Setting bounding box of product domain leads to "RuntimeError: Boolean value of Tensor with more than one value is ambiguous"

[CAarset]

First of all, thank you sincerely for this beautiful and easy-to-use physics-based neural network code; the ease of use for moving domains is particularly noteworthy.

I've ran into a minor yet breaking error when seeking to manually specify bounding boxes of product domains.

• Release: 1.0.6rc2
• Issue description: If you attempt to call the .bounding_box() of a ProductDomain after you've manually set the bounding box via .set_bounding_box, you will consistently throw the error

RuntimeError: Boolean value of Tensor with more than one value is ambiguous

This appears to be because of the line

if self.bounds: return self.bounds

in .bounding_box. If self.bounds was set by .set_bounding_box, this ends up evaluating the truth value of a torch tensor (self.bounds), which immediately causes the error. I suspect a suitable workaround would be changing the above code to

if self.bounds is not None: return self.bounds

since self.bounds is already pre-set to None in init.

• Minimal breaking example:

I = tp.spaces.R1('i')
interval = tp.domains.Interval(I,0,1)
box = interval * interval
print(box.bounding_box()) # Works correctly
box.set_bounding_box(box.bounding_box()) # In principle, a harmless overwrite
print(box.bounding_box()) # But now, throws an error

• Side note: The warning produced when the code is forced to estimate an unknown bounding box is misleading; it outputs

warnings.warn(
                f"""The bounding box of the ProductDomain dependens of the
                              values of domain_b. Therefor will sample
                              {N_APPROX_VOLUME} in domain_b, to compute a 
                              approixmation. If the bounds a known exactly, set 
                              them with .set_bounds()."""
            )

but the actual name of the method that sets bounds is .set_bounding_box, not .set_bounds. There are also typos in "dependens", "therefore,", "are known" and "an approixmation".

[TomF98]

Thank you for pointing this out and for suggesting a fix. This will be corrected in the next version, along with the incorrect documentation.