Tom Augspurger

Can you post the full traceback? Does `NeuralNetRegressor` implement `partial_fit`?

Thanks for the report. I think the root cause is that `iris_da` has some blocks with very few samples. ```python In [28]: iris_da.chunks Out[28]: ((2, 5, 1, 2, 4, 0,...

Ah, gotcha. In that case, I think it just comes down to how Python implements floating point arithmetic: ```python In [3]: 1 - 0.7 Out[3]: 0.30000000000000004 ``` I don't recall...

+1 to the general concept, and I think that dask-ml is a fine home for this utility to live.

Oh https://github.com/dask/dask-gateway/issues/255 does sound better... Probably just close this issue then? > would you be free to chat at some point to help debug what's going on? Most graceful uses...

I notice that the issue is in the (de)serailization of the Queue object. It looks like `distributed.Queue` defines how it pickles in a way that doesn't work with Dask Gateway...

Based on https://github.com/Azure/azure-sdk-for-python/issues/22073#issuecomment-991311505 isn't the right approach to replace https://github.com/dask/dask-cloudprovider/blob/main/dask_cloudprovider/azure/azurevm.py#L16 with the newer ``` from azure.identity import AzureCliCredential credential = AzureCliCredential() ``` possibly increasing the minimum required dependencies if necessary?

> I am not super-excited about checking in Parquet files to the repo (don't like binaries in GitHub), but it's the best way to keep the example simple. Let me...

Is it right to say that the parquet files need to be written with specific arguments (like partition_cols) for this performance benefit? If so I'd prefer to write them in...

Should be doable, as long as nothing in the .fit method is blocking on the main thread. > On Dec 24, 2021, at 10:03 AM, Javier Yepez ***@***.***> wrote: >...