diffusers
[WIP]Add Wan2.2 Animate Pipeline (Continuation of #12442 by tolgacangoz)
#12526
Merged

[WIP]Add Wan2.2 Animate Pipeline (Continuation of #12442 by tolgacangoz) #12526

dg845
tolgacangoz template1
3529a0ad
tolgacangoz temp2
4f2ee5e7
tolgacangoz up
778fb54f
tolgacangoz up
d77b6baf
tolgacangoz fix-copies
2fc6ac26
tolgacangoz Add support for Wan2.2-Animate-14B model in convert_wan_to_diffusers.py
d667d03a
tolgacangoz style
6182d44f
tolgacangoz Refactor WanAnimate model components
8c9fd890
tolgacangoz Enhance `WanAnimatePipeline` with new parameters for mode and tempora…
d01e9419
tolgacangoz Update `WanAnimatePipeline` to require additional video inputs and im…
7af953b2
tolgacangoz Add Wan 2.2 Animate 14B model support and introduce Wan-Animate frame…
a0372e36
tolgacangoz Add unit test template for `WanAnimatePipeline` functionality
05a01c66
tolgacangoz Add unit tests for `WanAnimateTransformer3DModel` in GGUF format
22b83ce8
tolgacangoz style
7fb67322
tolgacangoz Improve the template of `transformer_wan_animate.py`
3e6f893c
tolgacangoz Update `WanAnimatePipeline`
624a3148
tolgacangoz style
fc0edb59
tolgacangoz Refactor test for `WanAnimatePipeline` to include new input structure
eb7eeddd
tolgacangoz from `einops` to `torch`
8968b429
tolgacangoz Merge branch 'main' into integrations/wan2.2-animate
dce83a8e
tolgacangoz Add padding functionality to `WanAnimatePipeline` for video frames
75b2382d
tolgacangoz style
802896e9
tolgacangoz Enhance `WanAnimatePipeline` with additional input parameters for imp…
e06098f8
tolgacangoz up
84768f6d
tolgacangoz Refactor `WanAnimatePipeline` for improved tensor handling and mask g…
06e61380
tolgacangoz Refactor `WanAnimatePipeline` to streamline latent tensor processing …
5777ce04
tolgacangoz style
b8337c69
tolgacangoz Add new layers and functions to `transformer_wan_animate.py` for enha…
f4eb9a05
tolgacangoz Merge branch 'main' into integrations/wan2.2-animate
4e6651b9
tolgacangoz Refactor `transformer_wan_animate.py` to improve modularity and type …
d80ae195
tolgacangoz Refactor `transformer_wan_animate.py` to enhance modularity and updat…
348a9450
tolgacangoz Update the `ConvLayer` class to conditionally apply bias based on act…
7774421c
tolgacangoz Simplify
a5536e2f
tolgacangoz refactor transformer
6a8662d6
tolgacangoz Enhance `convert_wan_to_diffusers.py` for Animate model integration
96a126ad
tolgacangoz Merge branch 'main' into integrations/wan2.2-animate
050b3137
tolgacangoz Enhance `convert_wan_to_diffusers.py` and `WanAnimatePipeline` for im…
0566e5df
tolgacangoz simplify
fe02c25c
tolgacangoz Refactor `WanAnimatePipeline` to enhance reference image handling and…
04ab2623
tolgacangoz Enhance weight conversion logic in `convert_wan_to_diffusers.py`
7bfbd935
tolgacangoz Enhance documentation and tests for WanAnimatePipeline, adding exampl…
7092a28d
tolgacangoz Merge branch 'main' into integrations/wan2.2-animate
5d01574d
tolgacangoz Clarify contribution of M. Tolga Cangöz
9c0a65d8
tolgacangoz Update face_embedder key mappings in `convert_wan_to_diffusers.py`
28ac516f
tolgacangoz up
b71d3a9a
tolgacangoz up
5818d71c
tolgacangoz Fix image embedding extraction in WanAnimatePipeline to return the la…
bfda25dd
tolgacangoz Adjust default parameters in WanAnimatePipeline for num_frames, num_i…
0ac259c6
tolgacangoz Update example docstring parameters for num_frames and guidance_scale…
e2e95ede
tolgacangoz Refactor tests in WanAnimatePipeline: remove redundant assertions and…
7146bb05
dg845 Add fused relu for Wan animate activations
6ffdb993
HuggingFaceDocBuilderDev
tolgacangoz
tolgacangoz commented on 2025-10-22
a-free-a
yiyixuxu yiyixuxu changed the title Add Wan2.2 Animate Pipeline (Continuation of #12442 by tolgacangoz) [WIP]Add Wan2.2 Animate Pipeline (Continuation of #12442 by tolgacangoz) 153 days ago
dg845 dg845 marked this pull request as draft 153 days ago
dg845 Refactor motion encoder to use custom Conv2d and Linear with weight s…
