start

* device_map tests for all models.

* updates

* Update tests/models/test_modeling_common.py

Co-authored-by: Aryan <aryan@huggingface.co>

* fix device_map in test

---------

Co-authored-by: Aryan <aryan@huggingface.co>

docs/source/en/_toctree.yml

@@ -180,6 +180,8 @@
     title: Caching
   - local: optimization/memory
     title: Reduce memory usage
+  - local: optimization/pruna
+    title: Pruna
   - local: optimization/xformers
     title: xFormers
   - local: optimization/tome

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

@@ -36,6 +36,22 @@ Make sure to check out the Schedulers [guide](../../using-diffusers/schedulers)
   - all
   - __call__
 
+## Cosmos2TextToImagePipeline
+
+[[autodoc]] Cosmos2TextToImagePipeline
+  - all
+  - __call__
+
+## Cosmos2VideoToWorldPipeline
+
+[[autodoc]] Cosmos2VideoToWorldPipeline
+  - all
+  - __call__
+
 ## CosmosPipelineOutput
 
 [[autodoc]] pipelines.cosmos.pipeline_output.CosmosPipelineOutput
+
+## CosmosImagePipelineOutput
+
+[[autodoc]] pipelines.cosmos.pipeline_output.CosmosImagePipelineOutput

docs/source/en/optimization/pruna.md

@@ -0,0 +1,187 @@
+# Pruna
+
+[Pruna](https://github.com/PrunaAI/pruna) is a model optimization framework that offers various optimization methods - quantization, pruning, caching, compilation - for accelerating inference and reducing memory usage. A general overview of the optimization methods are shown below.
+
+
+| Technique    | Description                                                                                   | Speed | Memory | Quality |
+|--------------|-----------------------------------------------------------------------------------------------|:-----:|:------:|:-------:|
+| `batcher`    | Groups multiple inputs together to be processed simultaneously, improving computational efficiency and reducing processing time. | ✅    | ❌     | ➖      |
+| `cacher`     | Stores intermediate results of computations to speed up subsequent operations.               | ✅    | ➖     | ➖      |
+| `compiler`   | Optimises the model with instructions for specific hardware.                                 | ✅    | ➖     | ➖      |
+| `distiller`  | Trains a smaller, simpler model to mimic a larger, more complex model.                       | ✅    | ✅     | ❌      |
+| `quantizer`  | Reduces the precision of weights and activations, lowering memory requirements.              | ✅    | ✅     | ❌      |
+| `pruner`     | Removes less important or redundant connections and neurons, resulting in a sparser, more efficient network. | ✅    | ✅     | ❌      |
+| `recoverer`  | Restores the performance of a model after compression.                                       | ➖    | ➖     | ✅      |
+| `factorizer` | Factorization batches several small matrix multiplications into one large fused operation. | ✅ | ➖ | ➖ |
+| `enhancer`   | Enhances the model output by applying post-processing algorithms such as denoising or upscaling. | ❌ | - | ✅ |
+
+✅ (improves), ➖ (approx. the same), ❌ (worsens)
+
+Explore the full range of optimization methods in the [Pruna documentation](https://docs.pruna.ai/en/stable/docs_pruna/user_manual/configure.html#configure-algorithms).
+
+## Installation
+
+Install Pruna with the following command.
+
+```bash
+pip install pruna
+```
+
+
+## Optimize Diffusers models
+
+A broad range of optimization algorithms are supported for Diffusers models as shown below.
+
+<div class="flex justify-center">
+    <img src="https://huggingface.co/datasets/PrunaAI/documentation-images/resolve/main/diffusers/diffusers_combinations.png" alt="Overview of the supported optimization algorithms for diffusers models">
+</div>
+
+The example below optimizes [black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev)
+with a combination of factorizer, compiler, and cacher algorithms. This combination accelerates inference by up to 4.2x and cuts peak GPU memory usage from 34.7GB to 28.0GB, all while maintaining virtually the same output quality.
+
+> [!TIP]
+> Refer to the [Pruna optimization](https://docs.pruna.ai/en/stable/docs_pruna/user_manual/configure.html) docs to learn more about the optimization techniques used in this example.
+
+<div class="flex justify-center">
+    <img src="https://huggingface.co/datasets/PrunaAI/documentation-images/resolve/main/diffusers/flux_combination.png" alt="Optimization techniques used for FLUX.1-dev showing the combination of factorizer, compiler, and cacher algorithms">
+</div>
+
+Start by defining a `SmashConfig` with the optimization algorithms to use. To optimize the model, wrap the pipeline and the `SmashConfig` with `smash` and then use the pipeline as normal for inference.
+
+```python
+import torch
+from diffusers import FluxPipeline
+
+from pruna import PrunaModel, SmashConfig, smash
+
+# load the model
+# Try segmind/Segmind-Vega or black-forest-labs/FLUX.1-schnell with a small GPU memory
+pipe = FluxPipeline.from_pretrained(
+    "black-forest-labs/FLUX.1-dev",
+    torch_dtype=torch.bfloat16
+).to("cuda")
+
+# define the configuration
+smash_config = SmashConfig()
+smash_config["factorizer"] = "qkv_diffusers"
+smash_config["compiler"] = "torch_compile"
+smash_config["torch_compile_target"] = "module_list"
+smash_config["cacher"] = "fora"
+smash_config["fora_interval"] = 2
+
+# for the best results in terms of speed you can add these configs
+# however they will increase your warmup time from 1.5 min to 10 min
+# smash_config["torch_compile_mode"] = "max-autotune-no-cudagraphs"
+# smash_config["quantizer"] = "torchao"
+# smash_config["torchao_quant_type"] = "fp8dq"
+# smash_config["torchao_excluded_modules"] = "norm+embedding"
+
+# optimize the model
+smashed_pipe = smash(pipe, smash_config)
+
+# run the model
+smashed_pipe("a knitted purple prune").images[0]
+```
+
+<div class="flex justify-center">
+    <img src="https://huggingface.co/datasets/PrunaAI/documentation-images/resolve/main/diffusers/flux_smashed_comparison.png">
+</div>
+
+After optimization, we can share and load the optimized model using the Hugging Face Hub.
+
+```python
+# save the model
+smashed_pipe.save_to_hub("<username>/FLUX.1-dev-smashed")
+
+# load the model
+smashed_pipe = PrunaModel.from_hub("<username>/FLUX.1-dev-smashed")
+```
+
+## Evaluate and benchmark Diffusers models
+
+Pruna provides the [EvaluationAgent](https://docs.pruna.ai/en/stable/docs_pruna/user_manual/evaluate.html) to evaluate the quality of your optimized models.
+
+We can metrics we care about, such as total time and throughput, and the dataset to evaluate on. We can define a model and pass it to the `EvaluationAgent`.
+
+<hfoptions id="eval">
+<hfoption id="optimized model">
+
+We can load and evaluate an optimized model by using the `EvaluationAgent` and pass it to the `Task`.
+
+```python
+import torch
+from diffusers import FluxPipeline
+
+from pruna import PrunaModel
+from pruna.data.pruna_datamodule import PrunaDataModule
+from pruna.evaluation.evaluation_agent import EvaluationAgent
+from pruna.evaluation.metrics import (
+    ThroughputMetric,
+    TorchMetricWrapper,
+    TotalTimeMetric,
+)
+from pruna.evaluation.task import Task
+
+# define the device
+device = "cuda" if torch.cuda.is_available() else "mps" if torch.backends.mps.is_available() else "cpu"
+
+# load the model
+# Try PrunaAI/Segmind-Vega-smashed or PrunaAI/FLUX.1-dev-smashed with a small GPU memory
+smashed_pipe = PrunaModel.from_hub("PrunaAI/FLUX.1-dev-smashed")
+
+# Define the metrics
+metrics = [
+    TotalTimeMetric(n_iterations=20, n_warmup_iterations=5),
+    ThroughputMetric(n_iterations=20, n_warmup_iterations=5),
+    TorchMetricWrapper("clip"),
+]
+
+# Define the datamodule
+datamodule = PrunaDataModule.from_string("LAION256")
+datamodule.limit_datasets(10)
+
+# Define the task and evaluation agent
+task = Task(metrics, datamodule=datamodule, device=device)
+eval_agent = EvaluationAgent(task)
+
+# Evaluate smashed model and offload it to CPU
+smashed_pipe.move_to_device(device)
+smashed_pipe_results = eval_agent.evaluate(smashed_pipe)
+smashed_pipe.move_to_device("cpu")
+```
+
+</hfoption>
+<hfoption id="standalone model">
+
+Instead of comparing the optimized model to the base model, you can also evaluate the standalone `diffusers` model. This is useful if you want to evaluate the performance of the model without the optimization. We can do so by using the `PrunaModel` wrapper and run the `EvaluationAgent` on it.
+
+```python
+import torch
+from diffusers import FluxPipeline
+
+from pruna import PrunaModel
+
+# load the model
+# Try PrunaAI/Segmind-Vega-smashed or PrunaAI/FLUX.1-dev-smashed with a small GPU memory
+pipe = FluxPipeline.from_pretrained(
+    "black-forest-labs/FLUX.1-dev",
+    torch_dtype=torch.bfloat16
+).to("cpu")
+wrapped_pipe = PrunaModel(model=pipe)
+```
+
+</hfoption>
+</hfoptions>
+
+Now that you have seen how to optimize and evaluate your models, you can start using Pruna to optimize your own models. Luckily, we have many examples to help you get started.
+
+> [!TIP]
+> For more details about benchmarking Flux, check out the [Announcing FLUX-Juiced: The Fastest Image Generation Endpoint (2.6 times faster)!](https://huggingface.co/blog/PrunaAI/flux-fastest-image-generation-endpoint) blog post and the [InferBench](https://huggingface.co/spaces/PrunaAI/InferBench) Space.
+
+## Reference
+
+- [Pruna](https://github.com/pruna-ai/pruna)
+- [Pruna optimization](https://docs.pruna.ai/en/stable/docs_pruna/user_manual/configure.html#configure-algorithms)
+- [Pruna evaluation](https://docs.pruna.ai/en/stable/docs_pruna/user_manual/evaluate.html)
+- [Pruna tutorials](https://docs.pruna.ai/en/stable/docs_pruna/tutorials/index.html)
+

examples/advanced_diffusion_training/README_flux.md

@@ -76,6 +76,24 @@ This command will prompt you for a token. Copy-paste yours from your [settings/t
 > `pip install wandb`
 > Alternatively, you can use other tools / train without reporting by modifying the flag  `--report_to="wandb"`.
 
+### LoRA Rank and Alpha
+Two key LoRA hyperparameters are LoRA rank and LoRA alpha. 
+- `--rank`: Defines the dimension of the trainable LoRA matrices. A higher rank means more expressiveness and capacity to learn (and more parameters).
+- `--lora_alpha`: A scaling factor for the LoRA's output. The LoRA update is scaled by lora_alpha / lora_rank.
+- lora_alpha vs. rank:
+This ratio dictates the LoRA's effective strength:
+lora_alpha == rank: Scaling factor is 1. The LoRA is applied with its learned strength. (e.g., alpha=16, rank=16)
+lora_alpha < rank: Scaling factor < 1. Reduces the LoRA's impact. Useful for subtle changes or to prevent overpowering the base model. (e.g., alpha=8, rank=16)
+lora_alpha > rank: Scaling factor > 1. Amplifies the LoRA's impact. Allows a lower rank LoRA to have a stronger effect. (e.g., alpha=32, rank=16)
+
+> [!TIP]
+> A common starting point is to set `lora_alpha` equal to `rank`. 
+> Some also set `lora_alpha` to be twice the `rank` (e.g., lora_alpha=32 for lora_rank=16) 
+> to give the LoRA updates more influence without increasing parameter count. 
+> If you find your LoRA is "overcooking" or learning too aggressively, consider setting `lora_alpha` to half of `rank` 
+> (e.g., lora_alpha=8 for rank=16). Experimentation is often key to finding the optimal balance for your use case.
+
+
 ### Target Modules
 When LoRA was first adapted from language models to diffusion models, it was applied to the cross-attention layers in the Unet that relate the image representations with the prompts that describe them. 
 More recently, SOTA text-to-image diffusion models replaced the Unet with a diffusion Transformer(DiT). With this change, we may also want to explore 

examples/advanced_diffusion_training/test_dreambooth_lora_flux_advanced.py

@@ -13,6 +13,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+import json
 import logging
 import os
 import sys
@@ -20,6 +21,8 @@ import tempfile
 
 import safetensors
 
+from diffusers.loaders.lora_base import LORA_ADAPTER_METADATA_KEY
+
 
 sys.path.append("..")
 from test_examples_utils import ExamplesTestsAccelerate, run_command  # noqa: E402
@@ -281,3 +284,45 @@ class DreamBoothLoRAFluxAdvanced(ExamplesTestsAccelerate):
             run_command(self._launch_args + resume_run_args)
 
             self.assertEqual({x for x in os.listdir(tmpdir) if "checkpoint" in x}, {"checkpoint-6", "checkpoint-8"})
+
+    def test_dreambooth_lora_with_metadata(self):
+        # Use a `lora_alpha` that is different from `rank`.
+        lora_alpha = 8
+        rank = 4
+        with tempfile.TemporaryDirectory() as tmpdir:
+            test_args = f"""
+                {self.script_path}
+                --pretrained_model_name_or_path {self.pretrained_model_name_or_path}
+                --instance_data_dir {self.instance_data_dir}
+                --instance_prompt {self.instance_prompt}
+                --resolution 64
+                --train_batch_size 1
+                --gradient_accumulation_steps 1
+                --max_train_steps 2
+                --lora_alpha={lora_alpha}
+                --rank={rank}
+                --learning_rate 5.0e-04
+                --scale_lr
+                --lr_scheduler constant
+                --lr_warmup_steps 0
+                --output_dir {tmpdir}
+                """.split()
+
+            run_command(self._launch_args + test_args)
+            # save_pretrained smoke test
+            state_dict_file = os.path.join(tmpdir, "pytorch_lora_weights.safetensors")
+            self.assertTrue(os.path.isfile(state_dict_file))
+
+            # Check if the metadata was properly serialized.
+            with safetensors.torch.safe_open(state_dict_file, framework="pt", device="cpu") as f:
+                metadata = f.metadata() or {}
+
+            metadata.pop("format", None)
+            raw = metadata.get(LORA_ADAPTER_METADATA_KEY)
+            if raw:
+                raw = json.loads(raw)
+
+            loaded_lora_alpha = raw["transformer.lora_alpha"]
+            self.assertTrue(loaded_lora_alpha == lora_alpha)
+            loaded_lora_rank = raw["transformer.r"]
+            self.assertTrue(loaded_lora_rank == rank)

examples/advanced_diffusion_training/train_dreambooth_lora_flux_advanced.py

@@ -55,6 +55,7 @@ from diffusers import (
 )
 from diffusers.optimization import get_scheduler
 from diffusers.training_utils import (
+    _collate_lora_metadata,
     _set_state_dict_into_text_encoder,
     cast_training_params,
     compute_density_for_timestep_sampling,
@@ -431,6 +432,13 @@ def parse_args(input_args=None):
         help=("The dimension of the LoRA update matrices."),
     )
 
+    parser.add_argument(
+        "--lora_alpha",
+        type=int,
+        default=4,
+        help="LoRA alpha to be used for additional scaling.",
+    )
+
     parser.add_argument("--lora_dropout", type=float, default=0.0, help="Dropout probability for LoRA layers")
 
     parser.add_argument(
@@ -1556,7 +1564,7 @@ def main(args):
     # now we will add new LoRA weights to the attention layers
     transformer_lora_config = LoraConfig(
         r=args.rank,
-        lora_alpha=args.rank,
+        lora_alpha=args.lora_alpha,
         lora_dropout=args.lora_dropout,
         init_lora_weights="gaussian",
         target_modules=target_modules,
@@ -1565,7 +1573,7 @@ def main(args):
     if args.train_text_encoder:
         text_lora_config = LoraConfig(
             r=args.rank,
-            lora_alpha=args.rank,
+            lora_alpha=args.lora_alpha,
             lora_dropout=args.lora_dropout,
             init_lora_weights="gaussian",
             target_modules=["q_proj", "k_proj", "v_proj", "out_proj"],
@@ -1582,13 +1590,15 @@ def main(args):
         if accelerator.is_main_process:
             transformer_lora_layers_to_save = None
             text_encoder_one_lora_layers_to_save = None
-
+            modules_to_save = {}
             for model in models:
                 if isinstance(model, type(unwrap_model(transformer))):
                     transformer_lora_layers_to_save = get_peft_model_state_dict(model)
+                    modules_to_save["transformer"] = model
                 elif isinstance(model, type(unwrap_model(text_encoder_one))):
                     if args.train_text_encoder:  # when --train_text_encoder_ti we don't save the layers
                         text_encoder_one_lora_layers_to_save = get_peft_model_state_dict(model)
+                        modules_to_save["text_encoder"] = model
                 elif isinstance(model, type(unwrap_model(text_encoder_two))):
                     pass  # when --train_text_encoder_ti and --enable_t5_ti we don't save the layers
                 else:
@@ -1601,6 +1611,7 @@ def main(args):
                 output_dir,
                 transformer_lora_layers=transformer_lora_layers_to_save,
                 text_encoder_lora_layers=text_encoder_one_lora_layers_to_save,
+                **_collate_lora_metadata(modules_to_save),
             )
         if args.train_text_encoder_ti:
             embedding_handler.save_embeddings(f"{args.output_dir}/{Path(args.output_dir).name}_emb.safetensors")
@@ -2359,16 +2370,19 @@ def main(args):
     # Save the lora layers
     accelerator.wait_for_everyone()
     if accelerator.is_main_process:
+        modules_to_save = {}
         transformer = unwrap_model(transformer)
         if args.upcast_before_saving:
             transformer.to(torch.float32)
         else:
             transformer = transformer.to(weight_dtype)
         transformer_lora_layers = get_peft_model_state_dict(transformer)
+        modules_to_save["transformer"] = transformer
 
         if args.train_text_encoder:
             text_encoder_one = unwrap_model(text_encoder_one)
             text_encoder_lora_layers = get_peft_model_state_dict(text_encoder_one.to(torch.float32))
+            modules_to_save["text_encoder"] = text_encoder_one
         else:
             text_encoder_lora_layers = None
 
