Some of my algos get this apparently common timeout error when I make a small change to them:
TimeoutException: Call to before_trading_start timed out
I know how to put timing statements around blocks of code to see what is taking the most time,
but the logging limits of like a few dozen lines cause this standard debugging method to be unusable,
since the timeout often occurs later in the backtest.
Is there any way to debug the timeouts without normal console log statements being available? At this point,
it seems like a trial and error guessing game. I had tried one of the machine learning notebooks, but how
can one do advanced machine learning development in this time-limited / logging-limited environment?
It seems that the example ML notebooks are right at the edge of timing out, and small additions cause
them to timeout.
Is there any upgrade option planned so that you can have access to more CPU/wallclock time and/or better logging?
Can we know if the timeouts are defined by CPU time or wall clock time? (If wallclock, time timeouts may
be due to data feed / pipelining delays?) If we knew this, it may help narrow things down. I had actually backed
out the machine learning models and put in linear models, so I doubt if the timeouts are due to long linear model solve times.
