[ET Device Support] CUDA-native Qwen 3.5 MoE inference with device tensor pipeline

[Gasoonjia]

Stack from ghstack (oldest at bottom):

• -> [ET Device Support] CUDA-native Qwen 3.5 MoE inference with device tensor pipeline #18788
• Turn on device memory planing as default #20214
• [ET Device Support] Add device tensor helper functions to TensorPtr API #19828
• [ET Device Support] make device support config method-based #19582
• [ET Device Support] Add ExecutorchBackendConfig flags for skipping H2D/D2H copies #18760
• [ET Device Support] PropagateDevicePass inserts H2D/D2H copy ops at delegate boundaries #18730
• [ET Device Support] Add device tensor helper functions to TensorPtr API #18761

Integrate the ET device tensor pipeline into the Qwen 3.5 MoE model to
eliminate unnecessary H2D/D2H copies during inference.

• Export: Multi-method export (forward + sample) with device memory
  planning enabled and method-level H2D/D2H skipping.
• Runner: Custom CUDA-native inference loop that keeps logits on GPU
  between forward and sample, reuses CUDA tensors across iterations,
  and only copies the 8-byte token ID back to CPU for EOS checking.

Differential Revision: D100133933

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/18788

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❌ 3 New Failures, 1 Unrelated Failure, 1 Unclassified Failure

As of commit e6785b1 with merge base b47e588 ():

NEW FAILURES - The following jobs have failed:

• pull / test-multimodal-linux (gemma3-4b) / linux-job (gh)
  RuntimeError: Command docker exec -t 02abc3b32a32bf514442b376355a91830ff32d1c01af35a0d5dd0b75443bb8cc /exec failed with exit code 139
• pull / test-samsung-quantmodels-linux / linux-job (gh)
  RuntimeError: Command docker exec -t 381715f2a6151dbd5ea99067bdedc5400aa6d7d256978b1e2917ca4908c2c53f /exec failed with exit code 1
• Test CUDA Windows Export and E2E / test-model-cuda-windows-e2e (facebook, dinov2-small-imagenet1k-1-layer, non-quantized) / windows-job (gh)
  Process completed with exit code 1.

UNCLASSIFIED FAILURE - DrCI could not classify the following job because the workflow did not run on the merge base. The failure may be pre-existing on trunk or introduced by this PR:

• Test CUDA Builds / unittest-cuda / linux-job (gh) (this job did not run on the merge base, so DrCI cannot tell whether the failure is pre-existing)
  examples/models/qwen3_5_moe/test_sampler.py::TestSampler::test_output_shape_and_dtype

FLAKY - The following job failed but was likely due to flakiness present on trunk:

• Test Metal Backend / test-model-metal-e2e (mistralai, Voxtral-Mini-4B-Realtime-2602, quantized-int4-metal) / macos-job (gh) (detected as infra flaky with no log or failing log classifier)

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[github-actions]

This PR needs a release notes: label

If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with release notes:. This helps us keep track and include your important work in the next release notes.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "release notes: none"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.