Enhance reproducibility in Ulysses attention tests by replacing current_omni_platform.seed_everything with torch's manual_seed and adding support for CUDA devices. Introduce get_supported_tasks method in AsyncOmni class to dynamically determine available tasks based on stage capabilities.

Signed-off-by: tzhouam <tzhouam@connect.ust.hk>

tests/diffusion/attention/test_attention_sp.py

@@ -329,7 +329,9 @@ def ulysses_attention_on_test_model(
     """Run Ulysses attention test on a test model and save results for comparison."""
     # Use fixed seed for reproducibility across baseline and SP runs
     RANDOM_SEED = 42
-    current_omni_platform.seed_everything(RANDOM_SEED)
+    torch.manual_seed(RANDOM_SEED)
+    if torch.cuda.is_available():
+        torch.cuda.manual_seed_all(RANDOM_SEED)
 
     mode_str = "Baseline (no SP)" if is_baseline else f"SP (ulysses={ulysses_degree}, ring={ring_degree})"
     print(f"\n[{mode_str}] Rank {local_rank}/{world_size} - Random seed set to {RANDOM_SEED}")
@@ -414,7 +416,8 @@ def ulysses_attention_on_test_model(
             # Generate and save full input data with fixed seed
             # Reinitialize RNG to ensure reproducibility
             torch.manual_seed(42)
-            current_omni_platform.seed_everything(42)
+            if torch.cuda.is_available():
+                torch.cuda.manual_seed_all(42)
             full_hidden_states = torch.randn(
                 (batch_size, seq_len, hidden_size),
                 dtype=dtype,

vllm_omni/entrypoints/async_omni.py

@@ -658,6 +658,16 @@ class AsyncOmni(OmniBase):
             stage.submit(abort_task)
         return None
 
+    async def get_supported_tasks(self) -> set[str]:
+        tasks: set[str] = set()
+        has_comprehension = any(stage.is_comprehension for stage in self.stage_list)
+        if has_comprehension:
+            tasks.add("generate")
+        for stage in self.stage_list:
+            if getattr(stage, "final_output_type", None) == "audio":
+                tasks.add("speech")
+        return tasks if tasks else {"generate"}
+
     async def get_vllm_config(self) -> VllmConfig:
         for stage in self.stage_list:
             if stage.is_comprehension: