Open thvasilo opened 4 years ago
You can disable the timeout with --VoilaExecutor.timeout=None
(see voila --help-all
).
There is no reason for a cell to take more time to execute in Voila than in a notebook.
--VoilaExecutor.timeout=<Int>
Default: None
The time to wait (in seconds) for output from executions. If a cell
execution takes longer, a TimeoutError is raised.
`None` or `-1` will disable the timeout. If `timeout_func` is set, it
overrides `timeout`.
Hello @davidbrochart
Thanks, couldn't find that option in the version I'm using (0.1.21) but found --VoilaExecutePreprocessor.timeout
. However, setting that to None or -1 as the helptext suggested to disable it led to errors:
File "/Users/thvasilo/opt/anaconda3/lib/python3.7/site-packages/jinja2/runtime.py", line 430, in __init__
self._after = self._safe_next()
File "/Users/thvasilo/opt/anaconda3/lib/python3.7/site-packages/jinja2/runtime.py", line 450, in _safe_next
return next(self._iterator)
File "/Users/thvasilo/opt/anaconda3/lib/python3.7/site-packages/voila/handler.py", line 145, in _jinja_cell_generator
res = ep.preprocess_cell(cell, resources, cell_idx, store_history=False)
File "/Users/thvasilo/opt/anaconda3/lib/python3.7/site-packages/voila/execute.py", line 156, in preprocess_cell
result = super(VoilaExecutePreprocessor, self).preprocess_cell(cell, resources, cell_index)
File "/Users/thvasilo/opt/anaconda3/lib/python3.7/site-packages/nbconvert/preprocessors/execute.py", line 438, in preprocess_cell
reply, outputs = self.run_cell(cell, cell_index, store_history)
File "/Users/thvasilo/opt/anaconda3/lib/python3.7/site-packages/voila/execute.py", line 289, in run_cell
timeout = min(1, deadline - monotonic())
TypeError: unsupported operand type(s) for -: 'NoneType' and 'float'
What I had to do to get it working was set the above to a high number, say 200 to avoid the timeout.
You're right, I looked at a future version of Voila, so the option is currently --VoilaExecutePreprocessor.timeout
. There might be a bug, so for now setting it to a high enough value is a good workaround.
I'm trying to use Voila in combination with torchtext for some NLP visualizations.
At one point I use torchtext to load a dataset, that takes a few seconds, maybe 10 seconds to load the data, then 10-20sec to create a vocabulary:
What I'm seeing when I launch the notebook through the
voila
command is aTimeoutError: Cell execution timed out
.I have two questions then: What is causing the cell to take much longer to run within Voila vs. when I'm executing it in Jupyter, and if there's a way to increase the timeout so long-running cells have time to complete.