Spaces:
Running on Zero
Running on Zero
File size: 4,638 Bytes
3db1cbc | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 | # Copyright 2026 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.
from typing import Any
import numpy as np
import PIL
import torch
from ...configuration_utils import FrozenDict
from ...models import AutoencoderKLWan
from ...utils import logging
from ...video_processor import VideoProcessor
from ..modular_pipeline import ModularPipelineBlocks, PipelineState
from ..modular_pipeline_utils import ComponentSpec, InputParam, OutputParam
logger = logging.get_logger(__name__) # pylint: disable=invalid-name
class WanVaceTrimReferenceLatentsStep(ModularPipelineBlocks):
model_name = "wan-vace"
@property
def description(self) -> str:
return "Step that removes the prepended reference image frames from the denoised latents before decoding"
@property
def inputs(self) -> list[InputParam]:
return [
InputParam(
"latents",
required=True,
type_hint=torch.Tensor,
description="The denoised latents from the denoising step",
),
InputParam(
"num_reference_images",
type_hint=int,
default=0,
description="Number of reference image frames to remove from the front of the latents. Can be generated in vace_encoder step.",
),
]
@torch.no_grad()
def __call__(self, components, state: PipelineState) -> PipelineState:
block_state = self.get_block_state(state)
block_state.latents = block_state.latents[:, :, block_state.num_reference_images :]
self.set_block_state(state, block_state)
return components, state
class WanVaeDecoderStep(ModularPipelineBlocks):
model_name = "wan"
@property
def expected_components(self) -> list[ComponentSpec]:
return [
ComponentSpec("vae", AutoencoderKLWan),
ComponentSpec(
"video_processor",
VideoProcessor,
config=FrozenDict({"vae_scale_factor": 8}),
default_creation_method="from_config",
),
]
@property
def description(self) -> str:
return "Step that decodes the denoised latents into images"
@property
def inputs(self) -> list[tuple[str, Any]]:
return [
InputParam(
"latents",
required=True,
type_hint=torch.Tensor,
description="The denoised latents from the denoising step",
),
InputParam(
"output_type", default="np", type_hint=str, description="The output type of the decoded videos"
),
]
@property
def intermediate_outputs(self) -> list[str]:
return [
OutputParam(
"videos",
type_hint=list[list[PIL.Image.Image]] | list[torch.Tensor] | list[np.ndarray],
description="The generated videos, can be a PIL.Image.Image, torch.Tensor or a numpy array",
)
]
@torch.no_grad()
def __call__(self, components, state: PipelineState) -> PipelineState:
block_state = self.get_block_state(state)
vae_dtype = components.vae.dtype
latents = block_state.latents
latents_mean = (
torch.tensor(components.vae.config.latents_mean)
.view(1, components.vae.config.z_dim, 1, 1, 1)
.to(latents.device, latents.dtype)
)
latents_std = 1.0 / torch.tensor(components.vae.config.latents_std).view(
1, components.vae.config.z_dim, 1, 1, 1
).to(latents.device, latents.dtype)
latents = latents / latents_std + latents_mean
latents = latents.to(vae_dtype)
block_state.videos = components.vae.decode(latents, return_dict=False)[0]
output_type = getattr(block_state, "output_type", "np")
block_state.videos = components.video_processor.postprocess_video(block_state.videos, output_type=output_type)
self.set_block_state(state, block_state)
return components, state
|