All pull requests

Pull requests

• fix(qvae_cell): expand and deduplicate the --metrics selection

  #638·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• fix(dplm): reject empty aligned sets in comparison reports

  #636·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• fix(qvae_mnist): reject unsupported dataset names with the valid choices

  #634·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• fix(qvae_mnist): tolerate singleton classes in MLPClassifier.fit

  #632·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• fix(qdiffusion): validate EnergyModel visible-logit width

  #630·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• feat(example): demonstrate feature selection on a non-default axis

  #628·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• docs(theory): use #P-hard and semidefinite Hessian wording

  #626·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• fix(qvae): validate configured KL and weight-decay coefficients

  #624·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• docs(release-notes): correct the RISC-V glibc floor to 2.39

  #622·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• docs(theory): distinguish critical and trivial RG fixed points

  #620·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• feat(example): add a constrained Q-Diffusion generation example

  #612·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• fix(qvae): validate BM weight decay

  #610·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• docs(theory): fix math rendering and spacing on four theory pages

  #608·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• fix(dbn_digits): accept torch inputs in get_layer_activations

  #606·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• docs(background): use one unnormalized-density symbol in the EBM gradient derivation

  #605·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• docs(theory): fix the bias-free assumption and two overstated claims

  #602·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• fix(qdiffusion): validate EnergyModel BM dimensions

  #600·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• fix(bm_generation): validate cost weights

  #598·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• fix(rbm_digits): reconstruct from the input's hidden activations

  #596·zjncs opened Oct 20, 2026·

  ·
  1

• fix(gbrbm): discard exactly n_burnin initial Gibbs steps

  #595·zjncs opened Oct 20, 2026·

  ·
  1

• fix(gbrbm): apply the dtype argument and keep user-supplied weights

  #594·zjncs opened Oct 20, 2026·

  ·
  1

• docs(release-notes): record the #164 / #171 fixes merged since v0.3.0

  #593·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• docs(install): list the real package layout and class names

  #591·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• fix(qvae): reject non-finite dataset means

  #589·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1

• docs: use the real KPP class names in the layer table

  #587·iloveeyjafjalla opened Oct 20, 2026·

  ·
  1

  1