Support MultiThreading and Multiprocessing

[palnabarun]

Hi @arshpreetsingh

Can you please elaborate on the issue?

[arshpreetsingh]

Sure, With each function there could be decorator something like where user can control number of threads for specified function, Something like:

@fire_resource(threads=4) # it could use Threading API Or may be joblib in Sklearn def function_name(): return "running multiple threads"

same if we have to run functions concurrently.
@fire_resource(threads=4,concurrent=True) # it could use Multiprocessing API def function_name(): return "running concurrent processes" I just found that when Training a ML function we obviously need multiple threads but when making predictions it's "Really Good" to run it concurrently. Please feel free to ping for more detailed elaboration.

[anandology]

I think this is a documentation issue. Firefly is a Python WSGI application. It can be run using multiple threads or processes using any WSGI server. The recommended approach is using gunicorn.

To run it using 4 processes:

gunicorn --workers 4 firefly.main.app -e FIREFLY_FUNCTIONS="funcs.square,funcs.cube"

To run using 4 threads:

gunicorn --workers 4 firefly.main.app -e FIREFLY_FUNCTIONS="funcs.square,funcs.cube"

@arshpreetsingh does it answer your question?

[arshpreetsingh]

Yes, Thanks!