List view

• Incorporate PyTorch 2 forward-mode autograd

  Remove all usage of `torch._vmap_internals._vmap` and replace with new tools that came out in PyTorch 2. Specifically, all Jacobian and Hessian computation should now use the `torch.func.jacfwd` utility for forward-mode automatic differentiation

  No due date

  •0/4 issues closed
  0% complete4 open0 closed