Progress bars: printing iterations/second and unbounded processes

[wbthomason]

Hi! I'm wondering if there's a way to use rich.progress to print information - specifically the number of iterations per second - for a process with an unknown number of steps. With tqdm, I do this as follows:

from tqdm import tqdm
progress = tqdm(desc='My process', unit='')
while not process_is_done():
  progress.update()

Is there a way to do this or something similar with rich? Apologies if this is a question already answered in the docs - I read through what seemed to be the relevant sections, but may have missed something.

[willmcgugan]

Yes, tasks have a speed attribute which is iterations per second.

Have a look at TransferSpeedColumn, which assumes your units are bytes of data, but you could modify that to display iterations per second.

Probably worth adding that to the core lib, so feel free to PR that.

[wbthomason]

Thanks! I'll try to put together a PR once I get things working.

Also, looking at https://github.com/willmcgugan/rich/blob/0e8f4747cae99d9ccb2596c966af5737e955a932/rich/progress.py#L440, I'm guessing that there is no way to specify a task with an unknown total bound, correct? (I assume that field would need to be Optional[float] rather than float if it was possible)

[aziz0x00]

Hey there, is this problem of "number of iterations per second" is resolved or a PR for it is ok?

[deo-nikita]

I'd love this feature too

[deo-nikita]

so I can estimate the speed (iterations per second) at which my deep learning model works, not only % finished and time remaining

[NichtJens]

@aziz0x00 and @deo-nikita Since I had a similar usecase, I just submitted #3533. Hope it will help you too (if/once it's accepted).