dylanyunlong/project_6
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

feat(CRITICAL): 从 GitHub 扫描搬运 ixformer SDK + xllm 完整 GDN/MoE 代码

Browse Source 
来源:
  1. Chranos/ixformer (GitHub) → ixformer_sdk/ (230 files, 70K lines)
     - inference/functions/vllm.py: vllm_moe_topk_softmax 完整实现 (2033 lines)
     - inference/functions/moe.py: MoE ops 完整实现 (1380 lines)
     - contrib/vllm_flash_attn/: FA2 Python 接口 (1018 lines)
     - contrib/tgi/fused_moe.py: TGI fused MoE (429 lines)
     - csrc/include/ixformer/: C++ kernel headers + cmake

  2. Deep-Spark/xllm (GitHub) → upstream_ref/xllm_latest/ (+15 files)
     - npu_torch/qwen3_5_decoder_layer_impl.cpp/.h
     - npu_torch/qwen3_5_gated_delta_net.cpp/.h
     - npu_torch/qwen3_next_*.cpp/.h (6 files)
     - npu_torch/attention.cpp/.h + fused_moe.cpp/.h + CMakeLists.txt
     - models/llm/qwen3_5.h + qwen3_5_mtp.h + qwen3_next.h
     - models/vlm/qwen3_5.h

调用链完整性:
  ixformer_sdk/inference/functions/vllm.py
    → ops.infer.moe_topk_softmax() (C++ 层)
    → 这就是 base 镜像 libixformer.so 里的实现

  upstream_ref/xllm_latest/core/layers/ilu/fused_moe.cpp
    → ixformer::infer::topk_softmax() (直接 C++ 调用)
    → ixformer::infer::group_gemm() → 完整 7-step MoE pipeline
This commit is contained in:
project6-dev
2026-08-11 02:31:56 +00:00
parent a8b16da5da
commit 87a19d2d00
250 changed files with 76690 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

0
ixformer_sdk/contrib/DeepCache/svd/__init__.py Normal file
View File

659
ixformer_sdk/contrib/DeepCache/svd/pipeline_stable_video_diffusion.py Normal file
View File

