Add PRXPipeline in AUTO_TEXT2IMAGE_PIPELINES_MAPPING (#13257)

* fix docs

* update main example

docs/source/en/_toctree.yml

@@ -532,8 +532,6 @@
         title: ControlNet-XS with Stable Diffusion XL
       - local: api/pipelines/controlnet_union
         title: ControlNetUnion
-      - local: api/pipelines/cosmos
-        title: Cosmos
       - local: api/pipelines/ddim
         title: DDIM
       - local: api/pipelines/ddpm
@@ -677,6 +675,8 @@
         title: CogVideoX
       - local: api/pipelines/consisid
         title: ConsisID
+      - local: api/pipelines/cosmos
+        title: Cosmos
       - local: api/pipelines/framepack
         title: Framepack
       - local: api/pipelines/helios

docs/source/en/api/pipelines/cosmos.md

@@ -21,29 +21,31 @@
 > [!TIP]
 > Make sure to check out the Schedulers [guide](../../using-diffusers/schedulers) to learn how to explore the tradeoff between scheduler speed and quality, and see the [reuse components across pipelines](../../using-diffusers/loading#reuse-a-pipeline) section to learn how to efficiently load the same components into multiple pipelines.
 
-## Loading original format checkpoints
-
-Original format checkpoints that have not been converted to diffusers-expected format can be loaded using the `from_single_file` method.
+## Basic usage
 
 ```python
 import torch
-from diffusers import Cosmos2TextToImagePipeline, CosmosTransformer3DModel
+from diffusers import Cosmos2_5_PredictBasePipeline
+from diffusers.utils import export_to_video
 
-model_id = "nvidia/Cosmos-Predict2-2B-Text2Image"
-transformer = CosmosTransformer3DModel.from_single_file(
-    "https://huggingface.co/nvidia/Cosmos-Predict2-2B-Text2Image/blob/main/model.pt",
-    torch_dtype=torch.bfloat16,
-).to("cuda")
-pipe = Cosmos2TextToImagePipeline.from_pretrained(model_id, transformer=transformer, torch_dtype=torch.bfloat16)
+model_id = "nvidia/Cosmos-Predict2.5-2B"
+pipe = Cosmos2_5_PredictBasePipeline.from_pretrained(
+    model_id, revision="diffusers/base/post-trained", torch_dtype=torch.bfloat16
+)
 pipe.to("cuda")
 
-prompt = "A close-up shot captures a vibrant yellow scrubber vigorously working on a grimy plate, its bristles moving in circular motions to lift stubborn grease and food residue. The dish, once covered in remnants of a hearty meal, gradually reveals its original glossy surface. Suds form and bubble around the scrubber, creating a satisfying visual of cleanliness in progress. The sound of scrubbing fills the air, accompanied by the gentle clinking of the dish against the sink. As the scrubber continues its task, the dish transforms, gleaming under the bright kitchen lights, symbolizing the triumph of cleanliness over mess."
+prompt = "As the red light shifts to green, the red bus at the intersection begins to move forward, its headlights cutting through the falling snow. The snowy tire tracks deepen as the vehicle inches ahead, casting fresh lines onto the slushy road. Around it, streetlights glow warmer, illuminating the drifting flakes and wet reflections on the asphalt. Other cars behind start to edge forward, their beams joining the scene. The stillness of the urban street transitions into motion as the quiet snowfall is punctuated by the slow advance of traffic through the frosty city corridor."
 negative_prompt = "The video captures a series of frames showing ugly scenes, static with no motion, motion blur, over-saturation, shaky footage, low resolution, grainy texture, pixelated images, poorly lit areas, underexposed and overexposed scenes, poor color balance, washed out colors, choppy sequences, jerky movements, low frame rate, artifacting, color banding, unnatural transitions, outdated special effects, fake elements, unconvincing visuals, poorly edited content, jump cuts, visual noise, and flickering. Overall, the video is of poor quality."
 
 output = pipe(
-    prompt=prompt, negative_prompt=negative_prompt, generator=torch.Generator().manual_seed(1)
-).images[0]
-output.save("output.png")
+    image=None,
+    video=None,
+    prompt=prompt,
+    negative_prompt=negative_prompt,
+    num_frames=93,
+    generator=torch.Generator().manual_seed(1),
+).frames[0]
+export_to_video(output, "text2world.mp4", fps=16)
 ```
 
