Closed krystophny closed 3 years ago
On my machine import profit
takes about 1s
Using tuna the main culprits are identified as:
numpy = 107 ms
sklearn = 140 ms
paramz (GPy) = 130 ms
pylab (GPy) = 217 ms
IPython (GPy) = 231 ms
Improvement by about 50% by moving GPy
in profit.sur.gaussian_process
Rykath/profit/speed
Another possibility for improving the speed of a small simulation (e.g. mockup
) would be an InternalRunner
which directly runs the simulation (without spawning a subprocess for the Worker
).
Should I implement that?
Yes! This would also be very important for testing Monte Carlo methods runs with many points quickly. There is also a way to call Python functions directly as a simulation without starting any external process, right?
At the moment,
profit
and related scripts take too long to start. This should be fixed by cleverer imports in__init__.py
and possibly other methods.