[None][test] Add e2e example tests for flux1/2, ltx2, wan_i2v, and cosmos3

[chang-l]

Summary

Five example scripts were recently added under examples/visual_gen/models/
(flux1.py, flux2.py, ltx2.py, wan_i2v.py, cosmos3_ti2v.py) but none
had integration test coverage. This PR adds one smoke test per script and wires
them into CI.

Tests added (tests/integration/defs/examples/test_visual_gen.py):

• test_flux1_example — runs flux1.py + configs/flux1-dev-fp4-1gpu.yaml
• test_flux2_example — runs flux2.py + configs/flux2-dev-fp4-1gpu.yaml
• test_ltx2_example — runs ltx2.py + configs/ltx2-t2v-fp4-1gpu.yaml (passes --text_encoder_path separately since the shared YAML omits it intentionally)
• test_wan_i2v_example — runs wan_i2v.py + configs/wan2.2-i2v-fp4-1gpu.yaml (uses NVFP4 pre-quantized checkpoint, default cat_piano.png input image)
• test_cosmos3_example — runs cosmos3_ti2v.py + configs/cosmos3-nano-1gpu.yaml

Each test:

1. Skips gracefully if the model checkpoint is absent under LLM_MODELS_ROOT
2. Passes the shared YAML config from examples/visual_gen/configs/
3. Asserts the expected output file (.png or .mp4) is produced

CI registration (tests/integration/test_lists/test-db/l0_b200.yml):
All five tests registered in the B200 post-merge stage alongside the existing test_wan_t2v_example.

Test plan

• Pre-merge CI (B200 post-merge stage) runs all 5 new tests
• Tests with missing checkpoints skip cleanly rather than fail
• Validate test list entries exist in source files (AST) pre-commit hook passes (already verified locally)

🤖 Generated with Claude Code

Summary by CodeRabbit

• Tests
  • Added comprehensive end-to-end integration tests for visual generation examples, covering Flux1, Flux2, LTX-2, Wan 2.2 I2V, and Cosmos3.
  • New tests validate example scripts, configurations, and output generation for each model.

[coderabbitai]

📝 Walkthrough

Walkthrough

This PR expands VisualGen integration test coverage by adding a model checkpoint constant and five end-to-end example tests (Flux1, Flux2, LTX-2, Wan I2V, Cosmos3), then registering those tests in the pre-merge CI configuration.

Changes

VisualGen Example Tests

Layer / File(s)	Summary
Model checkpoint constant definition tests/integration/defs/examples/test_visual_gen.py	Defines WAN22_I2V_A14B_NVFP4_MODEL_SUBPATH constant for the Wan I2V A14B NVFP4 checkpoint.
Visual-generation example integration tests tests/integration/defs/examples/test_visual_gen.py	Implements five integration tests (test_flux1_example, test_flux2_example, test_ltx2_example, test_wan_i2v_example, test_cosmos3_example) that validate checkpoint resolution, confirm example scripts and YAML configs exist, execute examples with model-specific CLI arguments, and verify output media files are produced.
Test matrix configuration update tests/integration/test_lists/test-db/l0_b200.yml	Registers the five new example tests in the l0_b200 pre-merge PyTorch test suite (1 GPU, Ubuntu).

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Possibly related PRs

• NVIDIA/TensorRT-LLM#14685: Extends VisualGen integration test coverage in tests/integration/defs/examples/test_visual_gen.py by adding end-to-end example tests.
• NVIDIA/TensorRT-LLM#14981: Introduces the Wan I2V example script that the new test_wan_i2v_example integration test validates.
• NVIDIA/TensorRT-LLM#14976: Introduces LTX-2 example artifacts that the new test_ltx2_example test exercises and validates.