 ## Cosmos2_5_TransferPipeline

docs/source/en/api/pipelines/overview.md

@@ -44,6 +44,7 @@ The table below lists all the pipelines currently available in 🤗 Diffusers an
 | [ControlNet with Stable Diffusion XL](controlnet_sdxl) | text2image |
 | [ControlNet-XS](controlnetxs) | text2image |
 | [ControlNet-XS with Stable Diffusion XL](controlnetxs_sdxl) | text2image |
+| [Cosmos](cosmos) | text2video, video2video |
 | [Dance Diffusion](dance_diffusion) | unconditional audio generation |
 | [DDIM](ddim) | unconditional image generation |
 | [DDPM](ddpm) | unconditional image generation |

src/diffusers/models/_modeling_parallel.py

@@ -60,6 +60,16 @@ class ContextParallelConfig:
         rotate_method (`str`, *optional*, defaults to `"allgather"`):
             Method to use for rotating key/value states across devices in ring attention. Currently, only `"allgather"`
             is supported.
+        ulysses_anything (`bool`, *optional*, defaults to `False`):
+            Whether to enable "Ulysses Anything" mode, which supports arbitrary sequence lengths and head counts that
+            are not evenly divisible by `ulysses_degree`. When enabled, `ulysses_degree` must be greater than 1 and
+            `ring_degree` must be 1.
+        mesh (`torch.distributed.device_mesh.DeviceMesh`, *optional*):
+            A custom device mesh to use for context parallelism. If provided, this mesh will be used instead of
+            creating a new one. This is useful when combining context parallelism with other parallelism strategies
+            (e.g., FSDP, tensor parallelism) that share the same device mesh. The mesh must have both "ring" and
+            "ulysses" dimensions. Use size 1 for dimensions not being used (e.g., `mesh_shape=(2, 1, 4)` with
+            `mesh_dim_names=("ring", "ulysses", "fsdp")` for ring attention only with FSDP).
 
     """
 
@@ -68,6 +78,7 @@ class ContextParallelConfig:
     convert_to_fp32: bool = True
     # TODO: support alltoall
     rotate_method: Literal["allgather", "alltoall"] = "allgather"
+    mesh: torch.distributed.device_mesh.DeviceMesh | None = None
     # Whether to enable ulysses anything attention to support
     # any sequence lengths and any head numbers.
     ulysses_anything: bool = False
@@ -124,7 +135,7 @@ class ContextParallelConfig:
                 f"The product of `ring_degree` ({self.ring_degree}) and `ulysses_degree` ({self.ulysses_degree}) must not exceed the world size ({world_size})."
             )
 
-        self._flattened_mesh = self._mesh._flatten()
+        self._flattened_mesh = self._mesh["ring", "ulysses"]._flatten()
         self._ring_mesh = self._mesh["ring"]
         self._ulysses_mesh = self._mesh["ulysses"]
         self._ring_local_rank = self._ring_mesh.get_local_rank()

src/diffusers/models/modeling_utils.py

@@ -1567,7 +1567,7 @@ class ModelMixin(torch.nn.Module, PushToHubMixin):
         mesh = None
         if config.context_parallel_config is not None:
             cp_config = config.context_parallel_config
-            mesh = torch.distributed.device_mesh.init_device_mesh(
+            mesh = cp_config.mesh or torch.distributed.device_mesh.init_device_mesh(
                 device_type=device_type,
                 mesh_shape=cp_config.mesh_shape,
                 mesh_dim_names=cp_config.mesh_dim_names,