@@ -2377,6 +2391,7 @@ def main(args):
                 save_directory=args.output_dir,
                 transformer_lora_layers=transformer_lora_layers,
                 text_encoder_lora_layers=text_encoder_lora_layers,
+                **_collate_lora_metadata(modules_to_save),
             )
 
         if args.train_text_encoder_ti:

examples/dreambooth/README_flux.md

@@ -170,6 +170,23 @@ accelerate launch train_dreambooth_lora_flux.py \
   --push_to_hub
 ```
 
+### LoRA Rank and Alpha
+Two key LoRA hyperparameters are LoRA rank and LoRA alpha. 
+- `--rank`: Defines the dimension of the trainable LoRA matrices. A higher rank means more expressiveness and capacity to learn (and more parameters).
+- `--lora_alpha`: A scaling factor for the LoRA's output. The LoRA update is scaled by lora_alpha / lora_rank.
+- lora_alpha vs. rank:
+This ratio dictates the LoRA's effective strength:
+lora_alpha == rank: Scaling factor is 1. The LoRA is applied with its learned strength. (e.g., alpha=16, rank=16)
+lora_alpha < rank: Scaling factor < 1. Reduces the LoRA's impact. Useful for subtle changes or to prevent overpowering the base model. (e.g., alpha=8, rank=16)
+lora_alpha > rank: Scaling factor > 1. Amplifies the LoRA's impact. Allows a lower rank LoRA to have a stronger effect. (e.g., alpha=32, rank=16)
+
+> [!TIP]
+> A common starting point is to set `lora_alpha` equal to `rank`. 
+> Some also set `lora_alpha` to be twice the `rank` (e.g., lora_alpha=32 for lora_rank=16) 
+> to give the LoRA updates more influence without increasing parameter count. 
+> If you find your LoRA is "overcooking" or learning too aggressively, consider setting `lora_alpha` to half of `rank` 
+> (e.g., lora_alpha=8 for rank=16). Experimentation is often key to finding the optimal balance for your use case.
+
 ### Target Modules
 When LoRA was first adapted from language models to diffusion models, it was applied to the cross-attention layers in the Unet that relate the image representations with the prompts that describe them. 
 More recently, SOTA text-to-image diffusion models replaced the Unet with a diffusion Transformer(DiT). With this change, we may also want to explore 

examples/dreambooth/test_dreambooth_lora_flux.py

@@ -13,6 +13,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+import json
 import logging
 import os
 import sys
@@ -20,6 +21,8 @@ import tempfile
 
 import safetensors
 
+from diffusers.loaders.lora_base import LORA_ADAPTER_METADATA_KEY
+
 
 sys.path.append("..")
 from test_examples_utils import ExamplesTestsAccelerate, run_command  # noqa: E402
@@ -234,3 +237,45 @@ class DreamBoothLoRAFlux(ExamplesTestsAccelerate):
             run_command(self._launch_args + resume_run_args)
 
             self.assertEqual({x for x in os.listdir(tmpdir) if "checkpoint" in x}, {"checkpoint-6", "checkpoint-8"})
+
+    def test_dreambooth_lora_with_metadata(self):
+        # Use a `lora_alpha` that is different from `rank`.
+        lora_alpha = 8
+        rank = 4
+        with tempfile.TemporaryDirectory() as tmpdir:
+            test_args = f"""
+                {self.script_path}
+                --pretrained_model_name_or_path {self.pretrained_model_name_or_path}
+                --instance_data_dir {self.instance_data_dir}
+                --instance_prompt {self.instance_prompt}
+                --resolution 64
+                --train_batch_size 1
+                --gradient_accumulation_steps 1
+                --max_train_steps 2
+                --lora_alpha={lora_alpha}
+                --rank={rank}
+                --learning_rate 5.0e-04
+                --scale_lr
+                --lr_scheduler constant
+                --lr_warmup_steps 0
+                --output_dir {tmpdir}
+                """.split()
+
+            run_command(self._launch_args + test_args)
+            # save_pretrained smoke test
+            state_dict_file = os.path.join(tmpdir, "pytorch_lora_weights.safetensors")
+            self.assertTrue(os.path.isfile(state_dict_file))
+
+            # Check if the metadata was properly serialized.
+            with safetensors.torch.safe_open(state_dict_file, framework="pt", device="cpu") as f:
+                metadata = f.metadata() or {}
+
+            metadata.pop("format", None)
+            raw = metadata.get(LORA_ADAPTER_METADATA_KEY)
+            if raw:
+                raw = json.loads(raw)
+
+            loaded_lora_alpha = raw["transformer.lora_alpha"]
+            self.assertTrue(loaded_lora_alpha == lora_alpha)
+            loaded_lora_rank = raw["transformer.r"]
+            self.assertTrue(loaded_lora_rank == rank)

examples/dreambooth/train_dreambooth_lora_flux.py

@@ -27,7 +27,6 @@ from pathlib import Path
 
 import numpy as np
 import torch
-import torch.utils.checkpoint
 import transformers
 from accelerate import Accelerator
 from accelerate.logging import get_logger
@@ -53,6 +52,7 @@ from diffusers import (
 )
 from diffusers.optimization import get_scheduler
 from diffusers.training_utils import (
+    _collate_lora_metadata,
     _set_state_dict_into_text_encoder,
     cast_training_params,
     compute_density_for_timestep_sampling,
@@ -358,7 +358,12 @@ def parse_args(input_args=None):
         default=4,
         help=("The dimension of the LoRA update matrices."),
     )
-
+    parser.add_argument(
+        "--lora_alpha",
+        type=int,
+        default=4,
+        help="LoRA alpha to be used for additional scaling.",
+    )
     parser.add_argument("--lora_dropout", type=float, default=0.0, help="Dropout probability for LoRA layers")
 
     parser.add_argument(
@@ -1238,7 +1243,7 @@ def main(args):
     # now we will add new LoRA weights the transformer layers
     transformer_lora_config = LoraConfig(
         r=args.rank,
-        lora_alpha=args.rank,
+        lora_alpha=args.lora_alpha,
         lora_dropout=args.lora_dropout,
         init_lora_weights="gaussian",
         target_modules=target_modules,
@@ -1247,7 +1252,7 @@ def main(args):
     if args.train_text_encoder:
         text_lora_config = LoraConfig(
             r=args.rank,
-            lora_alpha=args.rank,
+            lora_alpha=args.lora_alpha,
             lora_dropout=args.lora_dropout,
             init_lora_weights="gaussian",
             target_modules=["q_proj", "k_proj", "v_proj", "out_proj"],
@@ -1264,12 +1269,14 @@ def main(args):
         if accelerator.is_main_process:
             transformer_lora_layers_to_save = None
             text_encoder_one_lora_layers_to_save = None
-
+            modules_to_save = {}
             for model in models:
                 if isinstance(model, type(unwrap_model(transformer))):
                     transformer_lora_layers_to_save = get_peft_model_state_dict(model)
+                    modules_to_save["transformer"] = model
                 elif isinstance(model, type(unwrap_model(text_encoder_one))):
                     text_encoder_one_lora_layers_to_save = get_peft_model_state_dict(model)
+                    modules_to_save["text_encoder"] = model
                 else:
                     raise ValueError(f"unexpected save model: {model.__class__}")
 
@@ -1280,6 +1287,7 @@ def main(args):
                 output_dir,
                 transformer_lora_layers=transformer_lora_layers_to_save,
                 text_encoder_lora_layers=text_encoder_one_lora_layers_to_save,
+                **_collate_lora_metadata(modules_to_save),
             )
 
     def load_model_hook(models, input_dir):
@@ -1889,16 +1897,19 @@ def main(args):
     # Save the lora layers
     accelerator.wait_for_everyone()
     if accelerator.is_main_process:
+        modules_to_save = {}
         transformer = unwrap_model(transformer)
         if args.upcast_before_saving:
             transformer.to(torch.float32)
         else:
             transformer = transformer.to(weight_dtype)
         transformer_lora_layers = get_peft_model_state_dict(transformer)
+        modules_to_save["transformer"] = transformer
 
         if args.train_text_encoder:
             text_encoder_one = unwrap_model(text_encoder_one)
             text_encoder_lora_layers = get_peft_model_state_dict(text_encoder_one.to(torch.float32))
+            modules_to_save["text_encoder"] = text_encoder_one
         else:
             text_encoder_lora_layers = None
 
@@ -1906,6 +1917,7 @@ def main(args):
             save_directory=args.output_dir,
             transformer_lora_layers=transformer_lora_layers,
             text_encoder_lora_layers=text_encoder_lora_layers,
+            **_collate_lora_metadata(modules_to_save),
         )
 
         # Final inference

examples/dreambooth/train_dreambooth_lora_hidream.py

@@ -29,7 +29,7 @@ from pathlib import Path
 import numpy as np
 import torch
 import transformers
-from accelerate import Accelerator
+from accelerate import Accelerator, DistributedType
 from accelerate.logging import get_logger
 from accelerate.utils import DistributedDataParallelKwargs, ProjectConfiguration, set_seed
 from huggingface_hub import create_repo, upload_folder
@@ -1181,13 +1181,15 @@ def main(args):
             transformer_lora_layers_to_save = None
 
             for model in models:
-                if isinstance(model, type(unwrap_model(transformer))):
+                if isinstance(unwrap_model(model), type(unwrap_model(transformer))):
+                    model = unwrap_model(model)
                     transformer_lora_layers_to_save = get_peft_model_state_dict(model)
                 else:
                     raise ValueError(f"unexpected save model: {model.__class__}")
 
                 # make sure to pop weight so that corresponding model is not saved again
-                weights.pop()
+                if weights:
+                    weights.pop()
 
             HiDreamImagePipeline.save_lora_weights(
                 output_dir,
@@ -1197,13 +1199,20 @@ def main(args):
     def load_model_hook(models, input_dir):
         transformer_ = None
 
-        while len(models) > 0:
-            model = models.pop()
+        if not accelerator.distributed_type == DistributedType.DEEPSPEED:
+            while len(models) > 0:
+                model = models.pop()
 
-            if isinstance(model, type(unwrap_model(transformer))):
-                transformer_ = model
-            else:
-                raise ValueError(f"unexpected save model: {model.__class__}")
+                if isinstance(unwrap_model(model), type(unwrap_model(transformer))):
+                    model = unwrap_model(model)
+                    transformer_ = model
+                else:
+                    raise ValueError(f"unexpected save model: {model.__class__}")
+        else:
+            transformer_ = HiDreamImageTransformer2DModel.from_pretrained(
+                args.pretrained_model_name_or_path, subfolder="transformer"
+            )
+            transformer_.add_adapter(transformer_lora_config)
 
         lora_state_dict = HiDreamImagePipeline.lora_state_dict(input_dir)
 
@@ -1655,7 +1664,7 @@ def main(args):
                 progress_bar.update(1)
                 global_step += 1
 
-                if accelerator.is_main_process:
+                if accelerator.is_main_process or accelerator.distributed_type == DistributedType.DEEPSPEED:
                     if global_step % args.checkpointing_steps == 0:
                         # _before_ saving state, check if this save would set us over the `checkpoints_total_limit`
                         if args.checkpoints_total_limit is not None:

scripts/convert_cosmos_to_diffusers.py

@@ -7,7 +7,17 @@ from accelerate import init_empty_weights
 from huggingface_hub import snapshot_download
 from transformers import T5EncoderModel, T5TokenizerFast
 
-from diffusers import AutoencoderKLCosmos, CosmosTextToWorldPipeline, CosmosTransformer3DModel, EDMEulerScheduler
+from diffusers import (
+    AutoencoderKLCosmos,
+    AutoencoderKLWan,
+    Cosmos2TextToImagePipeline,
+    Cosmos2VideoToWorldPipeline,
+    CosmosTextToWorldPipeline,
+    CosmosTransformer3DModel,
+    CosmosVideoToWorldPipeline,
+    EDMEulerScheduler,
+    FlowMatchEulerDiscreteScheduler,
+)
 
 
 def remove_keys_(key: str, state_dict: Dict[str, Any]):
@@ -29,7 +39,7 @@ def rename_transformer_blocks_(key: str, state_dict: Dict[str, Any]):
     state_dict[new_key] = state_dict.pop(key)
 
 
-TRANSFORMER_KEYS_RENAME_DICT = {
+TRANSFORMER_KEYS_RENAME_DICT_COSMOS_1_0 = {
     "t_embedder.1": "time_embed.t_embedder",
     "affline_norm": "time_embed.norm",
     ".blocks.0.block.attn": ".attn1",
@@ -56,7 +66,7 @@ TRANSFORMER_KEYS_RENAME_DICT = {
     "final_layer.linear": "proj_out",
 }
 
-TRANSFORMER_SPECIAL_KEYS_REMAP = {
+TRANSFORMER_SPECIAL_KEYS_REMAP_COSMOS_1_0 = {
     "blocks.block": rename_transformer_blocks_,
     "logvar.0.freqs": remove_keys_,
     "logvar.0.phases": remove_keys_,
@@ -64,6 +74,45 @@ TRANSFORMER_SPECIAL_KEYS_REMAP = {
     "pos_embedder.seq": remove_keys_,
 }
 
+TRANSFORMER_KEYS_RENAME_DICT_COSMOS_2_0 = {
+    "t_embedder.1": "time_embed.t_embedder",
+    "t_embedding_norm": "time_embed.norm",
+    "blocks": "transformer_blocks",
+    "adaln_modulation_self_attn.1": "norm1.linear_1",
+    "adaln_modulation_self_attn.2": "norm1.linear_2",
+    "adaln_modulation_cross_attn.1": "norm2.linear_1",
+    "adaln_modulation_cross_attn.2": "norm2.linear_2",
+    "adaln_modulation_mlp.1": "norm3.linear_1",
+    "adaln_modulation_mlp.2": "norm3.linear_2",
+    "self_attn": "attn1",
+    "cross_attn": "attn2",
+    "q_proj": "to_q",
+    "k_proj": "to_k",
+    "v_proj": "to_v",
+    "output_proj": "to_out.0",
+    "q_norm": "norm_q",
+    "k_norm": "norm_k",
+    "mlp.layer1": "ff.net.0.proj",
+    "mlp.layer2": "ff.net.2",
+    "x_embedder.proj.1": "patch_embed.proj",
+    # "extra_pos_embedder": "learnable_pos_embed",
+    "final_layer.adaln_modulation.1": "norm_out.linear_1",
+    "final_layer.adaln_modulation.2": "norm_out.linear_2",
+    "final_layer.linear": "proj_out",
+}
+
+TRANSFORMER_SPECIAL_KEYS_REMAP_COSMOS_2_0 = {
+    "accum_video_sample_counter": remove_keys_,
+    "accum_image_sample_counter": remove_keys_,
+    "accum_iteration": remove_keys_,
+    "accum_train_in_hours": remove_keys_,
+    "pos_embedder.seq": remove_keys_,
+    "pos_embedder.dim_spatial_range": remove_keys_,
+    "pos_embedder.dim_temporal_range": remove_keys_,
+    "_extra_state": remove_keys_,
+}
+
+
 TRANSFORMER_CONFIGS = {
     "Cosmos-1.0-Diffusion-7B-Text2World": {
         "in_channels": 16,
@@ -125,6 +174,66 @@ TRANSFORMER_CONFIGS = {
         "concat_padding_mask": True,
         "extra_pos_embed_type": "learnable",
     },
+    "Cosmos-2.0-Diffusion-2B-Text2Image": {
+        "in_channels": 16,
+        "out_channels": 16,
+        "num_attention_heads": 16,
+        "attention_head_dim": 128,
+        "num_layers": 28,
+        "mlp_ratio": 4.0,
+        "text_embed_dim": 1024,
+        "adaln_lora_dim": 256,
+        "max_size": (128, 240, 240),
+        "patch_size": (1, 2, 2),
+        "rope_scale": (1.0, 4.0, 4.0),
+        "concat_padding_mask": True,
+        "extra_pos_embed_type": None,
+    },
+    "Cosmos-2.0-Diffusion-14B-Text2Image": {
+        "in_channels": 16,
+        "out_channels": 16,
+        "num_attention_heads": 40,
+        "attention_head_dim": 128,
+        "num_layers": 36,
+        "mlp_ratio": 4.0,
+        "text_embed_dim": 1024,
+        "adaln_lora_dim": 256,
+        "max_size": (128, 240, 240),
+        "patch_size": (1, 2, 2),
+        "rope_scale": (1.0, 4.0, 4.0),
+        "concat_padding_mask": True,
+        "extra_pos_embed_type": None,
+    },
+    "Cosmos-2.0-Diffusion-2B-Video2World": {
+        "in_channels": 16 + 1,
+        "out_channels": 16,
+        "num_attention_heads": 16,
+        "attention_head_dim": 128,
+        "num_layers": 28,
+        "mlp_ratio": 4.0,
+        "text_embed_dim": 1024,
+        "adaln_lora_dim": 256,
+        "max_size": (128, 240, 240),
+        "patch_size": (1, 2, 2),
+        "rope_scale": (1.0, 3.0, 3.0),
+        "concat_padding_mask": True,
+        "extra_pos_embed_type": None,
+    },
+    "Cosmos-2.0-Diffusion-14B-Video2World": {
+        "in_channels": 16 + 1,
+        "out_channels": 16,
+        "num_attention_heads": 40,
+        "attention_head_dim": 128,
+        "num_layers": 36,
+        "mlp_ratio": 4.0,
+        "text_embed_dim": 1024,
+        "adaln_lora_dim": 256,
+        "max_size": (128, 240, 240),
+        "patch_size": (1, 2, 2),
+        "rope_scale": (20 / 24, 2.0, 2.0),
+        "concat_padding_mask": True,
+        "extra_pos_embed_type": None,
+    },
 }
 
 VAE_KEYS_RENAME_DICT = {
@@ -216,9 +325,18 @@ def get_state_dict(saved_dict: Dict[str, Any]) -> Dict[str, Any]:
     return state_dict
 
 
-def convert_transformer(transformer_type: str, ckpt_path: str):
+def convert_transformer(transformer_type: str, ckpt_path: str, weights_only: bool = True):
     PREFIX_KEY = "net."
-    original_state_dict = get_state_dict(torch.load(ckpt_path, map_location="cpu", weights_only=True))
+    original_state_dict = get_state_dict(torch.load(ckpt_path, map_location="cpu", weights_only=weights_only))
+
+    if "Cosmos-1.0" in transformer_type:
+        TRANSFORMER_KEYS_RENAME_DICT = TRANSFORMER_KEYS_RENAME_DICT_COSMOS_1_0
+        TRANSFORMER_SPECIAL_KEYS_REMAP = TRANSFORMER_SPECIAL_KEYS_REMAP_COSMOS_1_0
+    elif "Cosmos-2.0" in transformer_type:
+        TRANSFORMER_KEYS_RENAME_DICT = TRANSFORMER_KEYS_RENAME_DICT_COSMOS_2_0
+        TRANSFORMER_SPECIAL_KEYS_REMAP = TRANSFORMER_SPECIAL_KEYS_REMAP_COSMOS_2_0
+    else:
+        assert False
 
     with init_empty_weights():
         config = TRANSFORMER_CONFIGS[transformer_type]
@@ -281,13 +399,61 @@ def convert_vae(vae_type: str):
     return vae
 
 
+def save_pipeline_cosmos_1_0(args, transformer, vae):
+    text_encoder = T5EncoderModel.from_pretrained(args.text_encoder_path, torch_dtype=torch.bfloat16)
+    tokenizer = T5TokenizerFast.from_pretrained(args.tokenizer_path)
+    # The original code initializes EDM config with sigma_min=0.0002, but does not make use of it anywhere directly.
+    # So, the sigma_min values that is used is the default value of 0.002.
+    scheduler = EDMEulerScheduler(
+        sigma_min=0.002,
+        sigma_max=80,
+        sigma_data=0.5,
+        sigma_schedule="karras",
+        num_train_timesteps=1000,
+        prediction_type="epsilon",
+        rho=7.0,
+        final_sigmas_type="sigma_min",
+    )
+
+    pipe_cls = CosmosTextToWorldPipeline if "Text2World" in args.transformer_type else CosmosVideoToWorldPipeline
+    pipe = pipe_cls(
+        text_encoder=text_encoder,
+        tokenizer=tokenizer,
+        transformer=transformer,
+        vae=vae,
+        scheduler=scheduler,
+        safety_checker=lambda *args, **kwargs: None,
+    )
+    pipe.save_pretrained(args.output_path, safe_serialization=True, max_shard_size="5GB")
+
+
+def save_pipeline_cosmos_2_0(args, transformer, vae):
+    text_encoder = T5EncoderModel.from_pretrained(args.text_encoder_path, torch_dtype=torch.bfloat16)
+    tokenizer = T5TokenizerFast.from_pretrained(args.tokenizer_path)
+
+    scheduler = FlowMatchEulerDiscreteScheduler(use_karras_sigmas=True)
+
+    pipe_cls = Cosmos2TextToImagePipeline if "Text2Image" in args.transformer_type else Cosmos2VideoToWorldPipeline
+    pipe = pipe_cls(
+        text_encoder=text_encoder,
+        tokenizer=tokenizer,
+        transformer=transformer,
+        vae=vae,
+        scheduler=scheduler,
+        safety_checker=lambda *args, **kwargs: None,
+    )
+    pipe.save_pretrained(args.output_path, safe_serialization=True, max_shard_size="5GB")
+
+
 def get_args():
     parser = argparse.ArgumentParser()
     parser.add_argument("--transformer_type", type=str, default=None, choices=list(TRANSFORMER_CONFIGS.keys()))
     parser.add_argument(
         "--transformer_ckpt_path", type=str, default=None, help="Path to original transformer checkpoint"
     )
-    parser.add_argument("--vae_type", type=str, default=None, choices=list(VAE_CONFIGS.keys()), help="Type of VAE")
+    parser.add_argument(
+        "--vae_type", type=str, default=None, choices=["none", *list(VAE_CONFIGS.keys())], help="Type of VAE"
+    )
     parser.add_argument("--text_encoder_path", type=str, default="google-t5/t5-11b")
     parser.add_argument("--tokenizer_path", type=str, default="google-t5/t5-11b")
     parser.add_argument("--save_pipeline", action="store_true")
@@ -316,37 +482,26 @@ if __name__ == "__main__":
         assert args.tokenizer_path is not None
 
     if args.transformer_ckpt_path is not None:
-        transformer = convert_transformer(args.transformer_type, args.transformer_ckpt_path)
+        weights_only = "Cosmos-1.0" in args.transformer_type
+        transformer = convert_transformer(args.transformer_type, args.transformer_ckpt_path, weights_only)
         transformer = transformer.to(dtype=dtype)
         if not args.save_pipeline:
             transformer.save_pretrained(args.output_path, safe_serialization=True, max_shard_size="5GB")
 
     if args.vae_type is not None:
-        vae = convert_vae(args.vae_type)
+        if "Cosmos-1.0" in args.transformer_type:
+            vae = convert_vae(args.vae_type)
+        else:
+            vae = AutoencoderKLWan.from_pretrained(
+                "Wan-AI/Wan2.1-T2V-1.3B-Diffusers", subfolder="vae", torch_dtype=torch.float32
+            )
         if not args.save_pipeline:
             vae.save_pretrained(args.output_path, safe_serialization=True, max_shard_size="5GB")
 
     if args.save_pipeline:
-        text_encoder = T5EncoderModel.from_pretrained(args.text_encoder_path, torch_dtype=dtype)
-        tokenizer = T5TokenizerFast.from_pretrained(args.tokenizer_path)
-        # The original code initializes EDM config with sigma_min=0.0002, but does not make use of it anywhere directly.
-        # So, the sigma_min values that is used is the default value of 0.002.
-        scheduler = EDMEulerScheduler(
-            sigma_min=0.002,
-            sigma_max=80,
-            sigma_data=0.5,
-            sigma_schedule="karras",
-            num_train_timesteps=1000,
-            prediction_type="epsilon",
-            rho=7.0,
-            final_sigmas_type="sigma_min",
-        )
-
-        pipe = CosmosTextToWorldPipeline(
-            text_encoder=text_encoder,
-            tokenizer=tokenizer,
-            transformer=transformer,
-            vae=vae,
-            scheduler=scheduler,
-        )
-        pipe.save_pretrained(args.output_path, safe_serialization=True, max_shard_size="5GB")
+        if "Cosmos-1.0" in args.transformer_type:
+            save_pipeline_cosmos_1_0(args, transformer, vae)
+        elif "Cosmos-2.0" in args.transformer_type:
+            save_pipeline_cosmos_2_0(args, transformer, vae)
+        else:
