Splicing / variadic symbolic expressions

Would it be possible to make the following code snippet work?

import torch
from beartype import beartype
from jaxtyping import Float, jaxtyped
from torch import Tensor

class A:
    def __init__(self, shape: tuple[int, ...]):
        self.shape = shape

    @jaxtyped(typechecker=beartype)
    def forward(self, x: Float[Tensor, "... {self.shape}"]) -> Float[Tensor, "..."]:
        return x.flatten(start_dim=-len(self.shape)).sum(dim=-1)

a = A((3, 10, 5))
x = torch.randn((7, 3, 4, 5))
print(a.forward(x))

At the moment it does not work as far as I can tell, because {self.shape} is only matched against a single dimension of x. Is there a way to evaluate the expression and splice in the tuple value into the type before the type gets matched against the dimensions? Maybe with something like a *{self.shape} syntax?

Start by running the provided Python example with jaxtyping, beartype, and torch to reproduce how {self.shape} is matched. Trace the shape-expression parsing and matching entry points; done means tuple-valued instance attributes can be spliced into symbolic expressions and the example validates with the intended output shape.