Max Balandat
Max Balandat
So this sounds very much like a contextual problem; let's see if I we can make that a bit more concrete (and mathy). The way I understand this is that...
It returns the standard deviation. Ax is a bit sloppy about terminology in some places, largely b/c historically it has been used a lot in the context of field experiments...
The default in the service API is actually to use EI with sequential conditioning rather than solving the full `d * q` (where `q` is the number of trials) dimensional...
@saitcakmak looks like [D37162339](https://www.internalfb.com/diff/D37162339) landed - did the bot fail to close this?
What exactly is the „saasbo_nehvi.py“ code you are running? Can you share a reproducible example? @dme65, @sdaulton
I tried this but unfortunately the GPU ran out of memory after ~45 iterations. SAASBO + MOO can be quite expensive, especially if the Pareto Frontier of the function consists...
In general, you should get away with relatively few Sobol points. This is really just an initialization and mostly in place to prime the model to be able to fit...
Not sure I fully understand the issue here. Are you trying to train a multiclass classification model and optimize the hyperparameters for the training job? Can you provide some more...
Hi, so unfortunately I think in the current setup this might be quite challenging to do. The original design of Ax was, as you observed, to perform all data and...
> What's the model PI? A typo :) Should be API. You probably want to wait until #967 goes in / start on top of that, unless you really like...