[skip ci][Doc] Remove obsolete DiffusionParallelConfig in many example scripts (#6347)

Signed-off-by: Huang, Zeyu <11222265+fhfuih@users.noreply.github.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>

examples/offline_inference/custom_pipeline/image_to_image/image_edit.py

@@ -53,7 +53,6 @@ from typing import Any
 import torch
 from PIL import Image
 
-from vllm_omni.diffusion.data import DiffusionParallelConfig
 from vllm_omni.entrypoints.omni import Omni
 from vllm_omni.inputs.data import OmniDiffusionSamplingParams
 from vllm_omni.outputs import OmniRequestOutput
@@ -139,12 +138,6 @@ async def main():
 
     # ---- Torch setup ----
     generator = torch.Generator(device=current_omni_platform.device_type).manual_seed(args.seed)
-    parallel_config = DiffusionParallelConfig(
-        ulysses_degree=args.ulysses_degree,
-        ring_degree=args.ring_degree,
-        cfg_parallel_size=args.cfg_parallel_size,
-        tensor_parallel_size=args.tensor_parallel_size,
-    )
 
     # ---- Cache Config ----
     if args.cache_backend == "cache_dit":
@@ -171,7 +164,10 @@ async def main():
         vae_use_tiling=args.vae_use_tiling,
         cache_backend=args.cache_backend,
         cache_config=cache_config,
-        parallel_config=parallel_config,
+        ulysses_degree=args.ulysses_degree,
+        ring_degree=args.ring_degree,
+        cfg_parallel_size=args.cfg_parallel_size,
+        tensor_parallel_size=args.tensor_parallel_size,
         enforce_eager=args.enforce_eager,
         enable_cpu_offload=args.enable_cpu_offload,
         diffusion_load_format="dummy",

examples/offline_inference/diffusion/lingbot_world_v2.py

@@ -154,8 +154,6 @@ def _positive_finite(value: float, flag: str) -> float:
 def build_omni_kwargs(
     args: argparse.Namespace,
     paths: LingBotPaths,
-    *,
-    parallel_config: Any,
 ) -> dict[str, Any]:
     """Build Omni configuration without overriding checkpoint class discovery."""
 
@@ -167,7 +165,7 @@ def build_omni_kwargs(
     return {
         "model": model,
         "flow_shift": flow_shift,
-        "parallel_config": parallel_config,
+        "tensor_parallel_size": args.tensor_parallel_size,
         "enforce_eager": args.enforce_eager,
         "model_config": {"lingbot_action_root": str(paths.action_root)},
     }
@@ -262,12 +260,10 @@ def main(argv: Sequence[str] | None = None) -> Path:
     # helper tests work on machines without the native vLLM runtime.
     from diffusers.utils import export_to_video
 
-    from vllm_omni.diffusion.data import DiffusionParallelConfig
     from vllm_omni.entrypoints.omni import Omni
     from vllm_omni.inputs.data import OmniDiffusionSamplingParams
 
-    parallel_config = DiffusionParallelConfig(tensor_parallel_size=args.tensor_parallel_size)
-    omni_kwargs = build_omni_kwargs(args, paths, parallel_config=parallel_config)
+    omni_kwargs = build_omni_kwargs(args, paths)
     omni = Omni(**omni_kwargs)
     try:
         outputs = omni.generate(

examples/offline_inference/diffusion/lingbot_world_v2_realtime.py

@@ -125,7 +125,6 @@ async def run(argv: Sequence[str] | None = None) -> Path:
     # require a CUDA-enabled vLLM installation.
     import torch
 
