[Bug] AttributeError: 'SimpleCustomGP' object has no attribute 'num_outputs'

[aleczhanshi]

🐛 Bug

As I was trying to run the first code example in the tutorial (using botorch with ax), I got an error AttributeError: 'SimpleCustomGP' object has no attribute 'num_outputs'. Could anyone explain to me why this would happen?

Here's my versions of dependencies:
python 3.7.4
gpytorch 1.0.4
scipy

** Stack trace/error message **
Running optimization batch 1/5...
Traceback (most recent call last):
File "/home/aleczshi/anaconda3/lib/python3.7/site-packages/gpytorch/module.py", line 342, in getattr
return super().getattribute(name)
AttributeError: 'SimpleCustomGP' object has no attribute '_num_outputs'
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/home/aleczshi/anaconda3/lib/python3.7/site-packages/gpytorch/module.py", line 342, in getattr
return super().getattribute(name)
File "/home/aleczshi/anaconda3/lib/python3.7/site-packages/botorch/models/gpytorch.py", line 102, in num_outputs
return self._num_outputs
File "/home/aleczshi/anaconda3/lib/python3.7/site-packages/gpytorch/module.py", line 344, in getattr
raise e
File "/home/aleczshi/anaconda3/lib/python3.7/site-packages/gpytorch/module.py", line 339, in getattr
return super().getattr(name)
File "/home/aleczshi/anaconda3/lib/python3.7/site-packages/torch/nn/modules/module.py", line 576, in getattr
type(self).name, name))
AttributeError: 'SimpleCustomGP' object has no attribute '_num_outputs'
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "bo_ax.py", line 84, in
batch = exp.new_trial(generator_run=model.gen(1))
File "/home/aleczshi/anaconda3/lib/python3.7/site-packages/ax/modelbridge/base.py", line 609, in gen
model_gen_options=model_gen_options,
File "/home/aleczshi/anaconda3/lib/python3.7/site-packages/ax/modelbridge/array.py", line 212, in _gen
target_fidelities=target_fidelities,
File "/home/aleczshi/anaconda3/lib/python3.7/site-packages/ax/modelbridge/torch.py", line 214, in _model_gen
target_fidelities=target_fidelities,
File "/home/aleczshi/anaconda3/lib/python3.7/site-packages/ax/models/torch/botorch.py", line 334, in gen
outcome_constraints=outcome_constraints,
File "/home/aleczshi/anaconda3/lib/python3.7/site-packages/ax/models/torch/utils.py", line 178, in subset_model
if len(idcs) == model.num_outputs:
File "/home/aleczshi/anaconda3/lib/python3.7/site-packages/gpytorch/module.py", line 344, in getattr
raise e
File "/home/aleczshi/anaconda3/lib/python3.7/site-packages/gpytorch/module.py", line 339, in getattr
return super().getattr(name)

[Balandat]

Thanks for flagging this, we added the num_outputs property to the model API (see here) in order to reduce ambiguity / improve error messages, must have missed updating the tutorial for this change. I'll put up a fix shortly.

Since this is a GPyTorchModel with a single output, all you should need to do to fix this is give the model a _num_outputs attribute, i.e.

class SimpleCustomGP(ExactGP, GPyTorchModel):
    
    _num_outputs = 1

    def __init__(self, train_X, train_Y):
        [...]

[Balandat]

Fixed by #355.

Website is updated for the "latest": https://botorch.org/v/latest/tutorials/custom_botorch_model_in_ax