fix(train): use gpu type arg for workers; group tail_lines logs

[akhileshh]

No description provided.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 100.00%. Comparing base (a6bcbae) to head (79592ea).
Report is 2 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main      #918   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files          142       145    +3     
  Lines         6113      6129   +16     
=========================================
+ Hits          6113      6129   +16     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[trivoldus28]

Isn't there supposed to be a final logger.info("\n".join(result)) at the end for when len(log_stream) is not evenly divisible by 8?

[trivoldus28]

Or does tail lines always give you what you asked, but then there's possibly a long delay between the actual output and when you see it.

[trivoldus28]

I'd set the default to None for now I think.

edit1: oh nvm I meant for training.py but I see you removed it already

edit2: ah but the default = 8 here would still enable it.

[akhileshh]

Or does tail lines always give you what you asked, but then there's possibly a long delay between the actual output and when you see it.

Yeah its a blocking call that just polls for logs. I have now made it a param to training so user can decide.