-    from vllm_omni.diffusion.data import DiffusionParallelConfig
     from vllm_omni.diffusion.models.lingbot_world.actions import LingBotCameraControlReducer
     from vllm_omni.entrypoints.async_omni import AsyncOmni
     from vllm_omni.experimental.ar_diffusion.consumer import ARDiffusionOmniTickConsumer
@@ -141,7 +140,7 @@ async def run(argv: Sequence[str] | None = None) -> Path:
         model=args.model,
         engine_backend="vllm_omni.experimental.ar_diffusion.engine.ARDiffusionEngine",
         enforce_eager=args.enforce_eager,
-        parallel_config=DiffusionParallelConfig(tensor_parallel_size=args.tensor_parallel_size),
+        tensor_parallel_size=args.tensor_parallel_size,
         max_num_seqs=1,
         model_config={
             "ar_diffusion_height": args.height,

examples/offline_inference/image_to_image/image_edit.py

@@ -96,7 +96,6 @@ from typing import Any
 import torch
 from PIL import Image
 
-from vllm_omni.diffusion.data import DiffusionParallelConfig
 from vllm_omni.diffusion.utils.image_output import extract_images_from_outputs
 from vllm_omni.diffusion.utils.param_utils import apply_declared_extra_args
 from vllm_omni.entrypoints.omni import Omni
@@ -452,14 +451,6 @@ def main():
 
     generator = torch.Generator(device=current_omni_platform.device_type).manual_seed(args.seed)
 
-    parallel_config = DiffusionParallelConfig(
-        ulysses_degree=args.ulysses_degree,
-        ring_degree=args.ring_degree,
-        cfg_parallel_size=args.cfg_parallel_size,
-        tensor_parallel_size=args.tensor_parallel_size,
-        enable_expert_parallel=args.enable_expert_parallel,
-    )
-
     # Configure cache based on backend type
     cache_config = None
     if args.cache_backend == "cache_dit":
@@ -490,7 +481,11 @@ def main():
         vae_use_tiling=args.vae_use_tiling,
         cache_backend=args.cache_backend,
         cache_config=cache_config,
-        parallel_config=parallel_config,
+        ulysses_degree=args.ulysses_degree,
+        ring_degree=args.ring_degree,
+        cfg_parallel_size=args.cfg_parallel_size,
+        tensor_parallel_size=args.tensor_parallel_size,
+        enable_expert_parallel=args.enable_expert_parallel,
         enable_cpu_offload=args.enable_cpu_offload,
         enable_diffusion_pipeline_profiler=args.enable_diffusion_pipeline_profiler,
         profiler_config=args.profiler_config,

examples/offline_inference/image_to_video/image_to_video.py

@@ -56,7 +56,6 @@ import numpy as np
 import PIL.Image
 import torch
 
-from vllm_omni.diffusion.data import DiffusionParallelConfig
 from vllm_omni.diffusion.utils.param_utils import apply_declared_extra_args
 from vllm_omni.entrypoints.omni import Omni
 from vllm_omni.inputs.data import OmniDiffusionSamplingParams
@@ -503,17 +502,6 @@ def main():
         }
 
     profiler_enabled = args.profiler_config is not None
-    parallel_config = DiffusionParallelConfig(
-        ulysses_degree=args.ulysses_degree,
-        ring_degree=args.ring_degree,
-        cfg_parallel_size=args.cfg_parallel_size,
-        tensor_parallel_size=args.tensor_parallel_size,
-        vae_patch_parallel_size=args.vae_patch_parallel_size,
-        use_hsdp=args.use_hsdp,
-        hsdp_shard_size=args.hsdp_shard_size,
-        hsdp_replicate_size=args.hsdp_replicate_size,
-        pipeline_parallel_size=args.pipeline_parallel_size,
-    )
     omni_kwargs = dict(
         model=args.model,
         enable_layerwise_offload=args.enable_layerwise_offload,
@@ -524,7 +512,15 @@ def main():
         diffusion_kv_cache_skip_steps=args.diffusion_kv_cache_skip_steps,
         diffusion_kv_cache_skip_layers=args.diffusion_kv_cache_skip_layers,
         enable_cpu_offload=args.enable_cpu_offload,
-        parallel_config=parallel_config,
+        ulysses_degree=args.ulysses_degree,
+        ring_degree=args.ring_degree,
+        cfg_parallel_size=args.cfg_parallel_size,
+        tensor_parallel_size=args.tensor_parallel_size,
+        vae_patch_parallel_size=args.vae_patch_parallel_size,
+        use_hsdp=args.use_hsdp,
+        hsdp_shard_size=args.hsdp_shard_size,
+        hsdp_replicate_size=args.hsdp_replicate_size,
+        pipeline_parallel_size=args.pipeline_parallel_size,
         enforce_eager=args.enforce_eager,
         model_class_name=model_class_name,
         cache_backend=args.cache_backend,

