Playground v2.5 + SDXL Lightning: add PCC-gated nightly + demo scripts

[kamalrajkannan78]

Ticket

• closes Model Bringup: SDXL Lightning #4691
• closes Model Bringup: Playground v2.5 #4692

Problem description

• E2E pipeline tests for Playground v2.5 and SDXL-Lightning had no correctness gate — they ran the full pipeline on TT but never compared outputs against a CPU reference, so PCC drops on any component (text encoders, UNet, VAE) were silently ignored.
• No standalone Python demo scripts exist for users to run these models end-to-end.
• Nightly tests included post-processing (image save to PNG) that belongs in a demo, not a correctness gate.
• Redundant per-component tests exist as the only source of numerical checks — kept alive purely because the e2e nightly lacked PCC comparisons.

What's changed

Split responsibilities into three: correctness gate, demo, and cleanup.

• PCC-gated nightlies (tests/torch/models/{sdxl_lightning,playground_v2_5}/test_*_pipeline.py): after every TT component forward, the same input tensors are fed to a lazy-loaded CPU twin and PCC is compared inline. Any component's PCC below 0.99 fails the test immediately with a clear message. UNet asserts per denoising step (fail-fast on the first bad step). No more PNG saving in the test — correctness is measured against tensors, not files.
• Demo scripts (examples/pytorch/sdxl_lightning.py, examples/pytorch/playground_v2_5.py): copies of the pipeline classes without pytest / PCC / assertions, with a __main__ block that saves an output PNG. Meant for humans to python examples/pytorch/<model>.py and see the pipeline produce an image.
• Component test removal: dropped 8 redundant per-component tests (test_text_encoder.py, test_text_encoder_2.py, test_unet.py, test_vae_decoder.py) across both model directories. Their coverage is now subsumed by the per-component PCC assertions in the nightly pipeline test.

Impact: any component-level numerical regression on TT is now caught in nightly with a clear per-component PCC log and an assertion. Fewer test files to maintain, one canonical demo path per model.

Checklist

• Verify the changes local testing/ Run test single in both n150 & p150/p100

Logs

• Playground-2.5

  • Demo(n150 & p100) - playgroundv2_5_demo.zip
  • Nightly(n150) - playgroundv2_5_pcc_n150.zip
  • Nightly(p150) - https://github.com/tenstorrent/tt-xla/actions/runs/28516865044/job/84534221782

• SDXL-lightning

  • Demo(n150 & p100) - sdxl_lightning_demo.zip
  • Nightly(n150) - sdxl_lightning_pcc.zip
  • Nightly(p150) - https://github.com/tenstorrent/tt-xla/actions/runs/28516865044/job/84534221813

Note : E2E Demo outputs are present in demo zip files

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 33.85%. Comparing base (827ca57) to head (796e0ee).
⚠️ Report is 5 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #5480   +/-   ##
=======================================
  Coverage   33.84%   33.85%           
=======================================
  Files          37       37           
  Lines        4990     4989    -1     
=======================================
  Hits         1689     1689           
+ Misses       3301     3300    -1     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.