+            assert False

scripts/convert_stable_audio.py

@@ -1,4 +1,4 @@
-# Run this script to convert the Stable Cascade model weights to a diffusers pipeline.
+# Run this script to convert the Stable Audio model weights to a diffusers pipeline.
 import argparse
 import json
 import os

src/diffusers/__init__.py

@@ -363,6 +363,8 @@ else:
             "CogView4ControlPipeline",
             "CogView4Pipeline",
             "ConsisIDPipeline",
+            "Cosmos2TextToImagePipeline",
+            "Cosmos2VideoToWorldPipeline",
             "CosmosTextToWorldPipeline",
             "CosmosVideoToWorldPipeline",
             "CycleDiffusionPipeline",
@@ -953,6 +955,8 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             CogView4ControlPipeline,
             CogView4Pipeline,
             ConsisIDPipeline,
+            Cosmos2TextToImagePipeline,
+            Cosmos2VideoToWorldPipeline,
             CosmosTextToWorldPipeline,
             CosmosVideoToWorldPipeline,
             CycleDiffusionPipeline,

src/diffusers/loaders/lora_conversion_utils.py

@@ -1596,7 +1596,10 @@ def _convert_non_diffusers_wan_lora_to_diffusers(state_dict):
     converted_state_dict = {}
     original_state_dict = {k[len("diffusion_model.") :]: v for k, v in state_dict.items()}
 
-    num_blocks = len({k.split("blocks.")[1].split(".")[0] for k in original_state_dict if "blocks." in k})
+    block_numbers = {int(k.split(".")[1]) for k in original_state_dict if k.startswith("blocks.")}
+    min_block = min(block_numbers)
+    max_block = max(block_numbers)
+
     is_i2v_lora = any("k_img" in k for k in original_state_dict) and any("v_img" in k for k in original_state_dict)
     lora_down_key = "lora_A" if any("lora_A" in k for k in original_state_dict) else "lora_down"
     lora_up_key = "lora_B" if any("lora_B" in k for k in original_state_dict) else "lora_up"
@@ -1622,45 +1625,57 @@ def _convert_non_diffusers_wan_lora_to_diffusers(state_dict):
     # For the `diff_b` keys, we treat them as lora_bias.
     # https://huggingface.co/docs/peft/main/en/package_reference/lora#peft.LoraConfig.lora_bias
 
-    for i in range(num_blocks):
+    for i in range(min_block, max_block + 1):
         # Self-attention
         for o, c in zip(["q", "k", "v", "o"], ["to_q", "to_k", "to_v", "to_out.0"]):
-            converted_state_dict[f"blocks.{i}.attn1.{c}.lora_A.weight"] = original_state_dict.pop(
-                f"blocks.{i}.self_attn.{o}.{lora_down_key}.weight"
-            )
-            converted_state_dict[f"blocks.{i}.attn1.{c}.lora_B.weight"] = original_state_dict.pop(
-                f"blocks.{i}.self_attn.{o}.{lora_up_key}.weight"
-            )
-            if f"blocks.{i}.self_attn.{o}.diff_b" in original_state_dict:
-                converted_state_dict[f"blocks.{i}.attn1.{c}.lora_B.bias"] = original_state_dict.pop(
-                    f"blocks.{i}.self_attn.{o}.diff_b"
-                )
+            original_key = f"blocks.{i}.self_attn.{o}.{lora_down_key}.weight"
+            converted_key = f"blocks.{i}.attn1.{c}.lora_A.weight"
+            if original_key in original_state_dict:
+                converted_state_dict[converted_key] = original_state_dict.pop(original_key)
+
+            original_key = f"blocks.{i}.self_attn.{o}.{lora_up_key}.weight"
+            converted_key = f"blocks.{i}.attn1.{c}.lora_B.weight"
+            if original_key in original_state_dict:
+                converted_state_dict[converted_key] = original_state_dict.pop(original_key)
+
+            original_key = f"blocks.{i}.self_attn.{o}.diff_b"
+            converted_key = f"blocks.{i}.attn1.{c}.lora_B.bias"
+            if original_key in original_state_dict:
+                converted_state_dict[converted_key] = original_state_dict.pop(original_key)
 
         # Cross-attention
         for o, c in zip(["q", "k", "v", "o"], ["to_q", "to_k", "to_v", "to_out.0"]):
-            converted_state_dict[f"blocks.{i}.attn2.{c}.lora_A.weight"] = original_state_dict.pop(
-                f"blocks.{i}.cross_attn.{o}.{lora_down_key}.weight"
-            )
-            converted_state_dict[f"blocks.{i}.attn2.{c}.lora_B.weight"] = original_state_dict.pop(
-                f"blocks.{i}.cross_attn.{o}.{lora_up_key}.weight"
-            )
-            if f"blocks.{i}.cross_attn.{o}.diff_b" in original_state_dict:
-                converted_state_dict[f"blocks.{i}.attn2.{c}.lora_B.bias"] = original_state_dict.pop(
-                    f"blocks.{i}.cross_attn.{o}.diff_b"
-                )
+            original_key = f"blocks.{i}.cross_attn.{o}.{lora_down_key}.weight"
+            converted_key = f"blocks.{i}.attn2.{c}.lora_A.weight"
+            if original_key in original_state_dict:
+                converted_state_dict[converted_key] = original_state_dict.pop(original_key)
+
+            original_key = f"blocks.{i}.cross_attn.{o}.{lora_up_key}.weight"
+            converted_key = f"blocks.{i}.attn2.{c}.lora_B.weight"
+            if original_key in original_state_dict:
+                converted_state_dict[converted_key] = original_state_dict.pop(original_key)
+
+            original_key = f"blocks.{i}.cross_attn.{o}.diff_b"
+            converted_key = f"blocks.{i}.attn2.{c}.lora_B.bias"
+            if original_key in original_state_dict:
+                converted_state_dict[converted_key] = original_state_dict.pop(original_key)
 
         if is_i2v_lora:
             for o, c in zip(["k_img", "v_img"], ["add_k_proj", "add_v_proj"]):
-                converted_state_dict[f"blocks.{i}.attn2.{c}.lora_A.weight"] = original_state_dict.pop(
-                    f"blocks.{i}.cross_attn.{o}.{lora_down_key}.weight"
-                )
-                converted_state_dict[f"blocks.{i}.attn2.{c}.lora_B.weight"] = original_state_dict.pop(
-                    f"blocks.{i}.cross_attn.{o}.{lora_up_key}.weight"
-                )
-                if f"blocks.{i}.cross_attn.{o}.diff_b" in original_state_dict:
-                    converted_state_dict[f"blocks.{i}.attn2.{c}.lora_B.bias"] = original_state_dict.pop(
-                        f"blocks.{i}.cross_attn.{o}.diff_b"
-                    )
+                original_key = f"blocks.{i}.cross_attn.{o}.{lora_down_key}.weight"
+                converted_key = f"blocks.{i}.attn2.{c}.lora_A.weight"
+                if original_key in original_state_dict:
+                    converted_state_dict[converted_key] = original_state_dict.pop(original_key)
+
+                original_key = f"blocks.{i}.cross_attn.{o}.{lora_up_key}.weight"
+                converted_key = f"blocks.{i}.attn2.{c}.lora_B.weight"
+                if original_key in original_state_dict:
+                    converted_state_dict[converted_key] = original_state_dict.pop(original_key)
+
+                original_key = f"blocks.{i}.cross_attn.{o}.diff_b"
+                converted_key = f"blocks.{i}.attn2.{c}.lora_B.bias"
+                if original_key in original_state_dict:
+                    converted_state_dict[converted_key] = original_state_dict.pop(original_key)
 
         # FFN
         for o, c in zip(["ffn.0", "ffn.2"], ["net.0.proj", "net.2"]):
@@ -1674,10 +1689,10 @@ def _convert_non_diffusers_wan_lora_to_diffusers(state_dict):
             if original_key in original_state_dict:
                 converted_state_dict[converted_key] = original_state_dict.pop(original_key)
 
-            if f"blocks.{i}.{o}.diff_b" in original_state_dict:
-                converted_state_dict[f"blocks.{i}.ffn.{c}.lora_B.bias"] = original_state_dict.pop(
-                    f"blocks.{i}.{o}.diff_b"
-                )
+            original_key = f"blocks.{i}.{o}.diff_b"
+            converted_key = f"blocks.{i}.ffn.{c}.lora_B.bias"
+            if original_key in original_state_dict:
+                converted_state_dict[converted_key] = original_state_dict.pop(original_key)
 
     # Remaining.
     if original_state_dict:

src/diffusers/loaders/lora_pipeline.py

@@ -2031,18 +2031,36 @@ class FluxLoraLoaderMixin(LoraBaseMixin):
         if is_kohya:
             state_dict = _convert_kohya_flux_lora_to_diffusers(state_dict)
             # Kohya already takes care of scaling the LoRA parameters with alpha.
-            return (state_dict, None) if return_alphas else state_dict
+            return cls._prepare_outputs(
+                state_dict,
+                metadata=metadata,
+                alphas=None,
+                return_alphas=return_alphas,
+                return_metadata=return_lora_metadata,
+            )
 
         is_xlabs = any("processor" in k for k in state_dict)
         if is_xlabs:
             state_dict = _convert_xlabs_flux_lora_to_diffusers(state_dict)
             # xlabs doesn't use `alpha`.
-            return (state_dict, None) if return_alphas else state_dict
+            return cls._prepare_outputs(
+                state_dict,
+                metadata=metadata,
+                alphas=None,
+                return_alphas=return_alphas,
+                return_metadata=return_lora_metadata,
+            )
 
         is_bfl_control = any("query_norm.scale" in k for k in state_dict)
         if is_bfl_control:
             state_dict = _convert_bfl_flux_control_lora_to_diffusers(state_dict)
-            return (state_dict, None) if return_alphas else state_dict
+            return cls._prepare_outputs(
+                state_dict,
+                metadata=metadata,
+                alphas=None,
+                return_alphas=return_alphas,
+                return_metadata=return_lora_metadata,
+            )
 
         # For state dicts like
         # https://huggingface.co/TheLastBen/Jon_Snow_Flux_LoRA
@@ -2061,12 +2079,13 @@ class FluxLoraLoaderMixin(LoraBaseMixin):
                     )
 
         if return_alphas or return_lora_metadata:
-            outputs = [state_dict]
-            if return_alphas:
-                outputs.append(network_alphas)
-            if return_lora_metadata:
-                outputs.append(metadata)
-            return tuple(outputs)
+            return cls._prepare_outputs(
+                state_dict,
+                metadata=metadata,
+                alphas=network_alphas,
+                return_alphas=return_alphas,
+                return_metadata=return_lora_metadata,
+            )
         else:
             return state_dict
 
@@ -2785,6 +2804,15 @@ class FluxLoraLoaderMixin(LoraBaseMixin):
 
         raise ValueError("Either `base_module` or `base_weight_param_name` must be provided.")
 
+    @staticmethod
+    def _prepare_outputs(state_dict, metadata, alphas=None, return_alphas=False, return_metadata=False):
+        outputs = [state_dict]
+        if return_alphas:
+            outputs.append(alphas)
+        if return_metadata:
+            outputs.append(metadata)
+        return tuple(outputs) if (return_alphas or return_metadata) else state_dict
+
 
 # The reason why we subclass from `StableDiffusionLoraLoaderMixin` here is because Amused initially
 # relied on `StableDiffusionLoraLoaderMixin` for its LoRA support.

src/diffusers/loaders/peft.py

@@ -187,7 +187,9 @@ class PeftAdapterMixin:
                 Note that hotswapping adapters of the text encoder is not yet supported. There are some further
                 limitations to this technique, which are documented here:
                 https://huggingface.co/docs/peft/main/en/package_reference/hotswap
-            metadata: TODO
+            metadata:
+                LoRA adapter metadata. When supplied, the metadata inferred through the state dict isn't used to
+                initialize `LoraConfig`.
         """
         from peft import LoraConfig, inject_adapter_in_model, set_peft_model_state_dict
         from peft.tuners.tuners_utils import BaseTunerLayer

src/diffusers/models/autoencoders/autoencoder_kl_wan.py

@@ -749,6 +749,16 @@ class AutoencoderKLWan(ModelMixin, ConfigMixin, FromOriginalModelMixin):
         self.tile_sample_stride_height = 192
         self.tile_sample_stride_width = 192
 
+        # Precompute and cache conv counts for encoder and decoder for clear_cache speedup
+        self._cached_conv_counts = {
+            "decoder": sum(isinstance(m, WanCausalConv3d) for m in self.decoder.modules())
+            if self.decoder is not None
+            else 0,
+            "encoder": sum(isinstance(m, WanCausalConv3d) for m in self.encoder.modules())
+            if self.encoder is not None
+            else 0,
+        }
+
     def enable_tiling(
         self,
         tile_sample_min_height: Optional[int] = None,
@@ -801,18 +811,12 @@ class AutoencoderKLWan(ModelMixin, ConfigMixin, FromOriginalModelMixin):
         self.use_slicing = False
 
     def clear_cache(self):
-        def _count_conv3d(model):
-            count = 0
-            for m in model.modules():
-                if isinstance(m, WanCausalConv3d):
-                    count += 1
-            return count
-
-        self._conv_num = _count_conv3d(self.decoder)
+        # Use cached conv counts for decoder and encoder to avoid re-iterating modules each call
+        self._conv_num = self._cached_conv_counts["decoder"]
         self._conv_idx = [0]
         self._feat_map = [None] * self._conv_num
         # cache encode
-        self._enc_conv_num = _count_conv3d(self.encoder)
+        self._enc_conv_num = self._cached_conv_counts["encoder"]
         self._enc_conv_idx = [0]
         self._enc_feat_map = [None] * self._enc_conv_num
 

src/diffusers/models/transformers/transformer_cosmos.py

@@ -100,11 +100,15 @@ class CosmosAdaLayerNorm(nn.Module):
         embedded_timestep = self.linear_2(embedded_timestep)
 
         if temb is not None:
-            embedded_timestep = embedded_timestep + temb[:, : 2 * self.embedding_dim]
+            embedded_timestep = embedded_timestep + temb[..., : 2 * self.embedding_dim]
 
-        shift, scale = embedded_timestep.chunk(2, dim=1)
+        shift, scale = embedded_timestep.chunk(2, dim=-1)
         hidden_states = self.norm(hidden_states)
-        hidden_states = hidden_states * (1 + scale.unsqueeze(1)) + shift.unsqueeze(1)
+
+        if embedded_timestep.ndim == 2:
+            shift, scale = (x.unsqueeze(1) for x in (shift, scale))
+
+        hidden_states = hidden_states * (1 + scale) + shift
         return hidden_states
 
 
@@ -135,9 +139,13 @@ class CosmosAdaLayerNormZero(nn.Module):
         if temb is not None:
             embedded_timestep = embedded_timestep + temb
 
-        shift, scale, gate = embedded_timestep.chunk(3, dim=1)
+        shift, scale, gate = embedded_timestep.chunk(3, dim=-1)
         hidden_states = self.norm(hidden_states)
-        hidden_states = hidden_states * (1 + scale.unsqueeze(1)) + shift.unsqueeze(1)
+
+        if embedded_timestep.ndim == 2:
+            shift, scale, gate = (x.unsqueeze(1) for x in (shift, scale, gate))
+
+        hidden_states = hidden_states * (1 + scale) + shift
         return hidden_states, gate
 
 
@@ -255,19 +263,19 @@ class CosmosTransformerBlock(nn.Module):
         # 1. Self Attention
         norm_hidden_states, gate = self.norm1(hidden_states, embedded_timestep, temb)
         attn_output = self.attn1(norm_hidden_states, image_rotary_emb=image_rotary_emb)
-        hidden_states = hidden_states + gate.unsqueeze(1) * attn_output
+        hidden_states = hidden_states + gate * attn_output
 
         # 2. Cross Attention
         norm_hidden_states, gate = self.norm2(hidden_states, embedded_timestep, temb)
         attn_output = self.attn2(
             norm_hidden_states, encoder_hidden_states=encoder_hidden_states, attention_mask=attention_mask
         )
-        hidden_states = hidden_states + gate.unsqueeze(1) * attn_output
+        hidden_states = hidden_states + gate * attn_output
 
         # 3. Feed Forward
         norm_hidden_states, gate = self.norm3(hidden_states, embedded_timestep, temb)
         ff_output = self.ff(norm_hidden_states)
-        hidden_states = hidden_states + gate.unsqueeze(1) * ff_output
+        hidden_states = hidden_states + gate * ff_output
 
         return hidden_states
 
@@ -513,7 +521,23 @@ class CosmosTransformer3DModel(ModelMixin, ConfigMixin):
         hidden_states = hidden_states.flatten(1, 3)  # [B, T, H, W, C] -> [B, THW, C]
 
         # 4. Timestep embeddings
-        temb, embedded_timestep = self.time_embed(hidden_states, timestep)
+        if timestep.ndim == 1:
+            temb, embedded_timestep = self.time_embed(hidden_states, timestep)
+        elif timestep.ndim == 5:
+            assert timestep.shape == (batch_size, 1, num_frames, 1, 1), (
+                f"Expected timestep to have shape [B, 1, T, 1, 1], but got {timestep.shape}"
+            )
+            timestep = timestep.flatten()
+            temb, embedded_timestep = self.time_embed(hidden_states, timestep)
+            # We can do this because num_frames == post_patch_num_frames, as p_t is 1
+            temb, embedded_timestep = (
+                x.view(batch_size, post_patch_num_frames, 1, 1, -1)
+                .expand(-1, -1, post_patch_height, post_patch_width, -1)
+                .flatten(1, 3)
+                for x in (temb, embedded_timestep)
+            )  # [BT, C] -> [B, T, 1, 1, C] -> [B, T, H, W, C] -> [B, THW, C]
+        else:
+            assert False
 
         # 5. Transformer blocks
         for block in self.transformer_blocks:
@@ -544,8 +568,8 @@ class CosmosTransformer3DModel(ModelMixin, ConfigMixin):
         hidden_states = self.proj_out(hidden_states)
         hidden_states = hidden_states.unflatten(2, (p_h, p_w, p_t, -1))
         hidden_states = hidden_states.unflatten(1, (post_patch_num_frames, post_patch_height, post_patch_width))
-        # Please just kill me at this point. What even is this permutation order and why is it different from the patching order?
-        # Another few hours of sanity lost to the void.
+        # NOTE: The permutation order here is not the inverse operation of what happens when patching as usually expected.
+        # It might be a source of confusion to the reader, but this is correct
         hidden_states = hidden_states.permute(0, 7, 1, 6, 2, 4, 3, 5)
         hidden_states = hidden_states.flatten(6, 7).flatten(4, 5).flatten(2, 3)
 

src/diffusers/pipelines/__init__.py

@@ -158,7 +158,12 @@ else:
     _import_structure["cogview3"] = ["CogView3PlusPipeline"]
     _import_structure["cogview4"] = ["CogView4Pipeline", "CogView4ControlPipeline"]
     _import_structure["consisid"] = ["ConsisIDPipeline"]
-    _import_structure["cosmos"] = ["CosmosTextToWorldPipeline", "CosmosVideoToWorldPipeline"]
+    _import_structure["cosmos"] = [
+        "Cosmos2TextToImagePipeline",
+        "CosmosTextToWorldPipeline",
+        "CosmosVideoToWorldPipeline",
+        "Cosmos2VideoToWorldPipeline",
+    ]
     _import_structure["controlnet"].extend(
         [
             "BlipDiffusionControlNetPipeline",
@@ -561,7 +566,12 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             StableDiffusionControlNetXSPipeline,
             StableDiffusionXLControlNetXSPipeline,
         )
-        from .cosmos import CosmosTextToWorldPipeline, CosmosVideoToWorldPipeline
+        from .cosmos import (
+            Cosmos2TextToImagePipeline,
+            Cosmos2VideoToWorldPipeline,
+            CosmosTextToWorldPipeline,
+            CosmosVideoToWorldPipeline,
+        )
         from .deepfloyd_if import (
             IFImg2ImgPipeline,
             IFImg2ImgSuperResolutionPipeline,

src/diffusers/pipelines/cosmos/__init__.py

@@ -22,6 +22,8 @@ except OptionalDependencyNotAvailable:
 
     _dummy_objects.update(get_objects_from_module(dummy_torch_and_transformers_objects))
 else:
+    _import_structure["pipeline_cosmos2_text2image"] = ["Cosmos2TextToImagePipeline"]
+    _import_structure["pipeline_cosmos2_video2world"] = ["Cosmos2VideoToWorldPipeline"]
     _import_structure["pipeline_cosmos_text2world"] = ["CosmosTextToWorldPipeline"]
     _import_structure["pipeline_cosmos_video2world"] = ["CosmosVideoToWorldPipeline"]
 
@@ -33,6 +35,8 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
     except OptionalDependencyNotAvailable:
         from ...utils.dummy_torch_and_transformers_objects import *
     else:
+        from .pipeline_cosmos2_text2image import Cosmos2TextToImagePipeline
+        from .pipeline_cosmos2_video2world import Cosmos2VideoToWorldPipeline
         from .pipeline_cosmos_text2world import CosmosTextToWorldPipeline
         from .pipeline_cosmos_video2world import CosmosVideoToWorldPipeline
 

src/diffusers/pipelines/cosmos/pipeline_cosmos2_text2image.py

@@ -0,0 +1,673 @@
+# Copyright 2025 The NVIDIA Team and The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import inspect
+from typing import Callable, Dict, List, Optional, Union
+
+import numpy as np
+import torch
+from transformers import T5EncoderModel, T5TokenizerFast
+
+from ...callbacks import MultiPipelineCallbacks, PipelineCallback
+from ...models import AutoencoderKLWan, CosmosTransformer3DModel
+from ...schedulers import FlowMatchEulerDiscreteScheduler
+from ...utils import is_cosmos_guardrail_available, is_torch_xla_available, logging, replace_example_docstring
+from ...utils.torch_utils import randn_tensor
+from ...video_processor import VideoProcessor
+from ..pipeline_utils import DiffusionPipeline
+from .pipeline_output import CosmosImagePipelineOutput
+
+
+if is_cosmos_guardrail_available():
+    from cosmos_guardrail import CosmosSafetyChecker
+else:
+
+    class CosmosSafetyChecker:
+        def __init__(self, *args, **kwargs):
+            raise ImportError(
+                "`cosmos_guardrail` is not installed. Please install it to use the safety checker for Cosmos: `pip install cosmos_guardrail`."
+            )
+
+
+if is_torch_xla_available():
+    import torch_xla.core.xla_model as xm
+
+    XLA_AVAILABLE = True
+else:
+    XLA_AVAILABLE = False
+
+logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
+
+
+EXAMPLE_DOC_STRING = """
+    Examples:
+        ```python
+        >>> import torch
+        >>> from diffusers import Cosmos2TextToImagePipeline
+
+        >>> # Available checkpoints: nvidia/Cosmos-Predict2-2B-Text2Image, nvidia/Cosmos-Predict2-14B-Text2Image
+        >>> model_id = "nvidia/Cosmos-Predict2-2B-Text2Image"
+        >>> pipe = Cosmos2TextToImagePipeline.from_pretrained(model_id, 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."
+        >>> 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")
+        ```
+"""
+
+
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+def retrieve_timesteps(
+    scheduler,
+    num_inference_steps: Optional[int] = None,
+    device: Optional[Union[str, torch.device]] = None,
+    timesteps: Optional[List[int]] = None,
+    sigmas: Optional[List[float]] = None,
+    **kwargs,
+):
+    r"""
+    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
+    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
+
+    Args:
+        scheduler (`SchedulerMixin`):
+            The scheduler to get timesteps from.
+        num_inference_steps (`int`):
+            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
+            must be `None`.
+        device (`str` or `torch.device`, *optional*):
+            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
+        timesteps (`List[int]`, *optional*):
+            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
+            `num_inference_steps` and `sigmas` must be `None`.
+        sigmas (`List[float]`, *optional*):
+            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
+            `num_inference_steps` and `timesteps` must be `None`.
+
+    Returns:
+        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
+        second element is the number of inference steps.
+    """
+    if timesteps is not None and sigmas is not None:
+        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
+    if timesteps is not None:
+        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accepts_timesteps:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" timestep schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    elif sigmas is not None:
+        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accept_sigmas:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" sigmas schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    else:
+        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+    return timesteps, num_inference_steps
+
+
+class Cosmos2TextToImagePipeline(DiffusionPipeline):
+    r"""
+    Pipeline for text-to-image generation using [Cosmos Predict2](https://github.com/nvidia-cosmos/cosmos-predict2).
+
+    This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods
+    implemented for all pipelines (downloading, saving, running on a particular device, etc.).
+
+    Args:
+        text_encoder ([`T5EncoderModel`]):
+            Frozen text-encoder. Cosmos uses
+            [T5](https://huggingface.co/docs/transformers/model_doc/t5#transformers.T5EncoderModel); specifically the
+            [t5-11b](https://huggingface.co/google-t5/t5-11b) variant.
+        tokenizer (`T5TokenizerFast`):
+            Tokenizer of class
+            [T5Tokenizer](https://huggingface.co/docs/transformers/model_doc/t5#transformers.T5Tokenizer).
+        transformer ([`CosmosTransformer3DModel`]):
+            Conditional Transformer to denoise the encoded image latents.
+        scheduler ([`FlowMatchEulerDiscreteScheduler`]):
+            A scheduler to be used in combination with `transformer` to denoise the encoded image latents.
+        vae ([`AutoencoderKLWan`]):
+            Variational Auto-Encoder (VAE) Model to encode and decode videos to and from latent representations.
+    """
+
+    model_cpu_offload_seq = "text_encoder->transformer->vae"
+    _callback_tensor_inputs = ["latents", "prompt_embeds", "negative_prompt_embeds"]
+    # We mark safety_checker as optional here to get around some test failures, but it is not really optional
+    _optional_components = ["safety_checker"]
+
+    def __init__(
+        self,
+        text_encoder: T5EncoderModel,
+        tokenizer: T5TokenizerFast,
+        transformer: CosmosTransformer3DModel,
+        vae: AutoencoderKLWan,
+        scheduler: FlowMatchEulerDiscreteScheduler,
+        safety_checker: CosmosSafetyChecker = None,
+    ):
+        super().__init__()
+
+        if safety_checker is None:
+            safety_checker = CosmosSafetyChecker()
+
+        self.register_modules(
+            vae=vae,
+            text_encoder=text_encoder,
+            tokenizer=tokenizer,
+            transformer=transformer,
+            scheduler=scheduler,
+            safety_checker=safety_checker,
+        )
+
+        self.vae_scale_factor_temporal = 2 ** sum(self.vae.temperal_downsample) if getattr(self, "vae", None) else 4
+        self.vae_scale_factor_spatial = 2 ** len(self.vae.temperal_downsample) if getattr(self, "vae", None) else 8
+        self.video_processor = VideoProcessor(vae_scale_factor=self.vae_scale_factor_spatial)
+
+        self.sigma_max = 80.0
+        self.sigma_min = 0.002
+        self.sigma_data = 1.0
+        self.final_sigmas_type = "sigma_min"
+        if self.scheduler is not None:
+            self.scheduler.register_to_config(
+                sigma_max=self.sigma_max,
+                sigma_min=self.sigma_min,
+                sigma_data=self.sigma_data,
+                final_sigmas_type=self.final_sigmas_type,
+            )
+
+    # Copied from diffusers.pipelines.cosmos.pipeline_cosmos_text2world.CosmosTextToWorldPipeline._get_t5_prompt_embeds
+    def _get_t5_prompt_embeds(
+        self,
+        prompt: Union[str, List[str]] = None,
+        max_sequence_length: int = 512,
+        device: Optional[torch.device] = None,
+        dtype: Optional[torch.dtype] = None,
+    ):
+        device = device or self._execution_device
+        dtype = dtype or self.text_encoder.dtype
+        prompt = [prompt] if isinstance(prompt, str) else prompt
+
+        text_inputs = self.tokenizer(
+            prompt,
+            padding="max_length",
+            max_length=max_sequence_length,
+            truncation=True,
+            return_tensors="pt",
+            return_length=True,
+            return_offsets_mapping=False,
+        )
+        text_input_ids = text_inputs.input_ids
+        prompt_attention_mask = text_inputs.attention_mask.bool().to(device)
+
+        untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
+        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
+            removed_text = self.tokenizer.batch_decode(untruncated_ids[:, max_sequence_length - 1 : -1])
+            logger.warning(
+                "The following part of your input was truncated because `max_sequence_length` is set to "
+                f" {max_sequence_length} tokens: {removed_text}"
+            )
+
+        prompt_embeds = self.text_encoder(
+            text_input_ids.to(device), attention_mask=prompt_attention_mask
+        ).last_hidden_state
+        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
+
+        lengths = prompt_attention_mask.sum(dim=1).cpu()
+        for i, length in enumerate(lengths):
+            prompt_embeds[i, length:] = 0
+
+        return prompt_embeds
+
+    # Copied from diffusers.pipelines.cosmos.pipeline_cosmos_text2world.CosmosTextToWorldPipeline.encode_prompt with num_videos_per_prompt->num_images_per_prompt
+    def encode_prompt(
+        self,
+        prompt: Union[str, List[str]],
+        negative_prompt: Optional[Union[str, List[str]]] = None,
+        do_classifier_free_guidance: bool = True,
+        num_images_per_prompt: int = 1,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
+        max_sequence_length: int = 512,
+        device: Optional[torch.device] = None,
+        dtype: Optional[torch.dtype] = None,
+    ):
+        r"""
+        Encodes the prompt into text encoder hidden states.
+
+        Args:
+            prompt (`str` or `List[str]`, *optional*):
+                prompt to be encoded
+            negative_prompt (`str` or `List[str]`, *optional*):
+                The prompt or prompts not to guide the image generation. If not defined, one has to pass
+                `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
+                less than `1`).
+            do_classifier_free_guidance (`bool`, *optional*, defaults to `True`):
+                Whether to use classifier free guidance or not.
+            num_images_per_prompt (`int`, *optional*, defaults to 1):
+                Number of videos that should be generated per prompt. torch device to place the resulting embeddings on
+            prompt_embeds (`torch.Tensor`, *optional*):
+                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
+                provided, text embeddings will be generated from `prompt` input argument.
+            negative_prompt_embeds (`torch.Tensor`, *optional*):
+                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
+                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
+                argument.
+            device: (`torch.device`, *optional*):
+                torch device
+            dtype: (`torch.dtype`, *optional*):
+                torch dtype
+        """
+        device = device or self._execution_device
+
+        prompt = [prompt] if isinstance(prompt, str) else prompt
+        if prompt is not None:
+            batch_size = len(prompt)
+        else:
+            batch_size = prompt_embeds.shape[0]
+
+        if prompt_embeds is None:
+            prompt_embeds = self._get_t5_prompt_embeds(
+                prompt=prompt, max_sequence_length=max_sequence_length, device=device, dtype=dtype
+            )
+
+            # duplicate text embeddings for each generation per prompt, using mps friendly method
+            _, seq_len, _ = prompt_embeds.shape
+            prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
+            prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
+
+        if do_classifier_free_guidance and negative_prompt_embeds is None:
+            negative_prompt = negative_prompt or ""
+            negative_prompt = batch_size * [negative_prompt] if isinstance(negative_prompt, str) else negative_prompt
+
+            if prompt is not None and type(prompt) is not type(negative_prompt):
+                raise TypeError(
+                    f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
+                    f" {type(prompt)}."
+                )
+            elif batch_size != len(negative_prompt):
+                raise ValueError(
+                    f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
+                    f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
+                    " the batch size of `prompt`."
+                )
+
+            negative_prompt_embeds = self._get_t5_prompt_embeds(
+                prompt=negative_prompt, max_sequence_length=max_sequence_length, device=device, dtype=dtype
+            )
+
+            # duplicate text embeddings for each generation per prompt, using mps friendly method
+            _, seq_len, _ = negative_prompt_embeds.shape
+            negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
+            negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
+
+        return prompt_embeds, negative_prompt_embeds
+
+    def prepare_latents(
+        self,
+        batch_size: int,
+        num_channels_latents: 16,
+        height: int = 768,
+        width: int = 1360,
+        num_frames: int = 1,
+        dtype: Optional[torch.dtype] = None,
+        device: Optional[torch.device] = None,
+        generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
+        latents: Optional[torch.Tensor] = None,
+    ) -> torch.Tensor:
+        if latents is not None:
+            return latents.to(device=device, dtype=dtype) * self.scheduler.config.sigma_max
+
+        num_latent_frames = (num_frames - 1) // self.vae_scale_factor_temporal + 1
+        latent_height = height // self.vae_scale_factor_spatial
+        latent_width = width // self.vae_scale_factor_spatial
+        shape = (batch_size, num_channels_latents, num_latent_frames, latent_height, latent_width)
+
+        if isinstance(generator, list) and len(generator) != batch_size:
+            raise ValueError(
+                f"You have passed a list of generators of length {len(generator)}, but requested an effective batch"
+                f" size of {batch_size}. Make sure the batch size matches the length of the generators."
+            )
+
+        latents = randn_tensor(shape, generator=generator, device=device, dtype=dtype)
+        return latents * self.scheduler.config.sigma_max
+
+    # Copied from diffusers.pipelines.cosmos.pipeline_cosmos_text2world.CosmosTextToWorldPipeline.check_inputs
+    def check_inputs(
+        self,
+        prompt,
+        height,
+        width,
+        prompt_embeds=None,
+        callback_on_step_end_tensor_inputs=None,
+    ):
+        if height % 16 != 0 or width % 16 != 0:
+            raise ValueError(f"`height` and `width` have to be divisible by 16 but are {height} and {width}.")
+
+        if callback_on_step_end_tensor_inputs is not None and not all(
+            k in self._callback_tensor_inputs for k in callback_on_step_end_tensor_inputs
+        ):
+            raise ValueError(
+                f"`callback_on_step_end_tensor_inputs` has to be in {self._callback_tensor_inputs}, but found {[k for k in callback_on_step_end_tensor_inputs if k not in self._callback_tensor_inputs]}"
+            )
+
+        if prompt is not None and prompt_embeds is not None:
+            raise ValueError(
+                f"Cannot forward both `prompt`: {prompt} and `prompt_embeds`: {prompt_embeds}. Please make sure to"
+                " only forward one of the two."
+            )
+        elif prompt is None and prompt_embeds is None:
+            raise ValueError(
+                "Provide either `prompt` or `prompt_embeds`. Cannot leave both `prompt` and `prompt_embeds` undefined."
+            )
+        elif prompt is not None and (not isinstance(prompt, str) and not isinstance(prompt, list)):
+            raise ValueError(f"`prompt` has to be of type `str` or `list` but is {type(prompt)}")
+
+    @property
+    def guidance_scale(self):
+        return self._guidance_scale
+
+    @property
+    def do_classifier_free_guidance(self):
+        return self._guidance_scale > 1.0
+
+    @property
+    def num_timesteps(self):
+        return self._num_timesteps
+
+    @property
+    def current_timestep(self):
+        return self._current_timestep
+
+    @property
+    def interrupt(self):
+        return self._interrupt
+
+    @torch.no_grad()
+    @replace_example_docstring(EXAMPLE_DOC_STRING)
+    def __call__(
+        self,
+        prompt: Union[str, List[str]] = None,
+        negative_prompt: Optional[Union[str, List[str]]] = None,
+        height: int = 768,
+        width: int = 1360,
+        num_inference_steps: int = 35,
+        guidance_scale: float = 7.0,
+        num_images_per_prompt: Optional[int] = 1,
+        generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
+        latents: Optional[torch.Tensor] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
+        output_type: Optional[str] = "pil",
+        return_dict: bool = True,
+        callback_on_step_end: Optional[
+            Union[Callable[[int, int, Dict], None], PipelineCallback, MultiPipelineCallbacks]
+        ] = None,
+        callback_on_step_end_tensor_inputs: List[str] = ["latents"],
+        max_sequence_length: int = 512,
+    ):
+        r"""
+        The call function to the pipeline for generation.
+
+        Args:
+            prompt (`str` or `List[str]`, *optional*):
+                The prompt or prompts to guide the image generation. If not defined, one has to pass `prompt_embeds`.
+                instead.
+            height (`int`, defaults to `768`):
+                The height in pixels of the generated image.
+            width (`int`, defaults to `1360`):
+                The width in pixels of the generated image.
+            num_inference_steps (`int`, defaults to `35`):
+                The number of denoising steps. More denoising steps usually lead to a higher quality image at the
+                expense of slower inference.
+            guidance_scale (`float`, defaults to `7.0`):
+                Guidance scale as defined in [Classifier-Free Diffusion
+                Guidance](https://huggingface.co/papers/2207.12598). `guidance_scale` is defined as `w` of equation 2.
+                of [Imagen Paper](https://huggingface.co/papers/2205.11487). Guidance scale is enabled by setting
+                `guidance_scale > 1`.
+            num_images_per_prompt (`int`, *optional*, defaults to 1):
+                The number of images to generate per prompt.
+            generator (`torch.Generator` or `List[torch.Generator]`, *optional*):
+                A [`torch.Generator`](https://pytorch.org/docs/stable/generated/torch.Generator.html) to make
+                generation deterministic.
+            latents (`torch.Tensor`, *optional*):
+                Pre-generated noisy latents sampled from a Gaussian distribution, to be used as inputs for image
+                generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
+                tensor is generated by sampling using the supplied random `generator`.
+            prompt_embeds (`torch.Tensor`, *optional*):
+                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
+                provided, text embeddings will be generated from `prompt` input argument.
+            negative_prompt_embeds (`torch.FloatTensor`, *optional*):
+                Pre-generated negative text embeddings. For PixArt-Sigma this negative prompt should be "". If not
+                provided, negative_prompt_embeds will be generated from `negative_prompt` input argument.
+            output_type (`str`, *optional*, defaults to `"pil"`):
+                The output format of the generated image. Choose between `PIL.Image` or `np.array`.
+            return_dict (`bool`, *optional*, defaults to `True`):
+                Whether or not to return a [`CosmosImagePipelineOutput`] instead of a plain tuple.
+            callback_on_step_end (`Callable`, `PipelineCallback`, `MultiPipelineCallbacks`, *optional*):
+                A function or a subclass of `PipelineCallback` or `MultiPipelineCallbacks` that is called at the end of
+                each denoising step during the inference. with the following arguments: `callback_on_step_end(self:
+                DiffusionPipeline, step: int, timestep: int, callback_kwargs: Dict)`. `callback_kwargs` will include a
+                list of all tensors as specified by `callback_on_step_end_tensor_inputs`.
+            callback_on_step_end_tensor_inputs (`List`, *optional*):
+                The list of tensor inputs for the `callback_on_step_end` function. The tensors specified in the list
+                will be passed as `callback_kwargs` argument. You will only be able to include variables listed in the
+                `._callback_tensor_inputs` attribute of your pipeline class.
+
+        Examples:
+
+        Returns:
+            [`~CosmosImagePipelineOutput`] or `tuple`:
+                If `return_dict` is `True`, [`CosmosImagePipelineOutput`] is returned, otherwise a `tuple` is returned
+                where the first element is a list with the generated images and the second element is a list of `bool`s
+                indicating whether the corresponding generated image contains "not-safe-for-work" (nsfw) content.
+        """
+
+        if self.safety_checker is None:
+            raise ValueError(
+                f"You have disabled the safety checker for {self.__class__}. This is in violation of the "
+                "[NVIDIA Open Model License Agreement](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license). "
+                f"Please ensure that you are compliant with the license agreement."
+            )
+
+        if isinstance(callback_on_step_end, (PipelineCallback, MultiPipelineCallbacks)):
+            callback_on_step_end_tensor_inputs = callback_on_step_end.tensor_inputs
+
+        num_frames = 1
+
+        # 1. Check inputs. Raise error if not correct
+        self.check_inputs(prompt, height, width, prompt_embeds, callback_on_step_end_tensor_inputs)
+
+        self._guidance_scale = guidance_scale
+        self._current_timestep = None
+        self._interrupt = False
+
+        device = self._execution_device
+
+        if self.safety_checker is not None:
+            self.safety_checker.to(device)
+            if prompt is not None:
+                prompt_list = [prompt] if isinstance(prompt, str) else prompt
+                for p in prompt_list:
+                    if not self.safety_checker.check_text_safety(p):
+                        raise ValueError(
+                            f"Cosmos Guardrail detected unsafe text in the prompt: {p}. Please ensure that the "
+                            f"prompt abides by the NVIDIA Open Model License Agreement."
+                        )
+            self.safety_checker.to("cpu")
+
+        # 2. Define call parameters
+        if prompt is not None and isinstance(prompt, str):
+            batch_size = 1
+        elif prompt is not None and isinstance(prompt, list):
+            batch_size = len(prompt)
+        else:
+            batch_size = prompt_embeds.shape[0]
+
+        # 3. Encode input prompt
+        (
+            prompt_embeds,
+            negative_prompt_embeds,
+        ) = self.encode_prompt(
+            prompt=prompt,
+            negative_prompt=negative_prompt,
+            do_classifier_free_guidance=self.do_classifier_free_guidance,
+            num_images_per_prompt=num_images_per_prompt,
+            prompt_embeds=prompt_embeds,
+            negative_prompt_embeds=negative_prompt_embeds,
+            device=device,
+            max_sequence_length=max_sequence_length,
+        )
+
+        # 4. Prepare timesteps
+        sigmas_dtype = torch.float32 if torch.backends.mps.is_available() else torch.float64
+        sigmas = torch.linspace(0, 1, num_inference_steps, dtype=sigmas_dtype)
+        timesteps, num_inference_steps = retrieve_timesteps(self.scheduler, device=device, sigmas=sigmas)
+        if self.scheduler.config.get("final_sigmas_type", "zero") == "sigma_min":
+            # Replace the last sigma (which is zero) with the minimum sigma value
+            self.scheduler.sigmas[-1] = self.scheduler.sigmas[-2]
+
+        # 5. Prepare latent variables
+        transformer_dtype = self.transformer.dtype
+        num_channels_latents = self.transformer.config.in_channels
+        latents = self.prepare_latents(
+            batch_size * num_images_per_prompt,
+            num_channels_latents,
+            height,
+            width,
+            num_frames,
+            torch.float32,
+            device,
+            generator,
+            latents,
+        )
+
+        padding_mask = latents.new_zeros(1, 1, height, width, dtype=transformer_dtype)
+
+        # 6. Denoising loop
+        num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order
+        self._num_timesteps = len(timesteps)
+
+        with self.progress_bar(total=num_inference_steps) as progress_bar:
+            for i, t in enumerate(timesteps):
+                if self.interrupt:
+                    continue
+
+                self._current_timestep = t
+                current_sigma = self.scheduler.sigmas[i]
+
+                current_t = current_sigma / (current_sigma + 1)
+                c_in = 1 - current_t
+                c_skip = 1 - current_t
+                c_out = -current_t
+                timestep = current_t.expand(latents.shape[0]).to(transformer_dtype)  # [B, 1, T, 1, 1]
+
+                latent_model_input = latents * c_in
+                latent_model_input = latent_model_input.to(transformer_dtype)
+
+                noise_pred = self.transformer(
+                    hidden_states=latent_model_input,
+                    timestep=timestep,
+                    encoder_hidden_states=prompt_embeds,
+                    padding_mask=padding_mask,
+                    return_dict=False,
+                )[0]
+                noise_pred = (c_skip * latents + c_out * noise_pred.float()).to(transformer_dtype)
+
+                if self.do_classifier_free_guidance:
+                    noise_pred_uncond = self.transformer(
+                        hidden_states=latent_model_input,
+                        timestep=timestep,
+                        encoder_hidden_states=negative_prompt_embeds,
+                        padding_mask=padding_mask,
+                        return_dict=False,
+                    )[0]
+                    noise_pred_uncond = (c_skip * latents + c_out * noise_pred_uncond.float()).to(transformer_dtype)
+                    noise_pred = noise_pred + self.guidance_scale * (noise_pred - noise_pred_uncond)
+
+                noise_pred = (latents - noise_pred) / current_sigma
+                latents = self.scheduler.step(noise_pred, t, latents, return_dict=False)[0]
+
+                if callback_on_step_end is not None:
+                    callback_kwargs = {}
+                    for k in callback_on_step_end_tensor_inputs:
+                        callback_kwargs[k] = locals()[k]
+                    callback_outputs = callback_on_step_end(self, i, t, callback_kwargs)
+
+                    latents = callback_outputs.pop("latents", latents)
+                    prompt_embeds = callback_outputs.pop("prompt_embeds", prompt_embeds)
+                    negative_prompt_embeds = callback_outputs.pop("negative_prompt_embeds", negative_prompt_embeds)
+
+                # call the callback, if provided
+                if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):
+                    progress_bar.update()
+
+                if XLA_AVAILABLE:
+                    xm.mark_step()
+
+        self._current_timestep = None
+
+        if not output_type == "latent":
+            latents_mean = (
+                torch.tensor(self.vae.config.latents_mean)
+                .view(1, self.vae.config.z_dim, 1, 1, 1)
+                .to(latents.device, latents.dtype)
+            )
+            latents_std = 1.0 / torch.tensor(self.vae.config.latents_std).view(1, self.vae.config.z_dim, 1, 1, 1).to(
+                latents.device, latents.dtype
+            )
+            latents = latents / latents_std / self.scheduler.config.sigma_data + latents_mean
+            video = self.vae.decode(latents.to(self.vae.dtype), return_dict=False)[0]
+
+            if self.safety_checker is not None:
+                self.safety_checker.to(device)
+                video = self.video_processor.postprocess_video(video, output_type="np")
+                video = (video * 255).astype(np.uint8)
+                video_batch = []
+                for vid in video:
+                    vid = self.safety_checker.check_video_safety(vid)
+                    video_batch.append(vid)
+                video = np.stack(video_batch).astype(np.float32) / 255.0 * 2 - 1
+                video = torch.from_numpy(video).permute(0, 4, 1, 2, 3)
+                video = self.video_processor.postprocess_video(video, output_type=output_type)
+                self.safety_checker.to("cpu")
+            else:
+                video = self.video_processor.postprocess_video(video, output_type=output_type)
+            image = [batch[0] for batch in video]
+            if isinstance(video, torch.Tensor):
+                image = torch.stack(image)
+            elif isinstance(video, np.ndarray):
+                image = np.stack(image)
+        else:
+            image = latents[:, :, 0]
+
+        # Offload all models
+        self.maybe_free_model_hooks()
+
+        if not return_dict:
+            return (image,)
+
+        return CosmosImagePipelineOutput(images=image)

src/diffusers/pipelines/cosmos/pipeline_cosmos2_video2world.py

@@ -0,0 +1,792 @@
+# Copyright 2025 The NVIDIA Team and The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import inspect
+from typing import Callable, Dict, List, Optional, Union
+
+import numpy as np
+import torch
+from transformers import T5EncoderModel, T5TokenizerFast
+
+from ...callbacks import MultiPipelineCallbacks, PipelineCallback
+from ...image_processor import PipelineImageInput
+from ...models import AutoencoderKLWan, CosmosTransformer3DModel
+from ...schedulers import FlowMatchEulerDiscreteScheduler
+from ...utils import is_cosmos_guardrail_available, is_torch_xla_available, logging, replace_example_docstring
+from ...utils.torch_utils import randn_tensor
+from ...video_processor import VideoProcessor
+from ..pipeline_utils import DiffusionPipeline
+from .pipeline_output import CosmosPipelineOutput
+
+
+if is_cosmos_guardrail_available():
+    from cosmos_guardrail import CosmosSafetyChecker
+else:
+
+    class CosmosSafetyChecker:
+        def __init__(self, *args, **kwargs):
+            raise ImportError(
+                "`cosmos_guardrail` is not installed. Please install it to use the safety checker for Cosmos: `pip install cosmos_guardrail`."
+            )
+
+
+if is_torch_xla_available():
+    import torch_xla.core.xla_model as xm
+
+    XLA_AVAILABLE = True
+else:
+    XLA_AVAILABLE = False
+
+logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
+
+
+EXAMPLE_DOC_STRING = """
+    Examples:
+        ```python
+        >>> import torch
+        >>> from diffusers import Cosmos2VideoToWorldPipeline
+        >>> from diffusers.utils import export_to_video, load_image
+
+        >>> # Available checkpoints: nvidia/Cosmos-Predict2-2B-Video2World, nvidia/Cosmos-Predict2-14B-Video2World
+        >>> model_id = "nvidia/Cosmos-Predict2-2B-Video2World"
+        >>> pipe = Cosmos2VideoToWorldPipeline.from_pretrained(model_id, 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."
+        >>> 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."
+        >>> image = load_image(
+        ...     "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/yellow-scrubber.png"
+        ... )
+
+        >>> video = pipe(
+        ...     image=image, prompt=prompt, negative_prompt=negative_prompt, generator=torch.Generator().manual_seed(1)
+        ... ).frames[0]
+        >>> export_to_video(video, "output.mp4", fps=16)
+        ```
+"""
+
+
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+def retrieve_timesteps(
+    scheduler,
+    num_inference_steps: Optional[int] = None,
+    device: Optional[Union[str, torch.device]] = None,
+    timesteps: Optional[List[int]] = None,
+    sigmas: Optional[List[float]] = None,
+    **kwargs,
+):
+    r"""
+    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
+    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
+
+    Args:
+        scheduler (`SchedulerMixin`):
+            The scheduler to get timesteps from.
+        num_inference_steps (`int`):
+            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
+            must be `None`.
+        device (`str` or `torch.device`, *optional*):
+            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
+        timesteps (`List[int]`, *optional*):
+            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
+            `num_inference_steps` and `sigmas` must be `None`.
+        sigmas (`List[float]`, *optional*):
+            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
+            `num_inference_steps` and `timesteps` must be `None`.
+
+    Returns:
+        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
+        second element is the number of inference steps.
+    """
+    if timesteps is not None and sigmas is not None:
+        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
+    if timesteps is not None:
+        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accepts_timesteps:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" timestep schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    elif sigmas is not None:
+        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accept_sigmas:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" sigmas schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    else:
+        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+    return timesteps, num_inference_steps
+
+
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+def retrieve_latents(
+    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
+):
+    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
+        return encoder_output.latent_dist.sample(generator)
+    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
+        return encoder_output.latent_dist.mode()
+    elif hasattr(encoder_output, "latents"):
+        return encoder_output.latents
+    else:
+        raise AttributeError("Could not access latents of provided encoder_output")
+
+
+class Cosmos2VideoToWorldPipeline(DiffusionPipeline):
+    r"""
+    Pipeline for video-to-world generation using [Cosmos Predict2](https://github.com/nvidia-cosmos/cosmos-predict2).
+
+    This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods
+    implemented for all pipelines (downloading, saving, running on a particular device, etc.).
+
+    Args:
+        text_encoder ([`T5EncoderModel`]):
+            Frozen text-encoder. Cosmos uses
+            [T5](https://huggingface.co/docs/transformers/model_doc/t5#transformers.T5EncoderModel); specifically the
+            [t5-11b](https://huggingface.co/google-t5/t5-11b) variant.
+        tokenizer (`T5TokenizerFast`):
+            Tokenizer of class
+            [T5Tokenizer](https://huggingface.co/docs/transformers/model_doc/t5#transformers.T5Tokenizer).
+        transformer ([`CosmosTransformer3DModel`]):
+            Conditional Transformer to denoise the encoded image latents.
+        scheduler ([`FlowMatchEulerDiscreteScheduler`]):
+            A scheduler to be used in combination with `transformer` to denoise the encoded image latents.
+        vae ([`AutoencoderKLWan`]):
+            Variational Auto-Encoder (VAE) Model to encode and decode videos to and from latent representations.
+    """
+
+    model_cpu_offload_seq = "text_encoder->transformer->vae"
+    _callback_tensor_inputs = ["latents", "prompt_embeds", "negative_prompt_embeds"]
+    # We mark safety_checker as optional here to get around some test failures, but it is not really optional
+    _optional_components = ["safety_checker"]
+
+    def __init__(
+        self,
+        text_encoder: T5EncoderModel,
+        tokenizer: T5TokenizerFast,
+        transformer: CosmosTransformer3DModel,
+        vae: AutoencoderKLWan,
+        scheduler: FlowMatchEulerDiscreteScheduler,
+        safety_checker: CosmosSafetyChecker = None,
+    ):
+        super().__init__()
+
+        if safety_checker is None:
+            safety_checker = CosmosSafetyChecker()
+
+        self.register_modules(
+            vae=vae,
+            text_encoder=text_encoder,
+            tokenizer=tokenizer,
+            transformer=transformer,
+            scheduler=scheduler,
+            safety_checker=safety_checker,
+        )
+
+        self.vae_scale_factor_temporal = 2 ** sum(self.vae.temperal_downsample) if getattr(self, "vae", None) else 4
+        self.vae_scale_factor_spatial = 2 ** len(self.vae.temperal_downsample) if getattr(self, "vae", None) else 8
+        self.video_processor = VideoProcessor(vae_scale_factor=self.vae_scale_factor_spatial)
+
+        self.sigma_max = 80.0
+        self.sigma_min = 0.002
+        self.sigma_data = 1.0
+        self.final_sigmas_type = "sigma_min"
+        if self.scheduler is not None:
+            self.scheduler.register_to_config(
+                sigma_max=self.sigma_max,
+                sigma_min=self.sigma_min,
+                sigma_data=self.sigma_data,
+                final_sigmas_type=self.final_sigmas_type,
+            )
+
+    # Copied from diffusers.pipelines.cosmos.pipeline_cosmos_text2world.CosmosTextToWorldPipeline._get_t5_prompt_embeds
+    def _get_t5_prompt_embeds(
+        self,
+        prompt: Union[str, List[str]] = None,
+        max_sequence_length: int = 512,
+        device: Optional[torch.device] = None,
+        dtype: Optional[torch.dtype] = None,
+    ):
+        device = device or self._execution_device
+        dtype = dtype or self.text_encoder.dtype
+        prompt = [prompt] if isinstance(prompt, str) else prompt
+
+        text_inputs = self.tokenizer(
+            prompt,
+            padding="max_length",
+            max_length=max_sequence_length,
+            truncation=True,
+            return_tensors="pt",
+            return_length=True,
+            return_offsets_mapping=False,
+        )
+        text_input_ids = text_inputs.input_ids
+        prompt_attention_mask = text_inputs.attention_mask.bool().to(device)
+
+        untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
+        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
+            removed_text = self.tokenizer.batch_decode(untruncated_ids[:, max_sequence_length - 1 : -1])
+            logger.warning(
+                "The following part of your input was truncated because `max_sequence_length` is set to "
+                f" {max_sequence_length} tokens: {removed_text}"
+            )
+
+        prompt_embeds = self.text_encoder(
+            text_input_ids.to(device), attention_mask=prompt_attention_mask
+        ).last_hidden_state
+        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
+
+        lengths = prompt_attention_mask.sum(dim=1).cpu()
+        for i, length in enumerate(lengths):
+            prompt_embeds[i, length:] = 0
+
+        return prompt_embeds
+
+    # Copied from diffusers.pipelines.cosmos.pipeline_cosmos_text2world.CosmosTextToWorldPipeline.encode_prompt
+    def encode_prompt(
+        self,
+        prompt: Union[str, List[str]],
+        negative_prompt: Optional[Union[str, List[str]]] = None,
+        do_classifier_free_guidance: bool = True,
+        num_videos_per_prompt: int = 1,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
+        max_sequence_length: int = 512,
+        device: Optional[torch.device] = None,
+        dtype: Optional[torch.dtype] = None,
+    ):
+        r"""
+        Encodes the prompt into text encoder hidden states.
+
+        Args:
+            prompt (`str` or `List[str]`, *optional*):
+                prompt to be encoded
+            negative_prompt (`str` or `List[str]`, *optional*):
+                The prompt or prompts not to guide the image generation. If not defined, one has to pass
+                `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
+                less than `1`).
+            do_classifier_free_guidance (`bool`, *optional*, defaults to `True`):
+                Whether to use classifier free guidance or not.
+            num_videos_per_prompt (`int`, *optional*, defaults to 1):
+                Number of videos that should be generated per prompt. torch device to place the resulting embeddings on
+            prompt_embeds (`torch.Tensor`, *optional*):
+                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
+                provided, text embeddings will be generated from `prompt` input argument.
+            negative_prompt_embeds (`torch.Tensor`, *optional*):
+                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
+                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
+                argument.
+            device: (`torch.device`, *optional*):
+                torch device
+            dtype: (`torch.dtype`, *optional*):
+                torch dtype
+        """
+        device = device or self._execution_device
+
+        prompt = [prompt] if isinstance(prompt, str) else prompt
+        if prompt is not None:
+            batch_size = len(prompt)
+        else:
+            batch_size = prompt_embeds.shape[0]
+
+        if prompt_embeds is None:
+            prompt_embeds = self._get_t5_prompt_embeds(
+                prompt=prompt, max_sequence_length=max_sequence_length, device=device, dtype=dtype
+            )
+
+            # duplicate text embeddings for each generation per prompt, using mps friendly method
+            _, seq_len, _ = prompt_embeds.shape
+            prompt_embeds = prompt_embeds.repeat(1, num_videos_per_prompt, 1)
+            prompt_embeds = prompt_embeds.view(batch_size * num_videos_per_prompt, seq_len, -1)
+
+        if do_classifier_free_guidance and negative_prompt_embeds is None:
+            negative_prompt = negative_prompt or ""
+            negative_prompt = batch_size * [negative_prompt] if isinstance(negative_prompt, str) else negative_prompt
+
+            if prompt is not None and type(prompt) is not type(negative_prompt):
+                raise TypeError(
+                    f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
+                    f" {type(prompt)}."
+                )
+            elif batch_size != len(negative_prompt):
+                raise ValueError(
+                    f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
+                    f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
+                    " the batch size of `prompt`."
+                )
+
+            negative_prompt_embeds = self._get_t5_prompt_embeds(
+                prompt=negative_prompt, max_sequence_length=max_sequence_length, device=device, dtype=dtype
+            )
+
+            # duplicate text embeddings for each generation per prompt, using mps friendly method
+            _, seq_len, _ = negative_prompt_embeds.shape
+            negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_videos_per_prompt, 1)
+            negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_videos_per_prompt, seq_len, -1)
+
+        return prompt_embeds, negative_prompt_embeds
+
+    def prepare_latents(
+        self,
+        video: torch.Tensor,
+        batch_size: int,
+        num_channels_latents: 16,
+        height: int = 704,
+        width: int = 1280,
+        num_frames: int = 93,
+        do_classifier_free_guidance: bool = True,
+        dtype: Optional[torch.dtype] = None,
+        device: Optional[torch.device] = None,
+        generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
+        latents: Optional[torch.Tensor] = None,
+    ) -> torch.Tensor:
+        if isinstance(generator, list) and len(generator) != batch_size:
+            raise ValueError(
+                f"You have passed a list of generators of length {len(generator)}, but requested an effective batch"
+                f" size of {batch_size}. Make sure the batch size matches the length of the generators."
+            )
+
+        num_cond_frames = video.size(2)
+        if num_cond_frames >= num_frames:
+            # Take the last `num_frames` frames for conditioning
+            num_cond_latent_frames = (num_frames - 1) // self.vae_scale_factor_temporal + 1
+            video = video[:, :, -num_frames:]
+        else:
+            num_cond_latent_frames = (num_cond_frames - 1) // self.vae_scale_factor_temporal + 1
+            num_padding_frames = num_frames - num_cond_frames
+            last_frame = video[:, :, -1:]
+            padding = last_frame.repeat(1, 1, num_padding_frames, 1, 1)
+            video = torch.cat([video, padding], dim=2)
+
+        if isinstance(generator, list):
+            init_latents = [
+                retrieve_latents(self.vae.encode(video[i].unsqueeze(0)), generator=generator[i])
+                for i in range(batch_size)
+            ]
+        else:
+            init_latents = [retrieve_latents(self.vae.encode(vid.unsqueeze(0)), generator) for vid in video]
+
+        init_latents = torch.cat(init_latents, dim=0).to(dtype)
+
+        latents_mean = (
+            torch.tensor(self.vae.config.latents_mean).view(1, self.vae.config.z_dim, 1, 1, 1).to(device, dtype)
+        )
+        latents_std = (
+            torch.tensor(self.vae.config.latents_std).view(1, self.vae.config.z_dim, 1, 1, 1).to(device, dtype)
+        )
+        init_latents = (init_latents - latents_mean) / latents_std * self.scheduler.config.sigma_data
+
+        num_latent_frames = (num_frames - 1) // self.vae_scale_factor_temporal + 1
+        latent_height = height // self.vae_scale_factor_spatial
+        latent_width = width // self.vae_scale_factor_spatial
+        shape = (batch_size, num_channels_latents, num_latent_frames, latent_height, latent_width)
+
+        if latents is None:
+            latents = randn_tensor(shape, generator=generator, device=device, dtype=dtype)
+        else:
+            latents = latents.to(device=device, dtype=dtype)
+
+        latents = latents * self.scheduler.config.sigma_max
+
+        padding_shape = (batch_size, 1, num_latent_frames, latent_height, latent_width)
+        ones_padding = latents.new_ones(padding_shape)
+        zeros_padding = latents.new_zeros(padding_shape)
+
+        cond_indicator = latents.new_zeros(1, 1, latents.size(2), 1, 1)
+        cond_indicator[:, :, :num_cond_latent_frames] = 1.0
+        cond_mask = cond_indicator * ones_padding + (1 - cond_indicator) * zeros_padding
+
+        uncond_indicator = uncond_mask = None
+        if do_classifier_free_guidance:
+            uncond_indicator = latents.new_zeros(1, 1, latents.size(2), 1, 1)
+            uncond_indicator[:, :, :num_cond_latent_frames] = 1.0
+            uncond_mask = uncond_indicator * ones_padding + (1 - uncond_indicator) * zeros_padding
+
+        return latents, init_latents, cond_indicator, uncond_indicator, cond_mask, uncond_mask
+
+    # Copied from diffusers.pipelines.cosmos.pipeline_cosmos_text2world.CosmosTextToWorldPipeline.check_inputs
+    def check_inputs(
+        self,
+        prompt,
+        height,
+        width,
+        prompt_embeds=None,
+        callback_on_step_end_tensor_inputs=None,
+    ):
+        if height % 16 != 0 or width % 16 != 0:
+            raise ValueError(f"`height` and `width` have to be divisible by 16 but are {height} and {width}.")
+
+        if callback_on_step_end_tensor_inputs is not None and not all(
