indicate binary incompatibility. Expected 192 from C header, got 216 from PyObject return f(*args, **kwds) I'm not sure what the cause is....
Kaldi has some existing systems for audio-based language ID (see the egs/lre* directories), but their training datasets are inaccessible. It is probably most straightforward to build one ourselves using the...
I would like to run that model on all of our audio files. Greg made an initial attempt here: https://github.com/greg-landing/yamnet I was told that there were a few problems: 1)...
[Webdataset](https://github.com/webdataset/webdataset) format is preferable for data distribution for a few reasons: - Easy to use without installing dependencies because it's just .tar files. - Natively supports the concept of sharding....
We would like to make smaller subsets of the data ourselves for the sake of downstream users. Thje msot obvious way to do this is to sort each aligned segment...
Right now, we timeout when an audio file fails to align with its transcript within 200 seconds: https://github.com/mlcommons/peoples-speech/pull/27/files#diff-b790cd27585332e1eeca7dab897f1ccd7bcd483181132bd9914f2dd07062534fR401 This means 10% of our files timeout during alignment. One observation is...
Now that we have several people developing, we should start thinking about adding this. I'm not sure how github does this actually. We can also consider adding a pre-commit git...
https://github.com/mlcommons/peoples-speech/tree/main/third_party
GreedyCTCInfer::forward already allowed for this, so they did not implement the exact same interface. Now, they do. Also warn about not passing in the decoder_lengths argument. It is likely an...
# What does this PR do ? I demonstrate, using transcribe_speech.py, that simply casting the entire model to bfloat16 gives about 15% higher performance than using automatic mixed precision. The...