src/diffusers/pipelines/auto_pipeline.py

@@ -95,6 +95,7 @@ from .pag import (
     StableDiffusionXLPAGPipeline,
 )
 from .pixart_alpha import PixArtAlphaPipeline, PixArtSigmaPipeline
+from .prx import PRXPipeline
 from .qwenimage import (
     QwenImageControlNetPipeline,
     QwenImageEditInpaintPipeline,
@@ -185,6 +186,7 @@ AUTO_TEXT2IMAGE_PIPELINES_MAPPING = OrderedDict(
         ("z-image-controlnet-inpaint", ZImageControlNetInpaintPipeline),
         ("z-image-omni", ZImageOmniPipeline),
         ("ovis", OvisImagePipeline),
+        ("prx", PRXPipeline),
     ]
 )
 

src/diffusers/pipelines/cosmos/pipeline_cosmos2_5_transfer.py

@@ -82,13 +82,16 @@ EXAMPLE_DOC_STRING = """
         ```python
         >>> import cv2
         >>> import numpy as np
+        >>> from PIL import Image
         >>> import torch
         >>> from diffusers import Cosmos2_5_TransferPipeline, AutoModel
         >>> from diffusers.utils import export_to_video, load_video
 
         >>> model_id = "nvidia/Cosmos-Transfer2.5-2B"
         >>> # Load a Transfer2.5 controlnet variant (edge, depth, seg, or blur)
-        >>> controlnet = AutoModel.from_pretrained(model_id, revision="diffusers/controlnet/general/edge")
+        >>> controlnet = AutoModel.from_pretrained(
+        ...     model_id, revision="diffusers/controlnet/general/edge", torch_dtype=torch.bfloat16
+        ... )
         >>> pipe = Cosmos2_5_TransferPipeline.from_pretrained(
         ...     model_id, controlnet=controlnet, revision="diffusers/general", torch_dtype=torch.bfloat16
         ... )

src/diffusers/quantizers/quantization_config.py

@@ -36,7 +36,7 @@ from typing import Any, Callable
 
 from packaging import version
 
-from ..utils import is_torch_available, is_torchao_available, is_torchao_version, logging
+from ..utils import deprecate, is_torch_available, is_torchao_available, is_torchao_version, logging
 
 
 if is_torch_available():
@@ -844,6 +844,8 @@ class QuantoConfig(QuantizationConfigMixin):
         modules_to_not_convert: list[str] | None = None,
         **kwargs,
     ):
+        deprecation_message = "`QuantoConfig` is deprecated and will be removed in version 1.0.0."
+        deprecate("QuantoConfig", "1.0.0", deprecation_message)
         self.quant_method = QuantizationMethod.QUANTO
         self.weights_dtype = weights_dtype
         self.modules_to_not_convert = modules_to_not_convert

src/diffusers/quantizers/quanto/quanto_quantizer.py

@@ -3,6 +3,7 @@ from typing import TYPE_CHECKING, Any
 from diffusers.utils.import_utils import is_optimum_quanto_version
 
 from ...utils import (
+    deprecate,
     get_module_from_name,
     is_accelerate_available,
     is_accelerate_version,
@@ -42,6 +43,9 @@ class QuantoQuantizer(DiffusersQuantizer):
         super().__init__(quantization_config, **kwargs)
 
     def validate_environment(self, *args, **kwargs):
+        deprecation_message = "The Quanto quantizer is deprecated and will be removed in version 1.0.0."
+        deprecate("QuantoQuantizer", "1.0.0", deprecation_message)
+
         if not is_optimum_quanto_available():
             raise ImportError(
                 "Loading an optimum-quanto quantized model requires optimum-quanto library (`pip install optimum-quanto`)"

tests/models/testing_utils/parallelism.py

@@ -60,12 +60,7 @@ def _context_parallel_worker(rank, world_size, master_port, model_class, init_di
         model.eval()
 
         # Move inputs to device
-        inputs_on_device = {}
-        for key, value in inputs_dict.items():
-            if isinstance(value, torch.Tensor):
-                inputs_on_device[key] = value.to(device)
-            else:
-                inputs_on_device[key] = value
+        inputs_on_device = {k: v.to(device) if isinstance(v, torch.Tensor) else v for k, v in inputs_dict.items()}
 
