prodigy stuck - white "loading" screen

when running a prodigy ner.teach task, i get a white “loading” screen from time to time (after 10-20 labelings).
why is that happening?
when i look at the monitoring of the machine, i see that it does not even use half of the cpu/ memory

Probably because the ner model is training in the background and your inference queue is waiting for the model to finish training. I am not exactly sure if this is exactly the reason. Why don’t you try setting the prodigy log level to PRODIGY_LOGGING=basic or PRODIGY_LOGGING=verbose to see what’s happening

1 Like

ner.teach should update reasonably quickly, but yes, the loading interruptions can happen if updating takes longer or if Prodigy goes through the incoming raw data to find suitable examples to send out for annotation. If you’re annotating a less common label and the model doesn’t produce many suggestions, it may take a few batches to compile a full batch of suggestions to annotate.