class torch.fx.experimental.symbolic_shapes.StrictMinMaxConstraint(warn_only, vr)[source]

For clients: the size at this dimension must be within ‘vr’ (which specifies a lower and upper bound, inclusive-inclusive) AND it must be non-negative and should not be 0 or 1 (but see NB below).

For backends: there must not be any guards on this dimension which are not implied by the given lower and upper bound. Regardless of the lower bound, the backend can assume the size is non-negative and that it is not 0 or 1.

An unbounded StrictMinMaxConstraint can be thought of as a strict version of “RelaxedUnspecConstraint”.

NB: Export will often unsoundly assume that a graph works for 0/1, even though at trace time we assumed size is not 0 or 1. The idea is that if we produce a graph that works for a range of values, it will be OK for N=0/1 too.


Format the constrain equation


Access comprehensive developer documentation for PyTorch

View Docs


Get in-depth tutorials for beginners and advanced developers

View Tutorials


Find development resources and get your questions answered

View Resources