examples/offline_inference/longcat_video/end2end.py

@@ -15,7 +15,6 @@ import imageio
 import numpy as np
 import torch
 
-from vllm_omni.diffusion.data import DiffusionParallelConfig
 from vllm_omni.diffusion.models.longcat_video.pipeline_longcat_video_avatar import (
     prepare_longcat_video_avatar_model_for_omni,
 )
@@ -236,14 +235,6 @@ def main():
         dtype="bfloat16",
         enforce_eager=args.enforce_eager,
         additional_config=additional_config,
-        parallel_config=DiffusionParallelConfig(
-            ulysses_degree=1,
-            ring_degree=1,
-            cfg_parallel_size=1,
-            tensor_parallel_size=1,
-            vae_patch_parallel_size=1,
-            pipeline_parallel_size=1,
-        ),
     )
 
     audio_input: str | dict[str, str] | None = args.audio

examples/offline_inference/speech_to_video/speech_to_video.py

@@ -23,7 +23,6 @@ from typing import Any
 
 import torch
 
-from vllm_omni.diffusion.data import DiffusionParallelConfig
 from vllm_omni.entrypoints.omni import Omni
 from vllm_omni.inputs.data import OmniDiffusionSamplingParams
 from vllm_omni.platforms import current_omni_platform
@@ -250,17 +249,6 @@ def main():
             "scm_steps_policy": "dynamic",
         }
 
-    parallel_config = DiffusionParallelConfig(
-        tensor_parallel_size=args.tensor_parallel_size,
-        cfg_parallel_size=args.cfg_parallel_size,
-        vae_patch_parallel_size=args.vae_patch_parallel_size,
-        ring_degree=args.ring_degree,
-        ulysses_degree=args.ulysses_degree,
-        use_hsdp=args.use_hsdp,
-        hsdp_shard_size=args.hsdp_shard_size,
-        hsdp_replicate_size=args.hsdp_replicate_size,
-    )
-
     profiler_enabled = args.profiler_config is not None
 
     omni_kwargs = dict(
@@ -271,7 +259,14 @@ def main():
         vae_use_tiling=args.vae_use_tiling,
         enable_cpu_offload=args.enable_cpu_offload,
         enable_layerwise_offload=args.enable_layerwise_offload,
-        parallel_config=parallel_config,
+        tensor_parallel_size=args.tensor_parallel_size,
+        cfg_parallel_size=args.cfg_parallel_size,
+        vae_patch_parallel_size=args.vae_patch_parallel_size,
+        ring_degree=args.ring_degree,
+        ulysses_degree=args.ulysses_degree,
+        use_hsdp=args.use_hsdp,
+        hsdp_shard_size=args.hsdp_shard_size,
+        hsdp_replicate_size=args.hsdp_replicate_size,
         enforce_eager=args.enforce_eager,
         enable_diffusion_pipeline_profiler=args.enable_diffusion_pipeline_profiler,
         cache_backend=args.cache_backend,

examples/offline_inference/text_to_audio/text_to_audio.py

@@ -38,7 +38,6 @@ from typing import Any
 import numpy as np
 import torch
 
-from vllm_omni.diffusion.data import DiffusionParallelConfig
 from vllm_omni.diffusion.utils.param_utils import apply_declared_extra_args
 from vllm_omni.entrypoints.omni import Omni
 from vllm_omni.inputs.data import OmniDiffusionSamplingParams
@@ -306,17 +305,13 @@ def main():
     print(f"  Seed: {args.seed}")
     print(f"{'=' * 60}\n")
 