4556730c
dg845 Refactor WanAnimateFaceEncoder to make it easier to understand
c3e69fc2
dg845 Refactor Wan Animate transformer to reuse WanTimeTextImageEmbedding
7f4dde9d
dg845 Refactor Wan Animate face blocks to use an attention processor
4f204ec5
dg845 Refactor Wan Animate transformer, taking into account previous changes
57e9ea30
dg845 Remove unused imports in transformer_wan_animate
091b7ce2
dg845 Merge branch 'main' into add-wan2.2-animate-pipeline
8216aef3
dg845 Add initial Wan Animate transformer tests
50329d73
dg845 Refactor face block attn into its own Attention class and fix some bugs
275d324f
dg845 Fix issues (such as device placement issues) to get remaining transfo…
ac2962db
dg845 Update Wan Animate conversion script to reflect changes to transformer
0145135f
dg845 Add _repeated_blocks to Wan Animate transformer for regional compilation
bdbd1419
dg845 Refactor Wan Animate pipeline to make latent preparation code more clear
25371339
dg845 Update Wan Animate pipeline tests after transformer an pipeline changes
332d3c2c
dg845 Merge branch 'main' into add-wan2.2-animate-pipeline
99e56e33
dg845 Fix some batching and device placement issues in Wan Animate pipeline
00ddbb98
dg845 Remove reference_images tests for Wan Animate
1e1e7069
dg845 Get Wan Animate pipeline fp16 inference tests working
a56bee14
dg845 Skip test_callback_inputs since the Wan Animate pipline is not compat…
6fb5ca8d
dg845 Fix mask video shapes for Wan Animate replacement
1e61ed7a
dg845 Use a separate VaeImageProcessor for the reference image as it uses d…
e2846f6d
dg845 Fix some more Wan Animate pipeline shape errors
3a80241a
dg845 Fix more bugs in Wan Animate pipeline
86be6005
dg845 Ensure that the replacement mask only has one channel
6748d25e
dg845 Support Wan Animate image preprocessing, fix bugs, clean up code
f696682d
dg845 Add docs for WanAnimateTransformer3DModel
80d9f8bf
dg845 make style and make quality
d9c6bc68
dg845 Merge branch 'main' into add-wan2.2-animate-pipeline
4e415d3d
dg845 Fix first segment I2V mask for prev segement cond latents
cbfc0ad4
dg845 Use same Open CLIP checkpoint as other Wan2.1-based models
b80be86c
dg845 Merge branch 'main' into add-wan2.2-animate-pipeline
d87baa50
dg845 dg845 marked this pull request as ready for review 135 days ago
dg845 dg845 requested a review from yiyixuxu yiyixuxu 135 days ago
dg845
yiyixuxu
yiyixuxu commented on 2025-11-10
yiyixuxu
yiyixuxu commented on 2025-11-11
dg845 Copy Wan blocks for Wan Animate with # Copied from
6420f0e7
dg845 Get regional compilation working without recompilation
dd680ee6
dg845 Remove Wan2.2 TI2V timestep logic as Wan Animate is based on Wan 2.1
6d92b3e8
dg845 Move motion encoder batch inference logic to forward and remove the m…
d0c77503
dg845 Move (de)standardize latents logic into Wan Animate pipeline __call__
c2ec703f
dg845 Move Wan Animate ref image processing logic to its own VaeImageProces…
2f549ee6
dg845 make style and make quality
68da86a9
dg845 Make motion encoder inference batch size configurable from Wan Animat…
cb7977ec
dg845 Avoid list comprehension for batched motion encoder inference as it u…
847e4a2d
dg845 Address more review comments
e4b1db00
dg845 Merge branch 'main' into add-wan2.2-animate-pipeline
f0a0d21d
dg845 make style, make quality, make fix-copies
e96f6385
dg845 Make motion_encode_batch_size configurable in pipeline __call__
a6ddd02d
dg845 Merge branch 'main' into add-wan2.2-animate-pipeline
6ad82e51
dg845 Update Wan Animate pipeline example
e74373b7
dg845 Have Wan image processor take into account the spatial patch size as …
2259ded8
dg845 dg845 requested a review from yiyixuxu yiyixuxu 133 days ago
yiyixuxu
yiyixuxu approved these changes on 2025-11-12
yiyixuxu yiyixuxu merged d8e48058 into main 132 days ago
dg845 dg845 deleted the add-wan2.2-animate-pipeline branch 125 days ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone