Add initial Docker container #9
Conversation
edsu
force-pushed
the
docker
branch
4 times, most recently
from
d99c759 to
d9d3b83
Compare
edsu
force-pushed
the
docker
branch
2 times, most recently
from
a45b7d1 to
6f60075
Compare
This commit adds an initial Whisper Docker container, along with program run.py that pulls job files and media from an "todo" AWS SQS and S3 bucket respectively, and writes the Whisper output back to the bucket while placing a "done" message in another queue. See README.md for the details.
There was a problem hiding this comment.
a bunch of small questions and suggestions. the suggested docker command didn't work for me (it hung without processing any jobs or producing any output -- idk if this was a result of the env var stuff, didn't look into the issue much). but just running python speech_to_text.py --create test_video.mp4 and python speech_to_text.py to grab that one job worked for me, doing the processing directly on my laptop instead of the container, of course. i pulled the output from the S3 bucket, and it looked good! i did also add SPEECH_TO_TEXT_TODO_SQS_QUEUE and SPEECH_TO_TEXT_DONE_SQS_QUEUE values to my .env.
i think i'd be fine to merge this as-is and do follow up changes, since this is still at a pretty early stage, but fine with whatever.
- tightened up documentation - removed the secret environment files from getting cooked into the image! - added a --receive option to fetch messages off the DONE queue.
This commit adds an initial Whisper Docker container, along with a program that pulls jobs from SQS, media from S3, and then pushes results back to S3 and SQS. There are more details in the README about how it works, how to run an example, and tests.
Closes #3