fix(flux): skip packing noise when FLUX Denoise continues from latents without add_noise

[gokay-ai]

Summary

Chaining two FLUX Denoise nodes where the second one continues from the first one's latents with add_noise=false (and no mask) crashed with Tensor type unknown to einops <class 'NoneType'>. _run_diffusion already sets noise = None in that case, but still called pack(noise). It now skips the pack when there is no noise, same as init_latents/inpaint_mask right above it and the FLUX.2 node.

Related Issues / Discussions

Fixes #9790

QA Instructions

Added test_flux_add_noise_false_denoises_incoming_latents_without_noise in tests/app/invocations/test_denoise_noise_inputs.py. It runs _run_diffusion with incoming latents, add_noise=False and a schedule with more than one step, with the model, LoRA patching and denoise mocked. The existing add_noise=False test returns early on the single-step short-circuit, so it never got to the pack.

• Before the fix: the new test fails with the same einops error as in the issue.
• After: pytest tests/app/invocations/test_denoise_noise_inputs.py tests/app/invocations/test_flux_denoise.py gives 47 passed, 5 xfailed, and pytest tests/app/invocations -k flux gives 330 passed, 4 skipped, 1 xfailed.
• ruff check / ruff format --check pass on both files.

CPU and mocks only. I didn't run the real two-stage FLUX workflow on a GPU.

Checklist

• The PR has a short but descriptive title, suitable for a changelog
• Meaningful regression coverage added / updated where needed; obsolete tests/code removed
• Persisted-state and API changes include required migrations / compatibility validation
• Relevant performance/efficiency opportunities considered; material claims have evidence
• Material review findings resolved and relevant checks rerun
• Documentation added / updated (if applicable)
• Updated What's New copy (if doing a release after this PR)

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[Pfannkuchensack]

Verified on GPU (RTX 4090, FLUX.1 dev): on main, a two-stage chain (stage 1 up to 0.6, stage 2 continuing from its latents with add_noise=false) fails with the einops NoneType error. With this PR it completes, and the result closely matches a single 20-step run (mean abs diff 0.29/255). FLUX.2, Z-Image and Anima already guard noise the same way, so FLUX.1 was the only affected node. The new test fails without the fix.

No changes requested, looks good to me.

[Pfannkuchensack]

Heads-up on merge order: this PR and #9837 both change the same import line in tests/app/invocations/test_denoise_noise_inputs.py (this one adds from contextlib import nullcontext, #9837 adds from contextlib import ExitStack, nullcontext). Whichever merges second will hit a small conflict there. The correct resolution is to keep from contextlib import ExitStack, nullcontext; the rest merges cleanly (checked locally with both branches at their current heads).