Closed ahmadia closed 7 years ago
@mrocklin / @quasiben - I'd prefer one of you take a look at this and the discussion on the Slack tutorial channel before I merge.
I suspect that there was a reason why we chose to use the conda-forge channel. One thing that comes to mind is pyspark, though that probably won't conflict with defaults just on its own.
What's our next step? I can try to either see if the tutorial works off the versions of packages available in the default channel or I can try figuring out what's going wrong in conda-forge. I suppose it would be best if we could determine the problem in conda-forge and land a patch. I think it may be something like an incompatibility between the latest versions of pandas
and cloudpickle
.
Maybe I could try just pinning to earlier versions of pandas and cloudpickles...
Pinning cloudpickle to 0.2.2 resolves the issue. I'm going to close this and work from the issue for now.
Temporarily fixes https://github.com/pydata/parallel-tutorial/issues/20 while we sort out issues with the conda-forge channel.