[diffusion] Accept mesh benchmark artifacts (#36553)

python/sglang/multimodal_gen/.claude/skills/sglang-diffusion-benchmark-profile/SKILL.md

@@ -107,7 +107,8 @@ temporal shape. Treat any remaining temporal or conditioning signature miss as
 Eager fallback, not as a valid BCG measurement.
 
 A zero process exit is not sufficient evidence: every accepted row must also
-contain its requested perf dump and a generated image, video, or audio file.
+contain its requested perf dump and a generated image, video, audio, or 3D mesh
+file.
 The helper gives every cell a unique output name and rejects missing artifacts.
 
 On machines with a read-only Hugging Face cache, combine

python/sglang/multimodal_gen/.claude/skills/sglang-diffusion-benchmark-profile/benchmark-and-profile.md

@@ -161,10 +161,10 @@ Eager/BCG/BCG/Eager at `lossless`, then the same sequence at `high`, while
 holding one GPU set and one isolated checkpoint cache. The high+BCG cells test
 whether the combination is actually supported; do not average them when the
 runtime rejects the combination or the helper detects a late quality-fusion
-mount. The helper also hashes every generated artifact, first requires the two
+mount. The helper hashes every generated image, video, audio, or 3D mesh
-Eager rows at each quality to agree, then rejects any BCG row whose hash differs
+artifact. It first requires the two Eager rows at each quality to agree, then
-from that Eager reference. Cleanup occurs only after all eight runs, including
+rejects any BCG row whose hash differs from that Eager reference. Cleanup occurs
-on failure or interruption:
+only after all eight runs, including on failure or interruption:
 
 ```bash
 MODEL_CACHE_ROOT=/path/to/task-owned/model-caches

python/sglang/multimodal_gen/.claude/skills/sglang-diffusion-benchmark-profile/scripts/bench_diffusion_denoise.py

@@ -121,7 +121,16 @@ MODEL_WEIGHT_SUFFIXES = {
     ".pth",
     ".safetensors",
 }
-GENERATED_OUTPUT_SUFFIXES = {".jpeg", ".jpg", ".mp4", ".png", ".wav", ".webp"}
+GENERATED_OUTPUT_SUFFIXES = {
+    ".glb",
+    ".jpeg",
+    ".jpg",
+    ".mp4",
+    ".obj",
+    ".png",
+    ".wav",
+    ".webp",
+}
 NIGHTLY_PRESET_ORDER = (
     "flux",
     "flux2",

python/sglang/multimodal_gen/test/unit/test_diffusion_benchmark_skill.py

@@ -472,6 +472,41 @@ class TestDiffusionBenchmarkSkill(unittest.TestCase):
                 result["missing_artifacts"], ["perf dump", "generated output"]
             )
 
+    def test_mesh_artifacts_are_accepted_and_hashed(self):
+        with tempfile.TemporaryDirectory() as tmpdir:
+            temp_root = Path(tmpdir)
+            module = _load_benchmark_module(temp_root)
+            output_dir = temp_root / "outputs"
+            output_dir.mkdir()
+
+            def finish_run():
+                (output_dir / "hunyuan3d-shape_mesh-output.json").write_text(
+                    json.dumps({"total_duration_ms": 1000, "steps": []}),
+                    encoding="utf-8",
+                )
+                (output_dir / "hunyuan3d-shape-mesh-output.obj").write_bytes(
+                    b"v 0 0 0\n"
+                )
+                return 0
+
+            with patch.object(module.subprocess, "Popen") as popen:
+                popen.return_value.stdout = iter(())
+                popen.return_value.wait.side_effect = finish_run
+                result = module._run_benchmark_once_impl(
+                    "hunyuan3d-shape",
+                    "mesh-output",
+                    output_dir,
+                    warmup=False,
+                    cuda_visible_devices="0",
+                )
+
+            self.assertFalse(result["error"])
+            self.assertEqual(
+                result["output_artifacts"],
+                [str(output_dir / "hunyuan3d-shape-mesh-output.obj")],
+            )
+            self.assertEqual(len(result["output_sha256"]), 1)
+
     def test_high_bcg_rejects_quality_fusion_mounted_after_capture(self):
         with tempfile.TemporaryDirectory() as tmpdir:
             temp_root = Path(tmpdir)