CoEDL / elpis_next

A simple transcription workflow GUI for linguists and data scientists.
https://elpis-next.readthedocs.io/en/latest/
0 stars 0 forks source link
automatic-speech-recognition huggingface-transformers linguistics machine-learning

Elpis Next

This software is currently in development. Stay tuned for a stable release.

This is a slimmer rewrite of the elpis project.

Running the client and server

To run with docker:

This will spawn a client service at https://localhost:3000 and the server at https://localhost:5001. It will also create a data directory at ./data in which to store the state of the application. This data directory is bound to the host machine, so you can inspect what's going on in the application without needing to open another terminal session inside the container.

Troubleshooting

No space left on device (Docker)

If the error persists and you're trying to combine Docker Desktop with the WSL:

Cannot connect to the docker daemon
Error getting credentials (running docker compose build)

Make sure you're logged in with credentials:

Still an issue?

docker compose up runs but doesn't set anything up at localhost:3000

Remove these lines from ./compose.yaml:

    deploy:
      resources:
        limits:
          memory: 8gb

To run a development build locally (without docker):

In another window:

Note: For the server to work properly during training and transcription, you might need to install some additional local dependencies:

Troubleshooting

Troubleshooting libsndfile errors

If you get an error message on Mac OS that reads along the lines of OSError: cannot load library [...] libsndfile.dylib' (no such file) try this:

brew install libsndfile
...

brew list libsndfile | grep dylib
/opt/homebrew/Cellar/libsndfile/1.1.0/lib/libsndfile.1.dylib
/opt/homebrew/Cellar/libsndfile/1.1.0/lib/libsndfile.dylib

ls -l /opt/homebrew/Cellar/libsndfile/1.1.0/lib/
total 912
-r--r--r--  1 raf  admin  466656 Jun 12 14:42 libsndfile.1.dylib
lrwxr-xr-x  1 raf  admin      18 Mar 27 14:42 libsndfile.dylib -> libsndfile.1.dylib
drwxr-xr-x  3 raf  admin      96 Jun 12 14:42 pkgconfig

SF_PATH=$(poetry env info -p)/lib/python3.10/site-packages/_soundfile_data/

mkdir -p $SF_PATH

ln -s $(brew list libsndfile | grep dylib | head -n 1) $SF_PATH/libsndfile.dylib

poetry shell

python3
>>> import soundfile
>>>
No module named 'poetry.repositories.http_repository'

Install poetry

ERROR: There are no scenarios; must have at least one. (Yarn)

You've got the wrong yarn (cmdtest package)---make sure it's through npm install --global yarn

Note: restart the terminal after this if it's still giving you grief

Firefox can't establish a connection to the server at ws://localhost:3000/_next/webpack-hmr

Try with a different browser

Usage

Visit localhost:3000 and begin your transcription workflow!

Creating a dataset

(It may take a little while to process)

Training a model

Transcribing audio

Runtime Troubleshooting

ffmpeg was not found but is required to load audio files from filename

Ensure ffmpeg is installed in your environment; restart and try again