diff --git a/python/pyproject.toml b/python/pyproject.toml
index ce4d87ff3..bfe0dd649 100755
--- a/python/pyproject.toml
+++ b/python/pyproject.toml
@@ -89,7 +89,7 @@ diffusion = [
   "imageio==2.36.0",
   "imageio-ffmpeg==0.5.1",
   "moviepy>=2.0.0",
-  "opencv-python==4.10.0.84",
+  "opencv-python-headless==4.10.0.84",
   "remote-pdb",
   "st_attn ==0.0.7",
   "vsa==0.0.4",
diff --git a/python/sglang/multimodal_gen/configs/sample/sampling_params.py b/python/sglang/multimodal_gen/configs/sample/sampling_params.py
index 2adc7c94f..4d538a93f 100644
--- a/python/sglang/multimodal_gen/configs/sample/sampling_params.py
+++ b/python/sglang/multimodal_gen/configs/sample/sampling_params.py
@@ -16,6 +16,7 @@ from dataclasses import dataclass
 from enum import Enum, auto
 from typing import Any
 
+from sglang.multimodal_gen import envs
 from sglang.multimodal_gen.runtime.utils.logging_utils import init_logger
 from sglang.multimodal_gen.utils import StoreBoolean
 
@@ -207,6 +208,10 @@ class SamplingParams:
         if env_steps is not None and self.num_inference_steps is not None:
             self.num_inference_steps = int(env_steps)
 
+        # Auto-enable stage logging if dump path is provided
+        if self.perf_dump_path:
+            envs.SGLANG_DIFFUSION_STAGE_LOGGING = True
+
     def _validate(self):
         """
         check if the sampling params is correct by itself
diff --git a/python/sglang/multimodal_gen/runtime/entrypoints/cli/generate.py b/python/sglang/multimodal_gen/runtime/entrypoints/cli/generate.py
index 42c4489f5..0aad34db2 100644
--- a/python/sglang/multimodal_gen/runtime/entrypoints/cli/generate.py
+++ b/python/sglang/multimodal_gen/runtime/entrypoints/cli/generate.py
@@ -8,7 +8,6 @@ import dataclasses
 import os
 from typing import cast
 
-import sglang.multimodal_gen.envs as envs
 from sglang.multimodal_gen import DiffGenerator
 from sglang.multimodal_gen.configs.sample.sampling_params import SamplingParams
 from sglang.multimodal_gen.runtime.entrypoints.cli.cli_types import CLISubcommand
@@ -89,11 +88,6 @@ def generate_cmd(args: argparse.Namespace):
     """The entry point for the generate command."""
     args.request_id = "mocked_fake_id_for_offline_generate"
 
-    # Auto-enable stage logging if dump path is provided
-    if args.perf_dump_path:
-        os.environ["SGLANG_DIFFUSION_STAGE_LOGGING"] = "True"
-        envs.SGLANG_DIFFUSION_STAGE_LOGGING = True
-
     server_args = ServerArgs.from_cli_args(args)
     sampling_params_kwargs = SamplingParams.get_cli_args(args)
     generator = DiffGenerator.from_pretrained(
diff --git a/python/sglang/multimodal_gen/runtime/managers/gpu_worker.py b/python/sglang/multimodal_gen/runtime/managers/gpu_worker.py
index f6905d137..a656c03ec 100644
--- a/python/sglang/multimodal_gen/runtime/managers/gpu_worker.py
+++ b/python/sglang/multimodal_gen/runtime/managers/gpu_worker.py
@@ -26,7 +26,10 @@ from sglang.multimodal_gen.runtime.utils.logging_utils import (
     globally_suppress_loggers,
     init_logger,
 )
-from sglang.multimodal_gen.runtime.utils.perf_logger import PerformanceLogger
+from sglang.multimodal_gen.runtime.utils.perf_logger import (
+    PerformanceLogger,
+    StageProfiler,
+)
 
 logger = init_logger(__name__)
 
@@ -108,7 +111,7 @@ class GPUWorker:
             if output_batch.timings:
                 duration_ms = (time.monotonic() - start_time) * 1000
                 output_batch.timings.total_duration_ms = duration_ms
-                if req.perf_dump_path is not None:
+                if StageProfiler.metrics_enabled():
                     PerformanceLogger.log_request_summary(timings=output_batch.timings)
         except Exception as e:
             logger.error(
diff --git a/python/sglang/multimodal_gen/test/run_suite.py b/python/sglang/multimodal_gen/test/run_suite.py
index 35fea064f..9f47799cd 100644
--- a/python/sglang/multimodal_gen/test/run_suite.py
+++ b/python/sglang/multimodal_gen/test/run_suite.py
@@ -77,14 +77,45 @@ def parse_args():
 
 
 def collect_test_items(files, filter_expr=None):
-    """Collect test item node IDs from the given files using pytest --collect-only."""
+    """Collect test item node IDs from the given files using pytest --collect-only.
+
+    Raises:
+        RuntimeError: If pytest collection fails due to errors (e.g., syntax errors,
+            import errors, or other collection failures).
+    """
     cmd = [sys.executable, "-m", "pytest", "--collect-only", "-q"]
     if filter_expr:
         cmd.extend(["-k", filter_expr])
     cmd.extend(files)
 
+    logger.info(f"Collecting tests with command: {' '.join(cmd)}")
     result = subprocess.run(cmd, capture_output=True, text=True)
 
+    # Check for collection errors
+    # pytest exit codes:
+    #   0: success
+    #   1: tests collected but some had errors during collection
+    #   2: test execution interrupted
+    #   3: internal error
+    #   4: command line usage error
+    #   5: no tests collected (may be expected with filters)
+    if result.returncode not in (0, 5):
+        error_msg = (
+            f"pytest --collect-only failed with exit code {result.returncode}\n"
+            f"Command: {' '.join(cmd)}\n"
+        )
+        if result.stderr:
+            error_msg += f"stderr:\n{result.stderr}\n"
+        if result.stdout:
+            error_msg += f"stdout:\n{result.stdout}\n"
+        logger.error(error_msg)
+        raise RuntimeError(error_msg)
+
+    if result.returncode == 5:
+        logger.info(
+            "No tests were collected (exit code 5). This may be expected with filters."
+        )
+
     # Parse the output to extract test node IDs
     # pytest -q outputs lines like: test_file.py::TestClass::test_method[param]
     test_items = []
@@ -97,6 +128,7 @@ def collect_test_items(files, filter_expr=None):
             if "::" in test_id:
                 test_items.append(test_id)
 
+    logger.info(f"Collected {len(test_items)} test items")
     return test_items