elpis
elpis copied to clipboard
CoEDL
🙊 software for creating speech recognition models.
I have Samoan data that I have archived with PARADISEC. As per their data design, the files that belong to the same event have the same names - save the...
Include `nvidia-cuda-toolkit` and `gdown` to the Docker image, to help with debugging Elpis, and downloading data for CLI usage.
Currently, user has to edit the "create_dataset.py" CLI file to use a custom dataset. Would be nice if the options could be passed as args rather than having to edit...
Standardise in the docs - Consistent case for titles and headings - References to buttons in the docs. Currently some things use code markup, others use quotes.
Check if the machine has GPU. If not, make the default train settings lower. Set the `Number of epochs` to `1` and `Batch size` to ` 1`.
Likely that strings have been included in the latest PRs that aren't in the translator.
The image has the wrong port numbers https://elpis.readthedocs.io/en/latest/wiki/install-elpis-docker.html
In the uploading feature, when multiple models have been uploaded, only the first uploaded model is used for transcription. This happens when after uploading a model, the user returns to...
The Kaldi CLI fails when running 'transcribe.py' Traceback (most recent call last): File "/elpis/elpis/examples/cli/kaldi/transcribe.py", line 33, in t.prepare_audio(faudio) File "/elpis/elpis/engines/kaldi/objects/transcription.py", line 235, in prepare_audio self._process_audio_file(audio) File "/elpis/elpis/engines/kaldi/objects/transcription.py", line 63, in...