+            k in self._callback_tensor_inputs for k in callback_on_step_end_tensor_inputs
+        ):
+            raise ValueError(
+                f"`callback_on_step_end_tensor_inputs` has to be in {self._callback_tensor_inputs}, but found {[k for k in callback_on_step_end_tensor_inputs if k not in self._callback_tensor_inputs]}"
+            )
+
+        if prompt is not None and prompt_embeds is not None:
+            raise ValueError(
+                f"Cannot forward both `prompt`: {prompt} and `prompt_embeds`: {prompt_embeds}. Please make sure to"
+                " only forward one of the two."
+            )
+        elif prompt is None and prompt_embeds is None:
+            raise ValueError(
+                "Provide either `prompt` or `prompt_embeds`. Cannot leave both `prompt` and `prompt_embeds` undefined."
+            )
+        elif prompt is not None and (not isinstance(prompt, str) and not isinstance(prompt, list)):
+            raise ValueError(f"`prompt` has to be of type `str` or `list` but is {type(prompt)}")
+
+    @property
+    def guidance_scale(self):
+        return self._guidance_scale
+
+    @property
+    def do_classifier_free_guidance(self):
+        return self._guidance_scale > 1.0
+
+    @property
+    def num_timesteps(self):
+        return self._num_timesteps
+
+    @property
+    def current_timestep(self):
+        return self._current_timestep
+
+    @property
+    def interrupt(self):
+        return self._interrupt
+
+    @torch.no_grad()
+    @replace_example_docstring(EXAMPLE_DOC_STRING)
+    def __call__(
+        self,
+        image: PipelineImageInput = None,
+        video: List[PipelineImageInput] = None,
+        prompt: Union[str, List[str]] = None,
+        negative_prompt: Optional[Union[str, List[str]]] = None,
+        height: int = 704,
+        width: int = 1280,
+        num_frames: int = 93,
+        num_inference_steps: int = 35,
+        guidance_scale: float = 7.0,
+        fps: int = 16,
+        num_videos_per_prompt: Optional[int] = 1,
+        generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
+        latents: Optional[torch.Tensor] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
+        output_type: Optional[str] = "pil",
+        return_dict: bool = True,
+        callback_on_step_end: Optional[
+            Union[Callable[[int, int, Dict], None], PipelineCallback, MultiPipelineCallbacks]
+        ] = None,
+        callback_on_step_end_tensor_inputs: List[str] = ["latents"],
+        max_sequence_length: int = 512,
+        sigma_conditioning: float = 0.0001,
+    ):
+        r"""
+        The call function to the pipeline for generation.
+
+        Args:
+            image (`PIL.Image.Image`, `np.ndarray`, `torch.Tensor`, *optional*):
+                The image to be used as a conditioning input for the video generation.
+            video (`List[PIL.Image.Image]`, `np.ndarray`, `torch.Tensor`, *optional*):
+                The video to be used as a conditioning input for the video generation.
+            prompt (`str` or `List[str]`, *optional*):
+                The prompt or prompts to guide the image generation. If not defined, one has to pass `prompt_embeds`.
+                instead.
+            height (`int`, defaults to `704`):
+                The height in pixels of the generated image.
+            width (`int`, defaults to `1280`):
+                The width in pixels of the generated image.
+            num_frames (`int`, defaults to `93`):
+                The number of frames in the generated video.
+            num_inference_steps (`int`, defaults to `35`):
+                The number of denoising steps. More denoising steps usually lead to a higher quality image at the
+                expense of slower inference.
+            guidance_scale (`float`, defaults to `7.0`):
+                Guidance scale as defined in [Classifier-Free Diffusion
+                Guidance](https://huggingface.co/papers/2207.12598). `guidance_scale` is defined as `w` of equation 2.
+                of [Imagen Paper](https://huggingface.co/papers/2205.11487). Guidance scale is enabled by setting
+                `guidance_scale > 1`.
+            fps (`int`, defaults to `16`):
+                The frames per second of the generated video.
+            num_videos_per_prompt (`int`, *optional*, defaults to 1):
+                The number of images to generate per prompt.
+            generator (`torch.Generator` or `List[torch.Generator]`, *optional*):
+                A [`torch.Generator`](https://pytorch.org/docs/stable/generated/torch.Generator.html) to make
+                generation deterministic.
+            latents (`torch.Tensor`, *optional*):
+                Pre-generated noisy latents sampled from a Gaussian distribution, to be used as inputs for image
+                generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
+                tensor is generated by sampling using the supplied random `generator`.
+            prompt_embeds (`torch.Tensor`, *optional*):
+                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
+                provided, text embeddings will be generated from `prompt` input argument.
+            negative_prompt_embeds (`torch.FloatTensor`, *optional*):
+                Pre-generated negative text embeddings. For PixArt-Sigma this negative prompt should be "". If not
+                provided, negative_prompt_embeds will be generated from `negative_prompt` input argument.
+            output_type (`str`, *optional*, defaults to `"pil"`):
+                The output format of the generated image. Choose between `PIL.Image` or `np.array`.
+            return_dict (`bool`, *optional*, defaults to `True`):
+                Whether or not to return a [`CosmosPipelineOutput`] instead of a plain tuple.
+            callback_on_step_end (`Callable`, `PipelineCallback`, `MultiPipelineCallbacks`, *optional*):
+                A function or a subclass of `PipelineCallback` or `MultiPipelineCallbacks` that is called at the end of
+                each denoising step during the inference. with the following arguments: `callback_on_step_end(self:
+                DiffusionPipeline, step: int, timestep: int, callback_kwargs: Dict)`. `callback_kwargs` will include a
+                list of all tensors as specified by `callback_on_step_end_tensor_inputs`.
+            callback_on_step_end_tensor_inputs (`List`, *optional*):
+                The list of tensor inputs for the `callback_on_step_end` function. The tensors specified in the list
+                will be passed as `callback_kwargs` argument. You will only be able to include variables listed in the
+                `._callback_tensor_inputs` attribute of your pipeline class.
+            max_sequence_length (`int`, defaults to `512`):
+                The maximum number of tokens in the prompt. If the prompt exceeds this length, it will be truncated. If
+                the prompt is shorter than this length, it will be padded.
+            sigma_conditioning (`float`, defaults to `0.0001`):
+                The sigma value used for scaling conditioning latents. Ideally, it should not be changed or should be
+                set to a small value close to zero.
+
+        Examples:
+
+        Returns:
+            [`~CosmosPipelineOutput`] or `tuple`:
+                If `return_dict` is `True`, [`CosmosPipelineOutput`] is returned, otherwise a `tuple` is returned where
+                the first element is a list with the generated images and the second element is a list of `bool`s
+                indicating whether the corresponding generated image contains "not-safe-for-work" (nsfw) content.
+        """
+
+        if self.safety_checker is None:
+            raise ValueError(
+                f"You have disabled the safety checker for {self.__class__}. This is in violation of the "
+                "[NVIDIA Open Model License Agreement](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license). "
+                f"Please ensure that you are compliant with the license agreement."
+            )
+
+        if isinstance(callback_on_step_end, (PipelineCallback, MultiPipelineCallbacks)):
+            callback_on_step_end_tensor_inputs = callback_on_step_end.tensor_inputs
+
+        # 1. Check inputs. Raise error if not correct
+        self.check_inputs(prompt, height, width, prompt_embeds, callback_on_step_end_tensor_inputs)
+
+        self._guidance_scale = guidance_scale
+        self._current_timestep = None
+        self._interrupt = False
+
+        device = self._execution_device
+
+        if self.safety_checker is not None:
+            self.safety_checker.to(device)
+            if prompt is not None:
+                prompt_list = [prompt] if isinstance(prompt, str) else prompt
+                for p in prompt_list:
+                    if not self.safety_checker.check_text_safety(p):
+                        raise ValueError(
+                            f"Cosmos Guardrail detected unsafe text in the prompt: {p}. Please ensure that the "
+                            f"prompt abides by the NVIDIA Open Model License Agreement."
+                        )
+            self.safety_checker.to("cpu")
+
+        # 2. Define call parameters
+        if prompt is not None and isinstance(prompt, str):
+            batch_size = 1
+        elif prompt is not None and isinstance(prompt, list):
+            batch_size = len(prompt)
+        else:
+            batch_size = prompt_embeds.shape[0]
+
+        # 3. Encode input prompt
+        (
+            prompt_embeds,
+            negative_prompt_embeds,
+        ) = self.encode_prompt(
+            prompt=prompt,
+            negative_prompt=negative_prompt,
+            do_classifier_free_guidance=self.do_classifier_free_guidance,
+            num_videos_per_prompt=num_videos_per_prompt,
+            prompt_embeds=prompt_embeds,
+            negative_prompt_embeds=negative_prompt_embeds,
+            device=device,
+            max_sequence_length=max_sequence_length,
+        )
+
+        # 4. Prepare timesteps
+        sigmas_dtype = torch.float32 if torch.backends.mps.is_available() else torch.float64
+        sigmas = torch.linspace(0, 1, num_inference_steps, dtype=sigmas_dtype)
+        timesteps, num_inference_steps = retrieve_timesteps(self.scheduler, device=device, sigmas=sigmas)
+        if self.scheduler.config.final_sigmas_type == "sigma_min":
+            # Replace the last sigma (which is zero) with the minimum sigma value
+            self.scheduler.sigmas[-1] = self.scheduler.sigmas[-2]
+
+        # 5. Prepare latent variables
+        vae_dtype = self.vae.dtype
+        transformer_dtype = self.transformer.dtype
+
+        if image is not None:
+            video = self.video_processor.preprocess(image, height, width).unsqueeze(2)
+        else:
+            video = self.video_processor.preprocess_video(video, height, width)
+        video = video.to(device=device, dtype=vae_dtype)
+
+        num_channels_latents = self.transformer.config.in_channels - 1
+        latents, conditioning_latents, cond_indicator, uncond_indicator, cond_mask, uncond_mask = self.prepare_latents(
+            video,
+            batch_size * num_videos_per_prompt,
+            num_channels_latents,
+            height,
+            width,
+            num_frames,
+            self.do_classifier_free_guidance,
+            torch.float32,
+            device,
+            generator,
+            latents,
+        )
+        unconditioning_latents = None
+
+        cond_mask = cond_mask.to(transformer_dtype)
+        if self.do_classifier_free_guidance:
+            uncond_mask = uncond_mask.to(transformer_dtype)
+            unconditioning_latents = conditioning_latents
+
+        padding_mask = latents.new_zeros(1, 1, height, width, dtype=transformer_dtype)
+        sigma_conditioning = torch.tensor(sigma_conditioning, dtype=torch.float32, device=device)
+        t_conditioning = sigma_conditioning / (sigma_conditioning + 1)
+
+        # 6. Denoising loop
+        num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order
+        self._num_timesteps = len(timesteps)
+
+        with self.progress_bar(total=num_inference_steps) as progress_bar:
+            for i, t in enumerate(timesteps):
+                if self.interrupt:
+                    continue
+
+                self._current_timestep = t
+                current_sigma = self.scheduler.sigmas[i]
+
+                current_t = current_sigma / (current_sigma + 1)
+                c_in = 1 - current_t
+                c_skip = 1 - current_t
+                c_out = -current_t
+                timestep = current_t.view(1, 1, 1, 1, 1).expand(
+                    latents.size(0), -1, latents.size(2), -1, -1
+                )  # [B, 1, T, 1, 1]
+
+                cond_latent = latents * c_in
+                cond_latent = cond_indicator * conditioning_latents + (1 - cond_indicator) * cond_latent
+                cond_latent = cond_latent.to(transformer_dtype)
+                cond_timestep = cond_indicator * t_conditioning + (1 - cond_indicator) * timestep
+                cond_timestep = cond_timestep.to(transformer_dtype)
+
+                noise_pred = self.transformer(
+                    hidden_states=cond_latent,
+                    timestep=cond_timestep,
+                    encoder_hidden_states=prompt_embeds,
+                    fps=fps,
+                    condition_mask=cond_mask,
+                    padding_mask=padding_mask,
+                    return_dict=False,
+                )[0]
+                noise_pred = (c_skip * latents + c_out * noise_pred.float()).to(transformer_dtype)
+                noise_pred = cond_indicator * conditioning_latents + (1 - cond_indicator) * noise_pred
+
+                if self.do_classifier_free_guidance:
+                    uncond_latent = latents * c_in
+                    uncond_latent = uncond_indicator * unconditioning_latents + (1 - uncond_indicator) * uncond_latent
+                    uncond_latent = uncond_latent.to(transformer_dtype)
+                    uncond_timestep = uncond_indicator * t_conditioning + (1 - uncond_indicator) * timestep
+                    uncond_timestep = uncond_timestep.to(transformer_dtype)
+
+                    noise_pred_uncond = self.transformer(
+                        hidden_states=uncond_latent,
+                        timestep=uncond_timestep,
+                        encoder_hidden_states=negative_prompt_embeds,
+                        fps=fps,
+                        condition_mask=uncond_mask,
+                        padding_mask=padding_mask,
+                        return_dict=False,
+                    )[0]
+                    noise_pred_uncond = (c_skip * latents + c_out * noise_pred_uncond.float()).to(transformer_dtype)
+                    noise_pred_uncond = (
+                        uncond_indicator * unconditioning_latents + (1 - uncond_indicator) * noise_pred_uncond
+                    )
+                    noise_pred = noise_pred + self.guidance_scale * (noise_pred - noise_pred_uncond)
+
+                noise_pred = (latents - noise_pred) / current_sigma
+                latents = self.scheduler.step(noise_pred, t, latents, return_dict=False)[0]
+
+                if callback_on_step_end is not None:
+                    callback_kwargs = {}
+                    for k in callback_on_step_end_tensor_inputs:
+                        callback_kwargs[k] = locals()[k]
+                    callback_outputs = callback_on_step_end(self, i, t, callback_kwargs)
+
+                    latents = callback_outputs.pop("latents", latents)
+                    prompt_embeds = callback_outputs.pop("prompt_embeds", prompt_embeds)
+                    negative_prompt_embeds = callback_outputs.pop("negative_prompt_embeds", negative_prompt_embeds)
+
+                # call the callback, if provided
+                if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):
+                    progress_bar.update()
+
+                if XLA_AVAILABLE:
+                    xm.mark_step()
+
+        self._current_timestep = None
+
+        if not output_type == "latent":
+            latents_mean = (
+                torch.tensor(self.vae.config.latents_mean)
+                .view(1, self.vae.config.z_dim, 1, 1, 1)
+                .to(latents.device, latents.dtype)
+            )
+            latents_std = (
+                torch.tensor(self.vae.config.latents_std)
+                .view(1, self.vae.config.z_dim, 1, 1, 1)
+                .to(latents.device, latents.dtype)
+            )
+            latents = latents * latents_std / self.scheduler.config.sigma_data + latents_mean
+            video = self.vae.decode(latents.to(self.vae.dtype), return_dict=False)[0]
+
+            if self.safety_checker is not None:
+                self.safety_checker.to(device)
+                video = self.video_processor.postprocess_video(video, output_type="np")
+                video = (video * 255).astype(np.uint8)
+                video_batch = []
+                for vid in video:
+                    vid = self.safety_checker.check_video_safety(vid)
+                    video_batch.append(vid)
+                video = np.stack(video_batch).astype(np.float32) / 255.0 * 2 - 1
+                video = torch.from_numpy(video).permute(0, 4, 1, 2, 3)
+                video = self.video_processor.postprocess_video(video, output_type=output_type)
+                self.safety_checker.to("cpu")
+            else:
+                video = self.video_processor.postprocess_video(video, output_type=output_type)
+        else:
+            video = latents
+
+        # Offload all models
+        self.maybe_free_model_hooks()
+
+        if not return_dict:
+            return (video,)
+
+        return CosmosPipelineOutput(frames=video)

src/diffusers/pipelines/cosmos/pipeline_cosmos_text2world.py

@@ -131,7 +131,7 @@ def retrieve_timesteps(
 
 class CosmosTextToWorldPipeline(DiffusionPipeline):
     r"""
-    Pipeline for text-to-video generation using [Cosmos](https://github.com/NVIDIA/Cosmos).
+    Pipeline for text-to-world generation using [Cosmos Predict1](https://github.com/nvidia-cosmos/cosmos-predict1).
 
     This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods
     implemented for all pipelines (downloading, saving, running on a particular device, etc.).
@@ -426,12 +426,12 @@ class CosmosTextToWorldPipeline(DiffusionPipeline):
                 The height in pixels of the generated image.
             width (`int`, defaults to `1280`):
                 The width in pixels of the generated image.
-            num_frames (`int`, defaults to `129`):
+            num_frames (`int`, defaults to `121`):
                 The number of frames in the generated video.
-            num_inference_steps (`int`, defaults to `50`):
+            num_inference_steps (`int`, defaults to `36`):
                 The number of denoising steps. More denoising steps usually lead to a higher quality image at the
                 expense of slower inference.
-            guidance_scale (`float`, defaults to `6.0`):
+            guidance_scale (`float`, defaults to `7.0`):
                 Guidance scale as defined in [Classifier-Free Diffusion
                 Guidance](https://huggingface.co/papers/2207.12598). `guidance_scale` is defined as `w` of equation 2.
                 of [Imagen Paper](https://huggingface.co/papers/2205.11487). Guidance scale is enabled by setting
@@ -457,9 +457,6 @@ class CosmosTextToWorldPipeline(DiffusionPipeline):
                 The output format of the generated image. Choose between `PIL.Image` or `np.array`.
             return_dict (`bool`, *optional*, defaults to `True`):
                 Whether or not to return a [`CosmosPipelineOutput`] instead of a plain tuple.
-            clip_skip (`int`, *optional*):
-                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
-                the output of the pre-final layer will be used for computing the prompt embeddings.
             callback_on_step_end (`Callable`, `PipelineCallback`, `MultiPipelineCallbacks`, *optional*):
                 A function or a subclass of `PipelineCallback` or `MultiPipelineCallbacks` that is called at the end of
                 each denoising step during the inference. with the following arguments: `callback_on_step_end(self:

src/diffusers/pipelines/cosmos/pipeline_cosmos_video2world.py

@@ -174,7 +174,8 @@ def retrieve_latents(
 
 class CosmosVideoToWorldPipeline(DiffusionPipeline):
     r"""
-    Pipeline for image-to-video and video-to-video generation using [Cosmos](https://github.com/NVIDIA/Cosmos).
+    Pipeline for image-to-world and video-to-world generation using [Cosmos
+    Predict-1](https://github.com/nvidia-cosmos/cosmos-predict1).
 
     This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods
     implemented for all pipelines (downloading, saving, running on a particular device, etc.).
@@ -541,12 +542,12 @@ class CosmosVideoToWorldPipeline(DiffusionPipeline):
                 The height in pixels of the generated image.
             width (`int`, defaults to `1280`):
                 The width in pixels of the generated image.
-            num_frames (`int`, defaults to `129`):
+            num_frames (`int`, defaults to `121`):
                 The number of frames in the generated video.
-            num_inference_steps (`int`, defaults to `50`):
+            num_inference_steps (`int`, defaults to `36`):
                 The number of denoising steps. More denoising steps usually lead to a higher quality image at the
                 expense of slower inference.
-            guidance_scale (`float`, defaults to `6.0`):
+            guidance_scale (`float`, defaults to `7.0`):
                 Guidance scale as defined in [Classifier-Free Diffusion
                 Guidance](https://huggingface.co/papers/2207.12598). `guidance_scale` is defined as `w` of equation 2.
                 of [Imagen Paper](https://huggingface.co/papers/2205.11487). Guidance scale is enabled by setting
@@ -572,9 +573,6 @@ class CosmosVideoToWorldPipeline(DiffusionPipeline):
                 The output format of the generated image. Choose between `PIL.Image` or `np.array`.
             return_dict (`bool`, *optional*, defaults to `True`):
                 Whether or not to return a [`CosmosPipelineOutput`] instead of a plain tuple.
-            clip_skip (`int`, *optional*):
-                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
-                the output of the pre-final layer will be used for computing the prompt embeddings.
             callback_on_step_end (`Callable`, `PipelineCallback`, `MultiPipelineCallbacks`, *optional*):
                 A function or a subclass of `PipelineCallback` or `MultiPipelineCallbacks` that is called at the end of
                 each denoising step during the inference. with the following arguments: `callback_on_step_end(self:

src/diffusers/pipelines/cosmos/pipeline_output.py

@@ -1,14 +1,20 @@
 from dataclasses import dataclass
+from typing import List, Union
 
+import numpy as np
+import PIL.Image
 import torch
 
-from diffusers.utils import BaseOutput
+from diffusers.utils import BaseOutput, get_logger
+
+
+logger = get_logger(__name__)
 
 
 @dataclass
 class CosmosPipelineOutput(BaseOutput):
     r"""
-    Output class for Cosmos pipelines.
+    Output class for Cosmos any-to-world/video pipelines.
 
     Args:
         frames (`torch.Tensor`, `np.ndarray`, or List[List[PIL.Image.Image]]):
@@ -18,3 +24,17 @@ class CosmosPipelineOutput(BaseOutput):
     """
 
     frames: torch.Tensor
+
+
+@dataclass
+class CosmosImagePipelineOutput(BaseOutput):
+    """
+    Output class for Cosmos any-to-image pipelines.
+
+    Args:
+        images (`List[PIL.Image.Image]` or `np.ndarray`)
+            List of denoised PIL images of length `batch_size` or numpy array of shape `(batch_size, height, width,
+            num_channels)`. PIL images or numpy array present the denoised images of the diffusion pipeline.
+    """
+
+    images: Union[List[PIL.Image.Image], np.ndarray]

src/diffusers/training_utils.py

@@ -247,6 +247,14 @@ def _set_state_dict_into_text_encoder(
     set_peft_model_state_dict(text_encoder, text_encoder_state_dict, adapter_name="default")
 
 
+def _collate_lora_metadata(modules_to_save: Dict[str, torch.nn.Module]) -> Dict[str, Any]:
+    metadatas = {}
+    for module_name, module in modules_to_save.items():
+        if module is not None:
+            metadatas[f"{module_name}_lora_adapter_metadata"] = module.peft_config["default"].to_dict()
+    return metadatas
+
+
 def compute_density_for_timestep_sampling(
     weighting_scheme: str,
     batch_size: int,

src/diffusers/utils/doc.py

@@ -0,0 +1,300 @@
+
+# Copyright 2025 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+Doc utilities: Utilities related to documentation
+
+Adapted from:
+https://github.com/huggingface/transformers/blob/5a95ed5ca0826c867e35e52f698db4d8fc907bcb/src/transformers/utils/doc.py
+"""
+
+import functools
+import inspect
+import re
+import textwrap
+import types
+from collections import OrderedDict
+
+from ..pipelines.auto_pipeline import AUTO_TEXT2IMAGE_PIPELINES_MAPPING
+
+
+def get_docstring_indentation_level(func):
+    """Return the indentation level of the start of the docstring of a class or function (or method)."""
+    # We assume classes are always defined in the global scope
+    if inspect.isclass(func):
+        return 4
+    source = inspect.getsource(func)
+    first_line = source.splitlines()[0]
+    function_def_level = len(first_line) - len(first_line.lstrip())
+    return 4 + function_def_level
+
+
+def add_start_docstrings(*docstr):
+    def docstring_decorator(fn):
+        fn.__doc__ = "".join(docstr) + (fn.__doc__ if fn.__doc__ is not None else "")
+        return fn
+
+    return docstring_decorator
+
+
+def add_start_docstrings_to_model_forward(*docstr):
+    def docstring_decorator(fn):
+        class_name = f"[`{fn.__qualname__.split('.')[0]}`]"
+        intro = rf"""    The {class_name} forward method, overrides the `__call__` special method.
+
+    <Tip>
+
+    Although the recipe for forward pass needs to be defined within this function, one should call the [`Module`]
+    instance afterwards instead of this since the former takes care of running the pre and post processing steps while
+    the latter silently ignores them.
+
+    </Tip>
+"""
+
+        correct_indentation = get_docstring_indentation_level(fn)
+        current_doc = fn.__doc__ if fn.__doc__ is not None else ""
+        try:
+            first_non_empty = next(line for line in current_doc.splitlines() if line.strip() != "")
+            doc_indentation = len(first_non_empty) - len(first_non_empty.lstrip())
+        except StopIteration:
+            doc_indentation = correct_indentation
+
+        docs = docstr
+        # In this case, the correct indentation level (class method, 2 Python levels) was respected, and we should
+        # correctly reindent everything. Otherwise, the doc uses a single indentation level
+        if doc_indentation == 4 + correct_indentation:
+            docs = [textwrap.indent(textwrap.dedent(doc), " " * correct_indentation) for doc in docstr]
+            intro = textwrap.indent(textwrap.dedent(intro), " " * correct_indentation)
+
+        docstring = "".join(docs) + current_doc
+        fn.__doc__ = intro + docstring
+        return fn
+
+    return docstring_decorator
+
+
+def add_end_docstrings(*docstr):
+    def docstring_decorator(fn):
+        fn.__doc__ = (fn.__doc__ if fn.__doc__ is not None else "") + "".join(docstr)
+        return fn
+
+    return docstring_decorator
+
+
+PT_RETURN_INTRODUCTION = r"""
+    Returns:
+        [`{full_output_type}`] or `tuple(torch.FloatTensor)`: A [`{full_output_type}`] or a tuple of
+        `torch.FloatTensor` (if `return_dict=False` is passed) comprising various
+        elements depending on the model and inputs.
+
+"""
+
+TEXT_TO_IMAGE_PIPELINE_CLASSES = list({p[0] for p in AUTO_TEXT2IMAGE_PIPELINES_MAPPING})
+
+def _get_indent(t):
+    """Returns the indentation in the first line of t"""
+    search = re.search(r"^(\s*)\S", t)
+    return "" if search is None else search.groups()[0]
+
+
+def _convert_output_args_doc(output_args_doc):
+    """Convert output_args_doc to display properly."""
+    # Split output_arg_doc in blocks argument/description
+    indent = _get_indent(output_args_doc)
+    blocks = []
+    current_block = ""
+    for line in output_args_doc.split("\n"):
+        # If the indent is the same as the beginning, the line is the name of new arg.
+        if _get_indent(line) == indent:
+            if len(current_block) > 0:
+                blocks.append(current_block[:-1])
+            current_block = f"{line}\n"
+        else:
+            # Otherwise it's part of the description of the current arg.
+            # We need to remove 2 spaces to the indentation.
+            current_block += f"{line[2:]}\n"
+    blocks.append(current_block[:-1])
+
+    # Format each block for proper rendering
+    for i in range(len(blocks)):
+        blocks[i] = re.sub(r"^(\s+)(\S+)(\s+)", r"\1- **\2**\3", blocks[i])
+        blocks[i] = re.sub(r":\s*\n\s*(\S)", r" -- \1", blocks[i])
+
+    return "\n".join(blocks)
+
+
+def _prepare_output_docstrings(output_type, config_class, min_indent=None, add_intro=True):
+    """
+    Prepares the return part of the docstring using `output_type`.
+    """
+    output_docstring = output_type.__doc__
+    params_docstring = None
+    if output_docstring is not None:
+        # Remove the head of the docstring to keep the list of args only
+        lines = output_docstring.split("\n")
+        i = 0
+        while i < len(lines) and re.search(r"^\s*(Args|Parameters):\s*$", lines[i]) is None:
+            i += 1
+        if i < len(lines):
+            params_docstring = "\n".join(lines[(i + 1) :])
+            params_docstring = _convert_output_args_doc(params_docstring)
+        elif add_intro:
+            raise ValueError(
+                f"No `Args` or `Parameters` section is found in the docstring of `{output_type.__name__}`. Make sure it has "
+                "docstring and contain either `Args` or `Parameters`."
+            )
+
+    # Add the return introduction
+    if add_intro:
+        full_output_type = f"{output_type.__module__}.{output_type.__name__}"
+        intro = PT_RETURN_INTRODUCTION
+        intro = intro.format(full_output_type=full_output_type, config_class=config_class)
+    else:
+        full_output_type = str(output_type)
+        intro = f"\nReturns:\n    `{full_output_type}`"
+        if params_docstring is not None:
+            intro += ":\n"
+
+    result = intro
+    if params_docstring is not None:
+        result += params_docstring
+
+    # Apply minimum indent if necessary
+    if min_indent is not None:
+        lines = result.split("\n")
+        # Find the indent of the first nonempty line
+        i = 0
+        while len(lines[i]) == 0:
+            i += 1
+        indent = len(_get_indent(lines[i]))
+        # If too small, add indentation to all nonempty lines
+        if indent < min_indent:
+            to_add = " " * (min_indent - indent)
+            lines = [(f"{to_add}{line}" if len(line) > 0 else line) for line in lines]
+            result = "\n".join(lines)
+
+    return result
+
+
+FAKE_MODEL_DISCLAIMER = """
+    <Tip warning={true}>
+
+    This example uses a random model as the real ones are all very big. To get proper results, you should use
+    {real_checkpoint} instead of {fake_checkpoint}. If you get out-of-memory when loading that checkpoint, you can
+    refer to our optimization docs.
+
+    </Tip>
+"""
+
+
+PT_TEXT_TO_IMAGE_SAMPLE = r"""
+    Example:
+
+    ```python
+    >>> from diffusers import DiffusionPipeline
+    >>> import torch
+
+    >>> # If memory doesn't allow, enable optimizations like `enable_model_cpu_offload()`.
+    >>> pipe = DiffusionPipeline.from_pretrained("{checkpoint}", torch_dtype=torch.bfloat16).to("cuda")
+
+    >>> prompt = "a photo of a cute dog."
+    >>> image = pipe(prompt).images[0] # Configure other pipe call arguments as needed.
+    ```
+"""
+
+PT_SAMPLE_DOCSTRINGS = {
+    "Text2Image": PT_TEXT_TO_IMAGE_SAMPLE
+}
+PIPELINE_TASKS_TO_SAMPLE_DOCSTRINGS = OrderedDict(["text-to-image", PT_TEXT_TO_IMAGE_SAMPLE])
+
+def filter_outputs_from_example(docstring, **kwargs):
+    """
+    Removes the lines testing an output with the doctest syntax in a code sample when it's set to `None`.
+    """
+    for key, value in kwargs.items():
+        if value is not None:
+            continue
+
+        doc_key = "{" + key + "}"
+        docstring = re.sub(rf"\n([^\n]+)\n\s+{doc_key}\n", "\n", docstring)
+
+    return docstring
+
+
+def add_code_sample_docstrings(
+    *docstr,
+    checkpoint=None,
+    output_type=None,
+    config_class=None,
+    model_cls=None,
+):
+    def docstring_decorator(fn):
+        # model_class defaults to function's class if not specified otherwise
+        model_class = fn.__qualname__.split(".")[0] if model_cls is None else model_cls
+
+        sample_docstrings = PT_SAMPLE_DOCSTRINGS
+
+        # putting all kwargs for docstrings in a dict to be used
+        # with the `.format(**doc_kwargs)`. Note that string might
+        # be formatted with non-existing keys, which is fine.
+        doc_kwargs = {
+            "checkpoint": checkpoint,
+            "true": "{true}",  # For <Tip warning={true}> syntax that conflicts with formatting.
+        }
+
+        if model_class in TEXT_TO_IMAGE_PIPELINE_CLASSES:
+            code_sample = sample_docstrings["Text2Image"]
+        else:
+            raise ValueError(f"Docstring can't be built for model {model_class}")
+
+        code_sample = filter_outputs_from_example(code_sample)
+        func_doc = (fn.__doc__ or "") + "".join(docstr)
+        output_doc = "" if output_type is None else _prepare_output_docstrings(output_type, config_class)
+        built_doc = code_sample.format(**doc_kwargs)
+
+        fn.__doc__ = func_doc + output_doc + built_doc
+        return fn
+
+    return docstring_decorator
+
+
+def replace_return_docstrings(output_type=None, config_class=None):
+    def docstring_decorator(fn):
+        func_doc = fn.__doc__
+        lines = func_doc.split("\n")
+        i = 0
+        while i < len(lines) and re.search(r"^\s*Returns?:\s*$", lines[i]) is None:
+            i += 1
+        if i < len(lines):
+            indent = len(_get_indent(lines[i]))
+            lines[i] = _prepare_output_docstrings(output_type, config_class, min_indent=indent)
+            func_doc = "\n".join(lines)
+        else:
+            raise ValueError(
+                f"The function {fn} should have an empty 'Return:' or 'Returns:' in its docstring as placeholder, "
+                f"current docstring is:\n{func_doc}"
+            )
+        fn.__doc__ = func_doc
+        return fn
+
+    return docstring_decorator
+
+
+def copy_func(f):
+    """Returns a copy of a function f."""
+    # Based on http://stackoverflow.com/a/6528148/190597 (Glenn Maynard)
+    g = types.FunctionType(f.__code__, f.__globals__, name=f.__name__, argdefs=f.__defaults__, closure=f.__closure__)
+    g = functools.update_wrapper(g, f)
+    g.__kwdefaults__ = f.__kwdefaults__
+    return g

src/diffusers/utils/dummy_torch_and_transformers_objects.py

@@ -422,6 +422,36 @@ class ConsisIDPipeline(metaclass=DummyObject):
         requires_backends(cls, ["torch", "transformers"])
 
 
+class Cosmos2TextToImagePipeline(metaclass=DummyObject):
+    _backends = ["torch", "transformers"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch", "transformers"])
+
+    @classmethod
+    def from_config(cls, *args, **kwargs):
+        requires_backends(cls, ["torch", "transformers"])
+
+    @classmethod
+    def from_pretrained(cls, *args, **kwargs):
+        requires_backends(cls, ["torch", "transformers"])
+
+
+class Cosmos2VideoToWorldPipeline(metaclass=DummyObject):
+    _backends = ["torch", "transformers"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch", "transformers"])
+
+    @classmethod
+    def from_config(cls, *args, **kwargs):
+        requires_backends(cls, ["torch", "transformers"])
+
+    @classmethod
+    def from_pretrained(cls, *args, **kwargs):
+        requires_backends(cls, ["torch", "transformers"])
+
+
 class CosmosTextToWorldPipeline(metaclass=DummyObject):
     _backends = ["torch", "transformers"]
 

src/diffusers/utils/state_dict_utils.py

@@ -359,5 +359,8 @@ def _load_sft_state_dict_metadata(model_file: str):
         metadata = f.metadata() or {}
 
     metadata.pop("format", None)
-    raw = metadata.get(LORA_ADAPTER_METADATA_KEY)
-    return json.loads(raw) if raw else None
+    if metadata:
+        raw = metadata.get(LORA_ADAPTER_METADATA_KEY)
+        return json.loads(raw) if raw else None
+    else:
+        return None

tests/models/test_modeling_common.py

@@ -1736,6 +1736,45 @@ class ModelTesterMixin:
             f"AutoModel forward pass diff: {max_diff} exceeds threshold {expected_max_diff}",
         )
 
+    @parameterized.expand(
+        [
+            (-1, "You can't pass device_map as a negative int"),
+            ("foo", "When passing device_map as a string, the value needs to be a device name"),
+        ]
+    )
+    def test_wrong_device_map_raises_error(self, device_map, msg_substring):
+        init_dict, _ = self.prepare_init_args_and_inputs_for_common()
+        model = self.model_class(**init_dict)
+        with tempfile.TemporaryDirectory() as tmpdir:
+            model.save_pretrained(tmpdir)
+            with self.assertRaises(ValueError) as err_ctx:
+                _ = self.model_class.from_pretrained(tmpdir, device_map=device_map)
+
+        assert msg_substring in str(err_ctx.exception)
+
+    @parameterized.expand([0, "cuda", torch.device("cuda")])
+    @require_torch_gpu
+    def test_passing_non_dict_device_map_works(self, device_map):
+        init_dict, inputs_dict = self.prepare_init_args_and_inputs_for_common()
+        model = self.model_class(**init_dict).eval()
+        with tempfile.TemporaryDirectory() as tmpdir:
+            model.save_pretrained(tmpdir)
+            loaded_model = self.model_class.from_pretrained(tmpdir, device_map=device_map)
+            _ = loaded_model(**inputs_dict)
+
+    @parameterized.expand([("", "cuda"), ("", torch.device("cuda"))])
+    @require_torch_gpu
+    def test_passing_dict_device_map_works(self, name, device):
+        # There are other valid dict-based `device_map` values too. It's best to refer to
+        # the docs for those: https://huggingface.co/docs/accelerate/en/concept_guides/big_model_inference#the-devicemap.
+        init_dict, inputs_dict = self.prepare_init_args_and_inputs_for_common()
+        model = self.model_class(**init_dict).eval()
+        device_map = {name: device}
+        with tempfile.TemporaryDirectory() as tmpdir:
+            model.save_pretrained(tmpdir)
+            loaded_model = self.model_class.from_pretrained(tmpdir, device_map=device_map)
+            _ = loaded_model(**inputs_dict)
+
 
 @is_staging_test
 class ModelPushToHubTester(unittest.TestCase):

tests/models/unets/test_models_unet_2d_condition.py

@@ -46,7 +46,6 @@ from diffusers.utils.testing_utils import (
     require_peft_backend,
     require_torch_accelerator,
     require_torch_accelerator_with_fp16,
-    require_torch_gpu,
     skip_mps,
     slow,
     torch_all_close,
@@ -1084,42 +1083,6 @@ class UNet2DConditionModelTests(ModelTesterMixin, UNetTesterMixin, unittest.Test
         assert loaded_model
         assert new_output.sample.shape == (4, 4, 16, 16)
 
-    @parameterized.expand(
-        [
-            (-1, "You can't pass device_map as a negative int"),
-            ("foo", "When passing device_map as a string, the value needs to be a device name"),
-        ]
-    )
-    def test_wrong_device_map_raises_error(self, device_map, msg_substring):
-        with self.assertRaises(ValueError) as err_ctx:
-            _ = self.model_class.from_pretrained(
-                "hf-internal-testing/unet2d-sharded-dummy-subfolder", subfolder="unet", device_map=device_map
-            )
-
-        assert msg_substring in str(err_ctx.exception)
-
-    @parameterized.expand([0, "cuda", torch.device("cuda"), torch.device("cuda:0")])
-    @require_torch_gpu
-    def test_passing_non_dict_device_map_works(self, device_map):
-        _, inputs_dict = self.prepare_init_args_and_inputs_for_common()
-        loaded_model = self.model_class.from_pretrained(
-            "hf-internal-testing/unet2d-sharded-dummy-subfolder", subfolder="unet", device_map=device_map
-        )
-        output = loaded_model(**inputs_dict)
-        assert output.sample.shape == (4, 4, 16, 16)
-
-    @parameterized.expand([("", "cuda"), ("", torch.device("cuda"))])
-    @require_torch_gpu
-    def test_passing_dict_device_map_works(self, name, device_map):
-        # There are other valid dict-based `device_map` values too. It's best to refer to
-        # the docs for those: https://huggingface.co/docs/accelerate/en/concept_guides/big_model_inference#the-devicemap.
-        _, inputs_dict = self.prepare_init_args_and_inputs_for_common()
-        loaded_model = self.model_class.from_pretrained(
-            "hf-internal-testing/unet2d-sharded-dummy-subfolder", subfolder="unet", device_map={name: device_map}
-        )
-        output = loaded_model(**inputs_dict)
-        assert output.sample.shape == (4, 4, 16, 16)
-
     @require_peft_backend
     def test_load_attn_procs_raise_warning(self):
         init_dict, inputs_dict = self.prepare_init_args_and_inputs_for_common()

tests/pipelines/cosmos/test_cosmos2_text2image.py

@@ -0,0 +1,337 @@
+# Copyright 2024 The HuggingFace Team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import inspect
+import json
+import os
+import tempfile
+import unittest
+
+import numpy as np
+import torch
+from transformers import AutoTokenizer, T5EncoderModel
+
+from diffusers import (
+    AutoencoderKLWan,
+    Cosmos2TextToImagePipeline,
+    CosmosTransformer3DModel,
+    FlowMatchEulerDiscreteScheduler,
+)
+from diffusers.utils.testing_utils import enable_full_determinism, torch_device
+
+from ..pipeline_params import TEXT_TO_IMAGE_BATCH_PARAMS, TEXT_TO_IMAGE_IMAGE_PARAMS, TEXT_TO_IMAGE_PARAMS
+from ..test_pipelines_common import PipelineTesterMixin, to_np
+from .cosmos_guardrail import DummyCosmosSafetyChecker
+
+
+enable_full_determinism()
+
+
+class Cosmos2TextToImagePipelineWrapper(Cosmos2TextToImagePipeline):
+    @staticmethod
+    def from_pretrained(*args, **kwargs):
+        kwargs["safety_checker"] = DummyCosmosSafetyChecker()
+        return Cosmos2TextToImagePipeline.from_pretrained(*args, **kwargs)
+
+
+class Cosmos2TextToImagePipelineFastTests(PipelineTesterMixin, unittest.TestCase):
+    pipeline_class = Cosmos2TextToImagePipelineWrapper
+    params = TEXT_TO_IMAGE_PARAMS - {"cross_attention_kwargs"}
+    batch_params = TEXT_TO_IMAGE_BATCH_PARAMS
+    image_params = TEXT_TO_IMAGE_IMAGE_PARAMS
+    image_latents_params = TEXT_TO_IMAGE_IMAGE_PARAMS
+    required_optional_params = frozenset(
+        [
+            "num_inference_steps",
+            "generator",
+            "latents",
+            "return_dict",
+            "callback_on_step_end",
+            "callback_on_step_end_tensor_inputs",
+        ]
+    )
+    supports_dduf = False
+    test_xformers_attention = False
+    test_layerwise_casting = True
+    test_group_offloading = True
+
+    def get_dummy_components(self):
+        torch.manual_seed(0)
+        transformer = CosmosTransformer3DModel(
+            in_channels=16,
+            out_channels=16,
+            num_attention_heads=2,
+            attention_head_dim=16,
+            num_layers=2,
+            mlp_ratio=2,
+            text_embed_dim=32,
+            adaln_lora_dim=4,
+            max_size=(4, 32, 32),
+            patch_size=(1, 2, 2),
+            rope_scale=(2.0, 1.0, 1.0),
+            concat_padding_mask=True,
+            extra_pos_embed_type="learnable",
+        )
+
+        torch.manual_seed(0)
+        vae = AutoencoderKLWan(
+            base_dim=3,
+            z_dim=16,
+            dim_mult=[1, 1, 1, 1],
+            num_res_blocks=1,
+            temperal_downsample=[False, True, True],
+        )
+
+        torch.manual_seed(0)
+        scheduler = FlowMatchEulerDiscreteScheduler(use_karras_sigmas=True)
+        text_encoder = T5EncoderModel.from_pretrained("hf-internal-testing/tiny-random-t5")
+        tokenizer = AutoTokenizer.from_pretrained("hf-internal-testing/tiny-random-t5")
+
+        components = {
+            "transformer": transformer,
+            "vae": vae,
+            "scheduler": scheduler,
+            "text_encoder": text_encoder,
+            "tokenizer": tokenizer,
+            # We cannot run the Cosmos Guardrail for fast tests due to the large model size
+            "safety_checker": DummyCosmosSafetyChecker(),
+        }
+        return components
+
+    def get_dummy_inputs(self, device, seed=0):
+        if str(device).startswith("mps"):
+            generator = torch.manual_seed(seed)
+        else:
+            generator = torch.Generator(device=device).manual_seed(seed)
+
+        inputs = {
+            "prompt": "dance monkey",
+            "negative_prompt": "bad quality",
+            "generator": generator,
+            "num_inference_steps": 2,
