[CharlelieLrt]

Description

This PR adds CheckpointStateBundle so users can restart distillation with
their named models, optimizers, loss state, and optional random state.
The bundle composes the existing PhysicsNeMo checkpoint utilities without
changing their public API or saving private FastGen adapters:

bundle = CheckpointStateBundle(
    {"student": student, "fake_score": fake_score},
    optimizers={"student": student_optimizer, "fake_score": fake_optimizer},
    stateful={"loss": loss_fn},
)
bundle.save_checkpoint("checkpoints", epoch=20, include_rng=True)
bundle.load_checkpoint("checkpoints", epoch=20, restore_rng=True)

Users recreate matching components before loading and save at completed
optimizer updates. Distributed callers invoke the bundle on every rank and
restart with the same topology. The optional strategy in the next PR uses this
same bundle.

Checklist

• This PR follows the contribution guidelines.
• The documentation describes the API and its role in the stack.
• This PR remains a draft.

Dependencies

• Depends on

  • PhysicsNeMo #2097 extends the loss API to DMD2 and f-distill. This is the immediate parent branch.

• Earlier prerequisites

  • PhysicsNeMo #2094 adds named diffusion inputs and outputs.
  • PhysicsNeMo #2095 adds inverse SNR and TrigFlow components.
  • PhysicsNeMo #2096 adds private adapters and public sCM and MeanFlow losses.

• Depends on this PR

  • PhysicsNeMo #2099 adds the optional training strategy and the grouped changelog entry. It follows this PR directly in the stack.

• Parallel FastGen work

  • FastGen #43 allows the NumPy 2 dependency used by PhysicsNeMo. PhysicsNeMo Add sCM and MeanFlow distillation losses #2096 uses an immutable fork pin, so it does not need to wait for this proposal to merge.
  • FastGen #44 provides shared loss hooks for external training loops. The initial PhysicsNeMo integration does not require these hooks.

• Conditional follow-up

  • PhysicsNeMo #2100 replaces local compatibility code with the proposed FastGen loss hooks. It depends on PhysicsNeMo Add optional diffusion distillation strategy #2099 and FastGen [other] Add modulus.models.fcn_mip_plugin #44, not the other way around.

If FastGen maintainers decline FastGen #44, we would close PhysicsNeMo #2100.
The preceding PhysicsNeMo stack would remain available.

Review process

Reviewers can compare this PR with its immediate parent. After a parent merges,
we update the remaining branches and their PR bases to keep each diff focused.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[github-actions]

CODEOWNERS review map

Current for commit b19f3dce85e4. An approval covers every file listed for that owner; one owner is sufficient for shared files.

⏳ @CharlelieLrt — 4 file(s)

• docs/api/diffusion/distillation.rst
• physicsnemo/diffusion/distillation/init.py
• physicsnemo/diffusion/distillation/checkpoint.py
• test/diffusion/test_distillation_checkpoint.py

⏳ @megnvidia — 1 file(s)

• docs/api/diffusion/distillation.rst

No CODEOWNER

• CHANGELOG.md

Comment /codeowners-info to refresh.

[CharlelieLrt]

Continued in PhysicsNeMo #2203, from branch diffusion-checkpoint-bundle in CharlelieLrt/modulus. The replacement preserves this implementation and includes the requested documentation clarifications. Please continue the review there. This PR retains the earlier discussion.