Hello,
I have been experimenting with ML stuff with gaffer, motivated by the new gafferML nodes.
(The example works well for me on windows)
I am thinking that allot can be done just using gaffer to automate things using the standard gaffer nodes. I linked up a large language model to a box via a python expression:
However these things can be very slow to run, and it makes sense to me for them to be task nodes instead to more easily control the execution.
I tried to make the same setup with a pythoncommand node, but it is not clear to me how to pass the output variable back to other nodes in gaffer.
I could write the output to a file - and this would probably be required if I was submitting to a farm, but if everything will run locally I should be able to pass the variables more directly I think.
I guess this is possible but I couldn't get anything to work.
I couldn't work out how to do it though - perhaps it is trivially easy
I think this will be a situation with allot of ML stuff in gaffer, especially when running on older hardware.