-    parallel_config = DiffusionParallelConfig(
-        use_hsdp=args.use_hsdp,
-        hsdp_shard_size=args.hsdp_shard_size,
-        hsdp_replicate_size=args.hsdp_replicate_size,
-    )
-
     # Initialize Omni. AudioX requires an explicit pipeline class; Stable Audio
     # is resolved from its config.
     omni_kwargs: dict[str, Any] = dict(
         model=args.model,
-        parallel_config=parallel_config,
+        use_hsdp=args.use_hsdp,
+        hsdp_shard_size=args.hsdp_shard_size,
+        hsdp_replicate_size=args.hsdp_replicate_size,
         cache_backend=args.cache_backend,
         cache_config=cache_config,
         enable_diffusion_pipeline_profiler=args.enable_diffusion_pipeline_profiler,

examples/offline_inference/text_to_video/text_to_video.py

@@ -10,7 +10,6 @@ from typing import Any
 import numpy as np
 import torch
 
-from vllm_omni.diffusion.data import DiffusionParallelConfig
 from vllm_omni.diffusion.utils.param_utils import apply_declared_extra_args
 from vllm_omni.entrypoints.omni import Omni
 from vllm_omni.inputs.data import OmniDiffusionSamplingParams
@@ -462,17 +461,6 @@ def main():
             "scm_steps_policy": "dynamic",
         }
 
-    # Configure parallel settings
-    parallel_config = DiffusionParallelConfig(
-        ulysses_degree=args.ulysses_degree,
-        ring_degree=args.ring_degree,
-        cfg_parallel_size=args.cfg_parallel_size,
-        tensor_parallel_size=args.tensor_parallel_size,
-        vae_patch_parallel_size=args.vae_patch_parallel_size,
-        pipeline_parallel_size=args.pipeline_parallel_size,
-        enable_expert_parallel=args.enable_expert_parallel,
-    )
-
     profiler_enabled = args.profiler_config is not None
 
     omni_kwargs = dict(
@@ -481,7 +469,13 @@ def main():
         vae_use_slicing=args.vae_use_slicing,
         vae_use_tiling=args.vae_use_tiling,
         enable_cpu_offload=args.enable_cpu_offload,
-        parallel_config=parallel_config,
+        ulysses_degree=args.ulysses_degree,
+        ring_degree=args.ring_degree,
+        cfg_parallel_size=args.cfg_parallel_size,
+        tensor_parallel_size=args.tensor_parallel_size,
+        vae_patch_parallel_size=args.vae_patch_parallel_size,
+        pipeline_parallel_size=args.pipeline_parallel_size,
+        enable_expert_parallel=args.enable_expert_parallel,
         enforce_eager=args.enforce_eager,
         model_class_name=model_class_name,
         cache_backend=args.cache_backend,

examples/offline_inference/x_to_video_audio/x_to_video_audio.py

@@ -11,7 +11,6 @@ import numpy as np
 from PIL import Image
 from vllm.multimodal.media.audio import load_audio
 
-from vllm_omni.diffusion.data import DiffusionParallelConfig
 from vllm_omni.diffusion.utils.media_utils import mux_video_audio_bytes
 from vllm_omni.entrypoints.omni import Omni
 from vllm_omni.inputs.data import OmniDiffusionSamplingParams
@@ -197,14 +196,6 @@ def main() -> None:
             },
         )
 
-    parallel_config = DiffusionParallelConfig(
-        cfg_parallel_size=args.cfg_parallel_size,
-        tensor_parallel_size=args.tensor_parallel_size,
-        use_hsdp=args.use_hsdp,
-        hsdp_shard_size=args.hsdp_shard_size,
-        hsdp_replicate_size=args.hsdp_replicate_size,
-    )
-
     cache_config = None
     if args.cache_backend == "cache_dit":
         cache_config = {
@@ -222,7 +213,11 @@ def main() -> None:
 
     omni_kwargs = dict(
         model=args.model,
-        parallel_config=parallel_config,
+        cfg_parallel_size=args.cfg_parallel_size,
+        tensor_parallel_size=args.tensor_parallel_size,
+        use_hsdp=args.use_hsdp,
+        hsdp_shard_size=args.hsdp_shard_size,
+        hsdp_replicate_size=args.hsdp_replicate_size,
         model_type=args.model_type,
         enable_cpu_offload=args.enable_cpu_offload,
         enable_layerwise_offload=args.enable_layerwise_offload,