Suggested labels

VisualGen

Suggested reviewers

• laikhtewari
• venkywonka
• karljang

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and specifically summarizes the main change: adding end-to-end integration tests for five visual generation example scripts (flux1/2, ltx2, wan_i2v, cosmos3).
Description check	✅ Passed	The pull request description is comprehensive and well-structured, covering the summary of changes, test details, CI registration, and test plan checklist.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

🧹 Nitpick comments (1)

tests/integration/defs/examples/test_visual_gen.py (1)

1423-1429: 💤 Low value

Consider using _lpips_model_path() helper for consistency.

While the current checkpoint resolution pattern (direct os.path.join(conftest.llm_models_root(), ...)) matches the existing test_wan_t2v_example, the other four new tests in this PR use the _lpips_model_path() helper. For consistency within the new code, consider:

-    scratch_space = conftest.llm_models_root()
-    model_path = os.path.join(scratch_space, WAN22_I2V_A14B_NVFP4_MODEL_SUBPATH)
-    if not os.path.isdir(model_path):
-        pytest.skip(
-            f"Model not found: {model_path} "
-            f"(set LLM_MODELS_ROOT or place {WAN22_I2V_A14B_NVFP4_MODEL_SUBPATH} under models root)"
-        )
+    model_path = _lpips_model_path(WAN22_I2V_A14B_NVFP4_MODEL_SUBPATH)
+    _skip_if_missing(model_path, "Wan 2.2 I2V checkpoint", is_dir=True)

This would align with test_flux1_example, test_flux2_example, test_ltx2_example, and test_cosmos3_example.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@tests/integration/defs/examples/test_visual_gen.py` around lines 1423 - 1429,
Replace the ad-hoc model path construction in
tests/integration/defs/examples/test_visual_gen.py that uses scratch_space =
conftest.llm_models_root() and os.path.join(...,
WAN22_I2V_A14B_NVFP4_MODEL_SUBPATH) with the existing helper _lpips_model_path()
used by the other new tests; call
_lpips_model_path(WAN22_I2V_A14B_NVFP4_MODEL_SUBPATH) (or the equivalent helper
import) to obtain model_path and keep the same pytest.skip behavior if the
returned path is missing so the test matches test_flux1_example,
test_flux2_example, test_ltx2_example, and test_cosmos3_example.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@tests/integration/defs/examples/test_visual_gen.py`:
- Around line 1423-1429: Replace the ad-hoc model path construction in
tests/integration/defs/examples/test_visual_gen.py that uses scratch_space =
conftest.llm_models_root() and os.path.join(...,
WAN22_I2V_A14B_NVFP4_MODEL_SUBPATH) with the existing helper _lpips_model_path()
used by the other new tests; call
_lpips_model_path(WAN22_I2V_A14B_NVFP4_MODEL_SUBPATH) (or the equivalent helper
import) to obtain model_path and keep the same pytest.skip behavior if the
returned path is missing so the test matches test_flux1_example,
test_flux2_example, test_ltx2_example, and test_cosmos3_example.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 91452b27-922f-4ae1-aa7d-31d32381173f

📥 Commits

Reviewing files that changed from the base of the PR and between 9827c21 and a4566c6.

📒 Files selected for processing (2)

• tests/integration/defs/examples/test_visual_gen.py
• tests/integration/test_lists/test-db/l0_b200.yml

[chang-l]

/bot run

[tensorrt-cicd]

PR_Github #52875 [ run ] triggered by Bot. Commit: a4566c6 Link to invocation

[yibinl-nvidia]

LGTM for the LTX-2 example test

[tensorrt-cicd]

PR_Github #52875 [ run ] completed with state SUCCESS. Commit: a4566c6
/LLM/main/L0_MergeRequest_PR pipeline #42127 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

[chang-l]

/bot run --disable-fail-fast

[tensorrt-cicd]

PR_Github #52947 [ run ] triggered by Bot. Commit: 137f0c6 Link to invocation

[tensorrt-cicd]

PR_Github #52947 [ run ] completed with state SUCCESS. Commit: 137f0c6
/LLM/main/L0_MergeRequest_PR pipeline #42190 completed with status: 'SUCCESS'

CI Report

Link to invocation