Issues

Issues

• LM reuses a stale loss when inputs, targets or parameters change between steps

  pypose/pypose#475 · opened on Nov 12, 2026

• LQR and MPC leave cost tensors behind during module dtype/device conversion

  pypose/pypose#473 · opened on Nov 12, 2026

• EPnP fails on broadcastable point, pixel, and calibration batches

  pypose/pypose#470 · opened on Nov 12, 2026

• SO3 right-Jacobian gradients become NaN at the identity

  pypose/pypose#469 · opened on Nov 12, 2026

• Triggs correction fails for the built-in Scale kernel's zero second derivative

  pypose/pypose#466 · opened on Nov 12, 2026

• MPC raises an ambiguous tensor truth-value error for multiple systems

  pypose/pypose#465 · opened on Nov 12, 2026

• Cholesky solver ignores failed factorizations and can corrupt LM parameters

  pypose/pypose#462 · opened on Nov 12, 2026

• ICP fails when initial transformations introduce a batch dimension

  pypose/pypose#461 · opened on Nov 12, 2026

• Batched CG returns NaNs when systems converge at different iterations

  pypose/pypose#459 · opened on Nov 12, 2026

• EKF covariance collapses to zero and stops subsequent measurement updates in float32

  pypose/pypose#457 · opened on Nov 12, 2026

• Feature request: optional trajectory alignment using positions and orientations

  pypose/pypose#454 · opened on Nov 4, 2026

• vec2skew is slow on small batches because it builds the matrix with four stacks

  pypose/pypose#453 · opened on Nov 1, 2026

• IMUPreintegrator covariance transition order differs from sequential updates

  pypose/pypose#449 · opened on Oct 31, 2026

• cumops fails for non-power-of-two sequence lengths

  pypose/pypose#444 · opened on Oct 27, 2026

• LM accepts non-finite trial losses and leaves parameters outside the model domain

  pypose/pypose#441 · opened on Oct 25, 2026

• Fix EKF automatic Jacobian batching and add a shared batched filter example

  pypose/pypose#435 · opened on Oct 16, 2026
  1

• Accelerate PyPose by replacing / add additional LieType implementation with NVIDIA Warp

  pypose/pypose#386 · opened on Dec 26, 2025
  1

• torch.optim.Adam does not work as intended with pp.Parameter

  pypose/pypose#377 · opened on Jul 20, 2025

• Request for Support of Similitude Group SL(4)

  pypose/pypose#376 · opened on May 22, 2025

• Try odeint instead of discrete time in state_transition.

  pypose/pypose#375 · opened on May 19, 2025

• Replace the dict to tensordict

  pypose/pypose#374 · opened on Mar 9, 2025

• 2D Case

  pypose/pypose#347 · opened on Jun 27, 2024
  4

• MPC examples do not work

  pypose/pypose#344 · opened on May 24, 2024

• Jacobian computed using autograd of MPC solver function inaccurate when returning 'u'

  pypose/pypose#340 · opened on Feb 28, 2024
  1

• Best way to incorporating Schur complement trick with pypose

  pypose/pypose#338 · opened on Feb 23, 2024
  1