d6tdev

Same issue. This works `shap.plots._waterfall.waterfall_legacy(explainer.expected_value, shap_values[0])`

[d6tpipe](https://github.com/d6t/d6tpipe) supports that see `pipe.pull()` examples https://github.com/d6t/d6tpipe#example-usage and https://stackoverflow.com/questions/50100221/download-file-from-aws-s3-using-python/54622791#54622791

It's not just about the format, it appears missing values were filled so be good to know what preprocessing was done. Seems #158 is also asking about the missing values

This is an important but also somewhat involved topic. Can you pick a time to discuss in more detail? https://calendly.com/databolt/30min

Can you take a look at this and tell us if this works? https://d6tflow.readthedocs.io/en/latest/advtasksdyn.html

How about asking the user if you want to run the task if it's not complete on load. Not sure if it should auto run. For now suggest to always...

There a multiple issues that make this non-trivial to implement. 1) a target doesn't know which tasks it belongs to and 2) this could work better in `Task().outputLoad(auto_run=True)` but would...

Thanks for raising the issue, we'll need to take a look. You can [hide execution output](https://d6tflow.readthedocs.io/en/latest/run.html#hiding-execution-output) but if there is an error it will still show.

@marcietran since the tasks gets executed in workers it wasn't as straight forward to reraise the original error, we're still looking into it.

Can you add more detail on why it didnt succeed and why you can't use it? Eg commands you ran and error log