Is backtesting delayed on "requesting backtest" for others?

My other projects run fine, just seems to be this one that gets stuck on “Requesting Backtest”.  It was stuck for at least 8 hours today and never moved to the “Analyzing Strategy” step.  The main.py is relatively small (7107 bytes) but it loads a variable from a separate python file, and that file is 63744 bytes.

Hi Jesse Pangburn,
Can you share a backtest for which the issue is occurring to try to reproduce the issue?
Best,
Varad Kabade

Hi Varad,

Sure, thanks for trying to help!  Mostly they don't seem to ever finish and I have to just delete them to kill the backtest and try another, but the one I attached eventually completed successfully.

I tried reducing the data file size to ~38k today, and started a backtest ~10 minutes ago and it's still running.  So not entirely sure that's the problem.  I think I'll try putting the data file on the web and loading it via URL (not sure if that works on QuantConnect, we'll see).  As it is, building takes a little while in the web IDE.

For anyone who runs into this same problem, I fixed this by putting the data file I needed at a publicly available URL (used Dropbox but any public URL will do) and using the QCAlgorithm.Download method which is documented at https://www.quantconnect.com/docs/algorithm-reference/importing-custom-data#Importing-Custom-Data-Loading-Reference-Data 

In case it helps, that method just downloads strings so you could use a CSV file but I was already using pickle.  If you need pickle instead of CSV, then do this with your exported data wherever you're creating it:

import base64
base64_encoded = base64.b64encode(pickle.dumps(the_object_you_want_to_dump))
with open(f'your_filename.b64', 'wb') as save_file:
    save_file.write(base64_encoded)

Put that file at a place with a public URL.  In your backtest code, you want to download that file and unpickle it.  You can do that like this where “qcAlgorithm” is a variable of type QCAlgorithm (self in your backtest code usually):

download_base64 = qcAlgorithm.Download("https://www.dropbox.com/your_filename.b64?dl=1")
decoded_bytes = base64.b64decode(download_base64)
your_object = pickle.loads(decoded_bytes)

If you use Dropbox, make sure you change the URL from dl=0 to dl=1 at the end.

After I did this, the project backtests rapidly as usual.  Seems like having a large code file is bad news for backtesting, so don't encode reference data in a python file like I did.  Do it this way instead (or with a CSV).