Inquire about introducing checkpoint while running BirdNET

[prayagnshah]

Hi Team,

I am a computer student working with Environment and Climate Change Canada and lately, we have observed that when running BirdNET on a large number of audio files, such as 100 or 200 hours, if the computer unexpectedly shuts down, we are unable to determine which files were successfully analyzed and which were in progress.

How about introducing an argument named --checkpoint-frequency ? If the user uses this flag with an integer value, a checkpoint will be created, and upon successful completion, the checkpoint will be automatically deleted. For example, you can specify --checkpoint-frequency 40, which means a checkpoint will be created after every 40 audio files have been processed.

Please let us know if this something I can contribute it or is not necessary on your end?

Regards,

Prayag Shah

[tgruetzm]

There's another simple way to implement this that I use in my personal repo. I only process files when the result file doesn't already exist for that file. If there's a failure, I can re-run on any files that were not processed.