         # Enable context parallelism
         cp_config = ContextParallelConfig(**cp_dict)
@@ -89,6 +84,59 @@ def _context_parallel_worker(rank, world_size, master_port, model_class, init_di
             dist.destroy_process_group()
 
 
+def _custom_mesh_worker(
+    rank,
+    world_size,
+    master_port,
+    model_class,
+    init_dict,
+    cp_dict,
+    mesh_shape,
+    mesh_dim_names,
+    inputs_dict,
+    return_dict,
+):
+    """Worker function for context parallel testing with a user-provided custom DeviceMesh."""
+    try:
+        os.environ["MASTER_ADDR"] = "localhost"
+        os.environ["MASTER_PORT"] = str(master_port)
+        os.environ["RANK"] = str(rank)
+        os.environ["WORLD_SIZE"] = str(world_size)
+
+        dist.init_process_group(backend="nccl", rank=rank, world_size=world_size)
+
+        torch.cuda.set_device(rank)
+        device = torch.device(f"cuda:{rank}")
+
+        model = model_class(**init_dict)
+        model.to(device)
+        model.eval()
+
+        inputs_on_device = {k: v.to(device) if isinstance(v, torch.Tensor) else v for k, v in inputs_dict.items()}
+
+        # DeviceMesh must be created after init_process_group, inside each worker process.
+        mesh = torch.distributed.device_mesh.init_device_mesh(
+            "cuda", mesh_shape=mesh_shape, mesh_dim_names=mesh_dim_names
+        )
+        cp_config = ContextParallelConfig(**cp_dict, mesh=mesh)
+        model.enable_parallelism(config=cp_config)
+
+        with torch.no_grad():
+            output = model(**inputs_on_device, return_dict=False)[0]
+
+        if rank == 0:
+            return_dict["status"] = "success"
+            return_dict["output_shape"] = list(output.shape)
+
+    except Exception as e:
+        if rank == 0:
+            return_dict["status"] = "error"
+            return_dict["error"] = str(e)
+    finally:
+        if dist.is_initialized():
+            dist.destroy_process_group()
+
+
 @is_context_parallel
 @require_torch_multi_accelerator
 class ContextParallelTesterMixin:
@@ -126,3 +174,48 @@ class ContextParallelTesterMixin:
         assert return_dict.get("status") == "success", (
             f"Context parallel inference failed: {return_dict.get('error', 'Unknown error')}"
         )
+
+    @pytest.mark.parametrize(
+        "cp_type,mesh_shape,mesh_dim_names",
+        [
+            ("ring_degree", (2, 1, 1), ("ring", "ulysses", "fsdp")),
+            ("ulysses_degree", (1, 2, 1), ("ring", "ulysses", "fsdp")),
+        ],
+        ids=["ring-3d-fsdp", "ulysses-3d-fsdp"],
+    )
+    def test_context_parallel_custom_mesh(self, cp_type, mesh_shape, mesh_dim_names):
+        if not torch.distributed.is_available():
+            pytest.skip("torch.distributed is not available.")
+
+        if not hasattr(self.model_class, "_cp_plan") or self.model_class._cp_plan is None:
+            pytest.skip("Model does not have a _cp_plan defined for context parallel inference.")
+
+        world_size = 2
+        init_dict = self.get_init_dict()
+        inputs_dict = {k: v.cpu() if isinstance(v, torch.Tensor) else v for k, v in self.get_dummy_inputs().items()}
+        cp_dict = {cp_type: world_size}
+
+        master_port = _find_free_port()
+        manager = mp.Manager()
+        return_dict = manager.dict()
+
+        mp.spawn(
+            _custom_mesh_worker,
+            args=(
+                world_size,
+                master_port,
+                self.model_class,
+                init_dict,
+                cp_dict,
+                mesh_shape,
+                mesh_dim_names,
+                inputs_dict,
+                return_dict,
+            ),
+            nprocs=world_size,
+            join=True,
+        )
+
+        assert return_dict.get("status") == "success", (
+            f"Custom mesh context parallel inference failed: {return_dict.get('error', 'Unknown error')}"
+        )