+            "guidance_scale": 3.0,
+            "height": 32,
+            "width": 32,
+            "max_sequence_length": 16,
+            "output_type": "pt",
+        }
+
+        return inputs
+
+    def test_inference(self):
+        device = "cpu"
+
+        components = self.get_dummy_components()
+        pipe = self.pipeline_class(**components)
+        pipe.to(device)
+        pipe.set_progress_bar_config(disable=None)
+
+        inputs = self.get_dummy_inputs(device)
+        image = pipe(**inputs).images
+        generated_image = image[0]
+
+        self.assertEqual(generated_image.shape, (3, 32, 32))
+        expected_video = torch.randn(3, 32, 32)
+        max_diff = np.abs(generated_image - expected_video).max()
+        self.assertLessEqual(max_diff, 1e10)
+
+    def test_callback_inputs(self):
+        sig = inspect.signature(self.pipeline_class.__call__)
+        has_callback_tensor_inputs = "callback_on_step_end_tensor_inputs" in sig.parameters
+        has_callback_step_end = "callback_on_step_end" in sig.parameters
+
+        if not (has_callback_tensor_inputs and has_callback_step_end):
+            return
+
+        components = self.get_dummy_components()
+        pipe = self.pipeline_class(**components)
+        pipe = pipe.to(torch_device)
+        pipe.set_progress_bar_config(disable=None)
+        self.assertTrue(
+            hasattr(pipe, "_callback_tensor_inputs"),
+            f" {self.pipeline_class} should have `_callback_tensor_inputs` that defines a list of tensor variables its callback function can use as inputs",
+        )
+
+        def callback_inputs_subset(pipe, i, t, callback_kwargs):
+            # iterate over callback args
+            for tensor_name, tensor_value in callback_kwargs.items():
+                # check that we're only passing in allowed tensor inputs
+                assert tensor_name in pipe._callback_tensor_inputs
+
+            return callback_kwargs
+
+        def callback_inputs_all(pipe, i, t, callback_kwargs):
+            for tensor_name in pipe._callback_tensor_inputs:
+                assert tensor_name in callback_kwargs
+
+            # iterate over callback args
+            for tensor_name, tensor_value in callback_kwargs.items():
+                # check that we're only passing in allowed tensor inputs
+                assert tensor_name in pipe._callback_tensor_inputs
+
+            return callback_kwargs
+
+        inputs = self.get_dummy_inputs(torch_device)
+
+        # Test passing in a subset
+        inputs["callback_on_step_end"] = callback_inputs_subset
+        inputs["callback_on_step_end_tensor_inputs"] = ["latents"]
+        output = pipe(**inputs)[0]
+
+        # Test passing in a everything
+        inputs["callback_on_step_end"] = callback_inputs_all
+        inputs["callback_on_step_end_tensor_inputs"] = pipe._callback_tensor_inputs
+        output = pipe(**inputs)[0]
+
+        def callback_inputs_change_tensor(pipe, i, t, callback_kwargs):
+            is_last = i == (pipe.num_timesteps - 1)
+            if is_last:
+                callback_kwargs["latents"] = torch.zeros_like(callback_kwargs["latents"])
+            return callback_kwargs
+
+        inputs["callback_on_step_end"] = callback_inputs_change_tensor
+        inputs["callback_on_step_end_tensor_inputs"] = pipe._callback_tensor_inputs
+        output = pipe(**inputs)[0]
+        assert output.abs().sum() < 1e10
+
+    def test_inference_batch_single_identical(self):
+        self._test_inference_batch_single_identical(batch_size=3, expected_max_diff=1e-2)
+
+    def test_attention_slicing_forward_pass(
+        self, test_max_difference=True, test_mean_pixel_difference=True, expected_max_diff=1e-3
+    ):
+        if not self.test_attention_slicing:
+            return
+
+        components = self.get_dummy_components()
+        pipe = self.pipeline_class(**components)
+        for component in pipe.components.values():
+            if hasattr(component, "set_default_attn_processor"):
+                component.set_default_attn_processor()
+        pipe.to(torch_device)
+        pipe.set_progress_bar_config(disable=None)
+
+        generator_device = "cpu"
+        inputs = self.get_dummy_inputs(generator_device)
+        output_without_slicing = pipe(**inputs)[0]
+
+        pipe.enable_attention_slicing(slice_size=1)
+        inputs = self.get_dummy_inputs(generator_device)
+        output_with_slicing1 = pipe(**inputs)[0]
+
+        pipe.enable_attention_slicing(slice_size=2)
+        inputs = self.get_dummy_inputs(generator_device)
+        output_with_slicing2 = pipe(**inputs)[0]
+
+        if test_max_difference:
+            max_diff1 = np.abs(to_np(output_with_slicing1) - to_np(output_without_slicing)).max()
+            max_diff2 = np.abs(to_np(output_with_slicing2) - to_np(output_without_slicing)).max()
+            self.assertLess(
+                max(max_diff1, max_diff2),
+                expected_max_diff,
+                "Attention slicing should not affect the inference results",
+            )
+
+    def test_vae_tiling(self, expected_diff_max: float = 0.2):
+        generator_device = "cpu"
+        components = self.get_dummy_components()
+
+        pipe = self.pipeline_class(**components)
+        pipe.to("cpu")
+        pipe.set_progress_bar_config(disable=None)
+
+        # Without tiling
+        inputs = self.get_dummy_inputs(generator_device)
+        inputs["height"] = inputs["width"] = 128
+        output_without_tiling = pipe(**inputs)[0]
+
+        # With tiling
+        pipe.vae.enable_tiling(
+            tile_sample_min_height=96,
+            tile_sample_min_width=96,
+            tile_sample_stride_height=64,
+            tile_sample_stride_width=64,
+        )
+        inputs = self.get_dummy_inputs(generator_device)
+        inputs["height"] = inputs["width"] = 128
+        output_with_tiling = pipe(**inputs)[0]
+
+        self.assertLess(
+            (to_np(output_without_tiling) - to_np(output_with_tiling)).max(),
+            expected_diff_max,
+            "VAE tiling should not affect the inference results",
+        )
+
+    def test_save_load_optional_components(self, expected_max_difference=1e-4):
+        self.pipeline_class._optional_components.remove("safety_checker")
+        super().test_save_load_optional_components(expected_max_difference=expected_max_difference)
+        self.pipeline_class._optional_components.append("safety_checker")
+
+    def test_serialization_with_variants(self):
+        components = self.get_dummy_components()
+        pipe = self.pipeline_class(**components)
+        model_components = [
+            component_name
+            for component_name, component in pipe.components.items()
+            if isinstance(component, torch.nn.Module)
+        ]
+        model_components.remove("safety_checker")
+        variant = "fp16"
+
+        with tempfile.TemporaryDirectory() as tmpdir:
+            pipe.save_pretrained(tmpdir, variant=variant, safe_serialization=False)
+
+            with open(f"{tmpdir}/model_index.json", "r") as f:
+                config = json.load(f)
+
+            for subfolder in os.listdir(tmpdir):
+                if not os.path.isfile(subfolder) and subfolder in model_components:
+                    folder_path = os.path.join(tmpdir, subfolder)
+                    is_folder = os.path.isdir(folder_path) and subfolder in config
+                    assert is_folder and any(p.split(".")[1].startswith(variant) for p in os.listdir(folder_path))
+
+    def test_torch_dtype_dict(self):
+        components = self.get_dummy_components()
+        if not components:
+            self.skipTest("No dummy components defined.")
+
+        pipe = self.pipeline_class(**components)
+
+        specified_key = next(iter(components.keys()))
+
+        with tempfile.TemporaryDirectory(ignore_cleanup_errors=True) as tmpdirname:
+            pipe.save_pretrained(tmpdirname, safe_serialization=False)
+            torch_dtype_dict = {specified_key: torch.bfloat16, "default": torch.float16}
+            loaded_pipe = self.pipeline_class.from_pretrained(
+                tmpdirname, safety_checker=DummyCosmosSafetyChecker(), torch_dtype=torch_dtype_dict
+            )
+
+        for name, component in loaded_pipe.components.items():
+            if name == "safety_checker":
+                continue
+            if isinstance(component, torch.nn.Module) and hasattr(component, "dtype"):
+                expected_dtype = torch_dtype_dict.get(name, torch_dtype_dict.get("default", torch.float32))
+                self.assertEqual(
+                    component.dtype,
+                    expected_dtype,
+                    f"Component '{name}' has dtype {component.dtype} but expected {expected_dtype}",
+                )
+
+    @unittest.skip(
+        "The pipeline should not be runnable without a safety checker. The test creates a pipeline without passing in "
+        "a safety checker, which makes the pipeline default to the actual Cosmos Guardrail. The Cosmos Guardrail is "
+        "too large and slow to run on CI."
+    )
+    def test_encode_prompt_works_in_isolation(self):
+        pass

tests/pipelines/cosmos/test_cosmos2_video2world.py

@@ -0,0 +1,351 @@
+# Copyright 2024 The HuggingFace Team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import inspect
+import json
+import os
+import tempfile
+import unittest
+
+import numpy as np
+import PIL.Image
+import torch
+from transformers import AutoTokenizer, T5EncoderModel
+
+from diffusers import (
+    AutoencoderKLWan,
+    Cosmos2VideoToWorldPipeline,
+    CosmosTransformer3DModel,
+    FlowMatchEulerDiscreteScheduler,
+)
+from diffusers.utils.testing_utils import enable_full_determinism, torch_device
+
+from ..pipeline_params import TEXT_TO_IMAGE_BATCH_PARAMS, TEXT_TO_IMAGE_IMAGE_PARAMS, TEXT_TO_IMAGE_PARAMS
+from ..test_pipelines_common import PipelineTesterMixin, to_np
+from .cosmos_guardrail import DummyCosmosSafetyChecker
+
+
+enable_full_determinism()
+
+
+class Cosmos2VideoToWorldPipelineWrapper(Cosmos2VideoToWorldPipeline):
+    @staticmethod
+    def from_pretrained(*args, **kwargs):
+        kwargs["safety_checker"] = DummyCosmosSafetyChecker()
+        return Cosmos2VideoToWorldPipeline.from_pretrained(*args, **kwargs)
+
+
+class Cosmos2VideoToWorldPipelineFastTests(PipelineTesterMixin, unittest.TestCase):
+    pipeline_class = Cosmos2VideoToWorldPipelineWrapper
+    params = TEXT_TO_IMAGE_PARAMS - {"cross_attention_kwargs"}
+    batch_params = TEXT_TO_IMAGE_BATCH_PARAMS.union({"image", "video"})
+    image_params = TEXT_TO_IMAGE_IMAGE_PARAMS
+    image_latents_params = TEXT_TO_IMAGE_IMAGE_PARAMS
+    required_optional_params = frozenset(
+        [
+            "num_inference_steps",
+            "generator",
+            "latents",
+            "return_dict",
+            "callback_on_step_end",
+            "callback_on_step_end_tensor_inputs",
+        ]
+    )
+    supports_dduf = False
+    test_xformers_attention = False
+    test_layerwise_casting = True
+    test_group_offloading = True
+
+    def get_dummy_components(self):
+        torch.manual_seed(0)
+        transformer = CosmosTransformer3DModel(
+            in_channels=16 + 1,
+            out_channels=16,
+            num_attention_heads=2,
+            attention_head_dim=16,
+            num_layers=2,
+            mlp_ratio=2,
+            text_embed_dim=32,
+            adaln_lora_dim=4,
+            max_size=(4, 32, 32),
+            patch_size=(1, 2, 2),
+            rope_scale=(2.0, 1.0, 1.0),
+            concat_padding_mask=True,
+            extra_pos_embed_type="learnable",
+        )
+
+        torch.manual_seed(0)
+        vae = AutoencoderKLWan(
+            base_dim=3,
+            z_dim=16,
+            dim_mult=[1, 1, 1, 1],
+            num_res_blocks=1,
+            temperal_downsample=[False, True, True],
+        )
+
+        torch.manual_seed(0)
+        scheduler = FlowMatchEulerDiscreteScheduler(use_karras_sigmas=True)
+        text_encoder = T5EncoderModel.from_pretrained("hf-internal-testing/tiny-random-t5")
+        tokenizer = AutoTokenizer.from_pretrained("hf-internal-testing/tiny-random-t5")
+
+        components = {
+            "transformer": transformer,
+            "vae": vae,
+            "scheduler": scheduler,
+            "text_encoder": text_encoder,
+            "tokenizer": tokenizer,
+            # We cannot run the Cosmos Guardrail for fast tests due to the large model size
+            "safety_checker": DummyCosmosSafetyChecker(),
+        }
+        return components
+
+    def get_dummy_inputs(self, device, seed=0):
+        if str(device).startswith("mps"):
+            generator = torch.manual_seed(seed)
+        else:
+            generator = torch.Generator(device=device).manual_seed(seed)
+
+        image_height = 32
+        image_width = 32
+        image = PIL.Image.new("RGB", (image_width, image_height))
+
+        inputs = {
+            "image": image,
+            "prompt": "dance monkey",
+            "negative_prompt": "bad quality",
+            "generator": generator,
+            "num_inference_steps": 2,
+            "guidance_scale": 3.0,
+            "height": image_height,
+            "width": image_width,
+            "num_frames": 9,
+            "max_sequence_length": 16,
+            "output_type": "pt",
+        }
+
+        return inputs
+
+    def test_inference(self):
+        device = "cpu"
+
+        components = self.get_dummy_components()
+        pipe = self.pipeline_class(**components)
+        pipe.to(device)
+        pipe.set_progress_bar_config(disable=None)
+
+        inputs = self.get_dummy_inputs(device)
+        video = pipe(**inputs).frames
+        generated_video = video[0]
+
+        self.assertEqual(generated_video.shape, (9, 3, 32, 32))
+        expected_video = torch.randn(9, 3, 32, 32)
+        max_diff = np.abs(generated_video - expected_video).max()
+        self.assertLessEqual(max_diff, 1e10)
+
+    def test_components_function(self):
+        init_components = self.get_dummy_components()
+        init_components = {k: v for k, v in init_components.items() if not isinstance(v, (str, int, float))}
+        pipe = self.pipeline_class(**init_components)
+        self.assertTrue(hasattr(pipe, "components"))
+        self.assertTrue(set(pipe.components.keys()) == set(init_components.keys()))
+
+    def test_callback_inputs(self):
+        sig = inspect.signature(self.pipeline_class.__call__)
+        has_callback_tensor_inputs = "callback_on_step_end_tensor_inputs" in sig.parameters
+        has_callback_step_end = "callback_on_step_end" in sig.parameters
+
+        if not (has_callback_tensor_inputs and has_callback_step_end):
+            return
+
+        components = self.get_dummy_components()
+        pipe = self.pipeline_class(**components)
+        pipe = pipe.to(torch_device)
+        pipe.set_progress_bar_config(disable=None)
+        self.assertTrue(
+            hasattr(pipe, "_callback_tensor_inputs"),
+            f" {self.pipeline_class} should have `_callback_tensor_inputs` that defines a list of tensor variables its callback function can use as inputs",
+        )
+
+        def callback_inputs_subset(pipe, i, t, callback_kwargs):
+            # iterate over callback args
+            for tensor_name, tensor_value in callback_kwargs.items():
+                # check that we're only passing in allowed tensor inputs
+                assert tensor_name in pipe._callback_tensor_inputs
+
+            return callback_kwargs
+
+        def callback_inputs_all(pipe, i, t, callback_kwargs):
+            for tensor_name in pipe._callback_tensor_inputs:
+                assert tensor_name in callback_kwargs
+
+            # iterate over callback args
+            for tensor_name, tensor_value in callback_kwargs.items():
+                # check that we're only passing in allowed tensor inputs
+                assert tensor_name in pipe._callback_tensor_inputs
+
+            return callback_kwargs
+
+        inputs = self.get_dummy_inputs(torch_device)
+
+        # Test passing in a subset
+        inputs["callback_on_step_end"] = callback_inputs_subset
+        inputs["callback_on_step_end_tensor_inputs"] = ["latents"]
+        output = pipe(**inputs)[0]
+
+        # Test passing in a everything
+        inputs["callback_on_step_end"] = callback_inputs_all
+        inputs["callback_on_step_end_tensor_inputs"] = pipe._callback_tensor_inputs
+        output = pipe(**inputs)[0]
+
+        def callback_inputs_change_tensor(pipe, i, t, callback_kwargs):
+            is_last = i == (pipe.num_timesteps - 1)
+            if is_last:
+                callback_kwargs["latents"] = torch.zeros_like(callback_kwargs["latents"])
+            return callback_kwargs
+
+        inputs["callback_on_step_end"] = callback_inputs_change_tensor
+        inputs["callback_on_step_end_tensor_inputs"] = pipe._callback_tensor_inputs
+        output = pipe(**inputs)[0]
+        assert output.abs().sum() < 1e10
+
+    def test_inference_batch_single_identical(self):
+        self._test_inference_batch_single_identical(batch_size=3, expected_max_diff=1e-2)
+
+    def test_attention_slicing_forward_pass(
+        self, test_max_difference=True, test_mean_pixel_difference=True, expected_max_diff=1e-3
+    ):
+        if not self.test_attention_slicing:
+            return
+
+        components = self.get_dummy_components()
+        pipe = self.pipeline_class(**components)
+        for component in pipe.components.values():
+            if hasattr(component, "set_default_attn_processor"):
+                component.set_default_attn_processor()
+        pipe.to(torch_device)
+        pipe.set_progress_bar_config(disable=None)
+
+        generator_device = "cpu"
+        inputs = self.get_dummy_inputs(generator_device)
+        output_without_slicing = pipe(**inputs)[0]
+
+        pipe.enable_attention_slicing(slice_size=1)
+        inputs = self.get_dummy_inputs(generator_device)
+        output_with_slicing1 = pipe(**inputs)[0]
+
+        pipe.enable_attention_slicing(slice_size=2)
+        inputs = self.get_dummy_inputs(generator_device)
+        output_with_slicing2 = pipe(**inputs)[0]
+
+        if test_max_difference:
+            max_diff1 = np.abs(to_np(output_with_slicing1) - to_np(output_without_slicing)).max()
+            max_diff2 = np.abs(to_np(output_with_slicing2) - to_np(output_without_slicing)).max()
+            self.assertLess(
+                max(max_diff1, max_diff2),
+                expected_max_diff,
+                "Attention slicing should not affect the inference results",
+            )
+
+    def test_vae_tiling(self, expected_diff_max: float = 0.2):
+        generator_device = "cpu"
+        components = self.get_dummy_components()
+
+        pipe = self.pipeline_class(**components)
+        pipe.to("cpu")
+        pipe.set_progress_bar_config(disable=None)
+
+        # Without tiling
+        inputs = self.get_dummy_inputs(generator_device)
+        inputs["height"] = inputs["width"] = 128
+        output_without_tiling = pipe(**inputs)[0]
+
+        # With tiling
+        pipe.vae.enable_tiling(
+            tile_sample_min_height=96,
+            tile_sample_min_width=96,
+            tile_sample_stride_height=64,
+            tile_sample_stride_width=64,
+        )
+        inputs = self.get_dummy_inputs(generator_device)
+        inputs["height"] = inputs["width"] = 128
+        output_with_tiling = pipe(**inputs)[0]
+
+        self.assertLess(
+            (to_np(output_without_tiling) - to_np(output_with_tiling)).max(),
+            expected_diff_max,
+            "VAE tiling should not affect the inference results",
+        )
+
+    def test_save_load_optional_components(self, expected_max_difference=1e-4):
+        self.pipeline_class._optional_components.remove("safety_checker")
+        super().test_save_load_optional_components(expected_max_difference=expected_max_difference)
+        self.pipeline_class._optional_components.append("safety_checker")
+
+    def test_serialization_with_variants(self):
+        components = self.get_dummy_components()
+        pipe = self.pipeline_class(**components)
+        model_components = [
+            component_name
+            for component_name, component in pipe.components.items()
+            if isinstance(component, torch.nn.Module)
+        ]
+        model_components.remove("safety_checker")
+        variant = "fp16"
+
+        with tempfile.TemporaryDirectory() as tmpdir:
+            pipe.save_pretrained(tmpdir, variant=variant, safe_serialization=False)
+
+            with open(f"{tmpdir}/model_index.json", "r") as f:
+                config = json.load(f)
+
+            for subfolder in os.listdir(tmpdir):
+                if not os.path.isfile(subfolder) and subfolder in model_components:
+                    folder_path = os.path.join(tmpdir, subfolder)
+                    is_folder = os.path.isdir(folder_path) and subfolder in config
+                    assert is_folder and any(p.split(".")[1].startswith(variant) for p in os.listdir(folder_path))
+
+    def test_torch_dtype_dict(self):
+        components = self.get_dummy_components()
+        if not components:
+            self.skipTest("No dummy components defined.")
+
+        pipe = self.pipeline_class(**components)
+
+        specified_key = next(iter(components.keys()))
+
+        with tempfile.TemporaryDirectory(ignore_cleanup_errors=True) as tmpdirname:
+            pipe.save_pretrained(tmpdirname, safe_serialization=False)
+            torch_dtype_dict = {specified_key: torch.bfloat16, "default": torch.float16}
+            loaded_pipe = self.pipeline_class.from_pretrained(
+                tmpdirname, safety_checker=DummyCosmosSafetyChecker(), torch_dtype=torch_dtype_dict
+            )
+
+        for name, component in loaded_pipe.components.items():
+            if name == "safety_checker":
+                continue
+            if isinstance(component, torch.nn.Module) and hasattr(component, "dtype"):
+                expected_dtype = torch_dtype_dict.get(name, torch_dtype_dict.get("default", torch.float32))
+                self.assertEqual(
+                    component.dtype,
+                    expected_dtype,
+                    f"Component '{name}' has dtype {component.dtype} but expected {expected_dtype}",
+                )
+
+    @unittest.skip(
+        "The pipeline should not be runnable without a safety checker. The test creates a pipeline without passing in "
+        "a safety checker, which makes the pipeline default to the actual Cosmos Guardrail. The Cosmos Guardrail is "
+        "too large and slow to run on CI."
+    )
+    def test_encode_prompt_works_in_isolation(self):
+        pass