[Bagel]: Support SP

docs/.nav.yml

@@ -92,11 +92,13 @@ nav:
       - design/feature/cache_dit.md
       - design/feature/teacache.md
       - design/feature/async_chunk_design.md
+      - design/feature/vae_parallel.md
       - design/feature/diffusion_step_execution.md
     - Module Design:
       - design/module/ar_module.md
       - design/module/dit_module.md
       - design/module/entrypoint_module.md
+      - design/module/async_omni_architecture.md
   - Docs Guide: contributing/DOCS_GUIDE.md
 - API Reference:
   - api/README.md

docs/user_guide/diffusion/parallelism_acceleration.md

@@ -39,6 +39,7 @@ The following table shows which models are currently supported by parallelism me
 | **FLUX.1-dev**           | `black-forest-labs/FLUX.1-dev`       |     ❌      |    ❌    |      ✅       |        ✅        |         ❌          |      N/A        | ✅   |
 | **FLUX.2-dev**           | `black-forest-labs/FLUX.2-dev`       |     ❌      |    ❌    |      ❌       |        ✅        |         ❌          |      N/A        | ✅   |
 | **HunyuanImage3.0**      | `tencent/HunyuanImage-3.0`, `tencent/HunyuanImage-3.0-Instruct` |     ❌      |    ❌    |      ❌       |        ✅        |         ❌          |      ✅        | ❌   |
+| **Bagel**                  | `ByteDance-Seed/BAGEL-7B-MoT` |     ✅      |    ✅    |      ✅       |        ✅        |         ❌          |      N/A        | ❌   |
 | **DreamID-Omni**           | `XuGuo699/DreamID-Omni`       |     ❌      |    ❌    |      ✅       |        ❌        |         ❌          |      N/A        | ❌   |
 | **OmniGen2**               | `OmniGen2/OmniGen2`           |     ❌      |    ❌    |      ❌       |        ✅        |         ❌          |      N/A        | ❌   |
 

docs/user_guide/diffusion_acceleration.md

@@ -66,7 +66,7 @@ The following table shows which models are currently supported by each accelerat
 | **Qwen-Image-Layered** | `Qwen/Qwen-Image-Layered` | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |
 | **Z-Image** | `Tongyi-MAI/Z-Image-Turbo` | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ (TP=2 only) | ✅ |
 | **Stable-Diffusion3.5** | `stabilityai/stable-diffusion-3.5` | ❌ | ✅ | ❌ | ❌ | ✅ | ❌ | ❌ |
-| **Bagel** | `ByteDance-Seed/BAGEL-7B-MoT` | ✅ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ |
+| **Bagel** | `ByteDance-Seed/BAGEL-7B-MoT` | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |
 | **FLUX.1-dev** | `black-forest-labs/FLUX.1-dev` | ❌ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ |
 | **NextStep-1.1** | `stepfun-ai/NextStep-1.1` | ❌ | ❌ | ❌ | ❌ | ✅ | ✅ | ❌ |
 | **FLUX.2-klein** | `black-forest-labs/FLUX.2-klein-4B` | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ |

tests/diffusion/test_diffusion_model_runner.py

@@ -59,6 +59,7 @@ def _make_runner(cache_backend, cache_backend_name: str, enable_cache_dit_summar
     runner.kv_transfer_manager = SimpleNamespace(
         receive_kv_cache=lambda req, target_device=None: None,
         receive_multi_kv_cache=lambda req, cfg_kv_collect_func=None, target_device=None: None,
+        receive_multi_kv_cache_distributed=lambda req, cfg_kv_collect_func=None, target_device=None: None,
     )
     return runner
 

tests/e2e/online_serving/test_bagel_expansion.py

@@ -7,6 +7,8 @@
 - Cache-DiT
 - CFG-Parallel
 - Tensor-Parallel
+- Ulysses-SP
+- Ring-Attention
 
 assert_diffusion_response validates successful generation and the expected
 512x512 resolution.
@@ -31,7 +33,8 @@
 
 def _get_diffusion_feature_cases(model: str):
     """Return L4 diffusion feature cases for Bagel.
-    TeaCache, Cache-DiT, CFG-Parallel, Tensor-Parallel.
+    TeaCache, Cache-DiT, CFG-Parallel, Tensor-Parallel,
+    Ulysses-SP, Ring-Attention.
     """
 
     return [
@@ -87,6 +90,30 @@ def _get_diffusion_feature_cases(model: str):
             id="parallel_tp_2",
             marks=PARALLEL_FEATURE_MARKS,
         ),
+        # Ulysses-SP degree=2 (2 GPUs)
+        pytest.param(
+            OmniServerParams(
+                model=model,
+                server_args=[
+                    "--usp",
+                    "2",
+                ],
+            ),
+            id="sp_ulysses_2",
+            marks=PARALLEL_FEATURE_MARKS,
+        ),
+        # Ring-Attention degree=2 (2 GPUs)
+        pytest.param(
+            OmniServerParams(
+                model=model,
+                server_args=[
+                    "--ring",
+                    "2",
+                ],
+            ),
+            id="sp_ring_2",
+            marks=PARALLEL_FEATURE_MARKS,
+        ),
     ]
 
 
@@ -108,6 +135,8 @@ def test_bagel(
     - Cache-DiT
     - CFG-Parallel (size=2)
     - Tensor-Parallel (size=2)
+    - Ulysses-SP (degree=2)
+    - Ring-Attention (degree=2)
 
     Validation is delegated to assert_diffusion_response in tests.conftest,
     which checks output dimensions and basic correctness.