does econml support distributing train(parallel) in multiple boxes(just like lightgbm)? if not, any plan for this?

[ysong2123]

[kbattocchi]

We do not currently have any plans to support this, but it's on our radar. Could you expand on your use case for this a bit? It looks like LightGBM supports a few different modes of distribution, depending on whether you have a lot of features or a lot of data, or both - would one of those scenarios be particularly important to you? Likewise, it looks like LightGBM supports a number of different distributed computing platforms, although only Dask is directly maintained by the LightGBM team - would Dask support be sufficient, or would you need support on a different platform?

[nmriabov]

Hi there, I'm also interested in this, and for my use case, Dask support would be sufficient.