add diffusers compat imports and mp4 output format

- Add try/except imports for diffusers >= 0.21 compatibility (PR guoyww#435):
  ModelMixin, CrossAttention->Attention, DiffusionPipeline path changes
- Widen diffusers version range to >=0.16.0 (no upper bound)
- Add --format flag to animate.py for mp4/gif output (PR guoyww#37)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: GitLab CI

animatediff/models/attention.py

@@ -8,10 +8,17 @@
 from torch import nn
 
 from diffusers.configuration_utils import ConfigMixin, register_to_config
-from diffusers.modeling_utils import ModelMixin
+try:
+    from diffusers.modeling_utils import ModelMixin
+except ImportError:
+    from diffusers.models.modeling_utils import ModelMixin
 from diffusers.utils import BaseOutput
 from diffusers.utils.import_utils import is_xformers_available
-from diffusers.models.attention import CrossAttention, FeedForward, AdaLayerNorm
+try:
+    from diffusers.models.attention import CrossAttention, FeedForward, AdaLayerNorm
+except ImportError:
+    from diffusers.models.attention_processor import Attention as CrossAttention
+    from diffusers.models.attention import FeedForward, AdaLayerNorm
 
 from einops import rearrange, repeat
 

animatediff/models/motion_module.py

@@ -8,10 +8,17 @@
 import torchvision
 
 from diffusers.configuration_utils import ConfigMixin, register_to_config
-from diffusers.modeling_utils import ModelMixin
+try:
+    from diffusers.modeling_utils import ModelMixin
+except ImportError:
+    from diffusers.models.modeling_utils import ModelMixin
 from diffusers.utils import BaseOutput
 from diffusers.utils.import_utils import is_xformers_available
-from diffusers.models.attention import CrossAttention, FeedForward
+try:
+    from diffusers.models.attention import CrossAttention, FeedForward
+except ImportError:
+    from diffusers.models.attention_processor import Attention as CrossAttention
+    from diffusers.models.attention import FeedForward
 
 from einops import rearrange, repeat
 import math

animatediff/models/sparse_controlnet.py

@@ -23,7 +23,10 @@
 from diffusers.configuration_utils import ConfigMixin, register_to_config
 from diffusers.utils import BaseOutput, logging
 from diffusers.models.embeddings import TimestepEmbedding, Timesteps
-from diffusers.modeling_utils import ModelMixin
+try:
+    from diffusers.modeling_utils import ModelMixin
+except ImportError:
+    from diffusers.models.modeling_utils import ModelMixin
 
 
 from .unet_blocks import (

animatediff/models/unet.py

@@ -12,7 +12,10 @@
 import torch.utils.checkpoint
 
 from diffusers.configuration_utils import ConfigMixin, register_to_config
-from diffusers.modeling_utils import ModelMixin
+try:
+    from diffusers.modeling_utils import ModelMixin
+except ImportError:
+    from diffusers.models.modeling_utils import ModelMixin
 from diffusers.utils import BaseOutput, logging
 from diffusers.models.embeddings import TimestepEmbedding, Timesteps
 from .unet_blocks import (

animatediff/pipelines/pipeline_animation.py

@@ -14,7 +14,10 @@
 
 from diffusers.configuration_utils import FrozenDict
 from diffusers.models import AutoencoderKL
-from diffusers.pipeline_utils import DiffusionPipeline
+try:
+    from diffusers.pipeline_utils import DiffusionPipeline
+except ImportError:
+    from diffusers.pipelines.pipeline_utils import DiffusionPipeline
 from diffusers.schedulers import (
     DDIMScheduler,
     DPMSolverMultistepScheduler,

requirements.txt

@@ -1,6 +1,6 @@
 torch>=2.0.0
 torchvision>=0.15.0
-diffusers>=0.16.0,<0.18.0
+diffusers>=0.16.0
 transformers>=4.25.0
 xformers>=0.0.20; sys_platform != 'darwin'
 imageio>=2.27.0

scripts/animate.py

@@ -35,6 +35,7 @@ def main(args):
 
     time_str = datetime.datetime.now().strftime("%Y-%m-%dT%H-%M-%S")
     savedir = f"samples/{Path(args.config).stem}-{time_str}"
+    extension = args.format
     os.makedirs(savedir)
 
     config  = OmegaConf.load(args.config)
@@ -170,13 +171,13 @@ def image_norm(image):
             samples.append(sample)
 
             prompt = "-".join((prompt.replace("/", "").split(" ")[:10]))
-            save_videos_grid(sample, f"{savedir}/sample/{sample_idx}-{prompt}.gif")
-            print(f"save to {savedir}/sample/{prompt}.gif")
+            save_videos_grid(sample, f"{savedir}/sample/{sample_idx}-{prompt}.{extension}")
+            print(f"save to {savedir}/sample/{prompt}.{extension}")
 
             sample_idx += 1
 
     samples = torch.concat(samples)
-    save_videos_grid(samples, f"{savedir}/sample.gif", n_rows=4)
+    save_videos_grid(samples, f"{savedir}/sample.{extension}", n_rows=4)
 
     OmegaConf.save(config, f"{savedir}/config.yaml")
 
@@ -192,6 +193,7 @@ def image_norm(image):
     parser.add_argument("--H", type=int, default=512)
 
     parser.add_argument("--without-xformers", action="store_true")
+    parser.add_argument("--format", type=str, default="gif", choices=["gif", "mp4"])
 
     args = parser.parse_args()
     main(args)