@@ -0,0 +1,659 @@
# Copyright 2023 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 dataclasses import dataclass
from typing import Callable, Dict, List, Optional, Union
import numpy as np
import PIL.Image
import torch
from transformers import CLIPImageProcessor, CLIPVisionModelWithProjection
from diffusers.image_processor import VaeImageProcessor
from diffusers.models import AutoencoderKLTemporalDecoder, UNetSpatioTemporalConditionModel
from diffusers.schedulers import EulerDiscreteScheduler
from diffusers.utils import BaseOutput, logging
from diffusers.utils.torch_utils import randn_tensor
from .pipeline_utils import DiffusionPipeline
logger = logging.get_logger(__name__) # pylint: disable=invalid-name
def _append_dims(x, target_dims):
"""Appends dimensions to the end of a tensor until it has target_dims dimensions."""
dims_to_append = target_dims - x.ndim
if dims_to_append < 0:
raise ValueError(f"input has {x.ndim} dims but target_dims is {target_dims}, which is less")
return x[(...,) + (None,) * dims_to_append]
def tensor2vid(video: torch.Tensor, processor, output_type="np"):
# Based on:
# https://github.com/modelscope/modelscope/blob/1509fdb973e5871f37148a4b5e5964cafd43e64d/modelscope/pipelines/multi_modal/text_to_video_synthesis_pipeline.py#L78
batch_size, channels, num_frames, height, width = video.shape
outputs = []
for batch_idx in range(batch_size):
batch_vid = video[batch_idx].permute(1, 0, 2, 3)
batch_output = processor.postprocess(batch_vid, output_type)
outputs.append(batch_output)
return outputs
@dataclass
class StableVideoDiffusionPipelineOutput(BaseOutput):
r"""
Output class for zero-shot text-to-video pipeline.
Args:
frames (`[List[PIL.Image.Image]`, `np.ndarray`]):
List of denoised PIL images of length `batch_size` or NumPy array of shape `(batch_size, height, width,
num_channels)`.
"""
frames: Union[List[PIL.Image.Image], np.ndarray]
class StableVideoDiffusionPipeline(DiffusionPipeline):
r"""
Pipeline to generate video from an input image using Stable Video Diffusion.
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:
vae ([`AutoencoderKL`]):
Variational Auto-Encoder (VAE) model to encode and decode images to and from latent representations.
image_encoder ([`~transformers.CLIPVisionModelWithProjection`]):
Frozen CLIP image-encoder ([laion/CLIP-ViT-H-14-laion2B-s32B-b79K](https://huggingface.co/laion/CLIP-ViT-H-14-laion2B-s32B-b79K)).
unet ([`UNetSpatioTemporalConditionModel`]):cache_interval=5, cache_branch=0,
A `UNetSpatioTemporalConditionModel` to denoise the encoded image latents.
scheduler ([`EulerDiscreteScheduler`]):
A scheduler to be used in combination with `unet` to denoise the encoded image latents.
feature_extractor ([`~transformers.CLIPImageProcessor`]):
A `CLIPImageProcessor` to extract features from generated images.
"""
model_cpu_offload_seq = "image_encoder->unet->vae"
_callback_tensor_inputs = ["latents"]
def __init__(
self,
vae: AutoencoderKLTemporalDecoder,
image_encoder: CLIPVisionModelWithProjection,
unet: UNetSpatioTemporalConditionModel,
scheduler: EulerDiscreteScheduler,
feature_extractor: CLIPImageProcessor,
):
super().__init__()
self.register_modules(
vae=vae,
image_encoder=image_encoder,
unet=unet,
scheduler=scheduler,
feature_extractor=feature_extractor,
)
self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
def _encode_image(self, image, device, num_videos_per_prompt, do_classifier_free_guidance):
dtype = next(self.image_encoder.parameters()).dtype
if not isinstance(image, torch.Tensor):
image = self.image_processor.pil_to_numpy(image)
image = self.image_processor.numpy_to_pt(image)
# We normalize the image before resizing to match with the original implementation.
# Then we unnormalize it after resizing.
image = image * 2.0 - 1.0
image = _resize_with_antialiasing(image, (224, 224))
image = (image + 1.0) / 2.0
# Normalize the image with for CLIP input
image = self.feature_extractor(
images=image,
do_normalize=True,
do_center_crop=False,
do_resize=False,
do_rescale=False,
return_tensors="pt",
).pixel_values
image = image.to(device=device, dtype=dtype)
image_embeddings = self.image_encoder(image).image_embeds
image_embeddings = image_embeddings.unsqueeze(1)
# duplicate image embeddings for each generation per prompt, using mps friendly method
bs_embed, seq_len, _ = image_embeddings.shape
image_embeddings = image_embeddings.repeat(1, num_videos_per_prompt, 1)
image_embeddings = image_embeddings.view(bs_embed * num_videos_per_prompt, seq_len, -1)
if do_classifier_free_guidance:
negative_image_embeddings = torch.zeros_like(image_embeddings)
# For classifier free guidance, we need to do two forward passes.
# Here we concatenate the unconditional and text embeddings into a single batch
# to avoid doing two forward passes
image_embeddings = torch.cat([negative_image_embeddings, image_embeddings])
return image_embeddings
def _encode_vae_image(
self,
image: torch.Tensor,
device,
num_videos_per_prompt,
do_classifier_free_guidance,
):
image = image.to(device=device)
image_latents = self.vae.encode(image).latent_dist.mode()
if do_classifier_free_guidance:
negative_image_latents = torch.zeros_like(image_latents)
# For classifier free guidance, we need to do two forward passes.
# Here we concatenate the unconditional and text embeddings into a single batch
# to avoid doing two forward passes
image_latents = torch.cat([negative_image_latents, image_latents])
# duplicate image_latents for each generation per prompt, using mps friendly method
image_latents = image_latents.repeat(num_videos_per_prompt, 1, 1, 1)
return image_latents
def _get_add_time_ids(
self,
fps,
motion_bucket_id,
noise_aug_strength,
dtype,
batch_size,
num_videos_per_prompt,
do_classifier_free_guidance,
):
add_time_ids = [fps, motion_bucket_id, noise_aug_strength]
passed_add_embed_dim = self.unet.config.addition_time_embed_dim * len(add_time_ids)
expected_add_embed_dim = self.unet.add_embedding.linear_1.in_features
if expected_add_embed_dim != passed_add_embed_dim:
raise ValueError(
f"Model expects an added time embedding vector of length {expected_add_embed_dim}, but a vector of {passed_add_embed_dim} was created. The model has an incorrect config. Please check `unet.config.time_embedding_type` and `text_encoder_2.config.projection_dim`."
)
add_time_ids = torch.tensor([add_time_ids], dtype=dtype)
add_time_ids = add_time_ids.repeat(batch_size * num_videos_per_prompt, 1)
if do_classifier_free_guidance:
add_time_ids = torch.cat([add_time_ids, add_time_ids])
return add_time_ids
def decode_latents(self, latents, num_frames, decode_chunk_size=14):
# [batch, frames, channels, height, width] -> [batch*frames, channels, height, width]
latents = latents.flatten(0, 1)
latents = 1 / self.vae.config.scaling_factor * latents
accepts_num_frames = "num_frames" in set(inspect.signature(self.vae.forward).parameters.keys())
# decode decode_chunk_size frames at a time to avoid OOM
frames = []
for i in range(0, latents.shape[0], decode_chunk_size):
num_frames_in = latents[i : i + decode_chunk_size].shape[0]
decode_kwargs = {}
if accepts_num_frames:
# we only pass num_frames_in if it's expected
decode_kwargs["num_frames"] = num_frames_in
frame = self.vae.decode(latents[i : i + decode_chunk_size], **decode_kwargs).sample
frames.append(frame)
frames = torch.cat(frames, dim=0)
# [batch*frames, channels, height, width] -> [batch, channels, frames, height, width]
frames = frames.reshape(-1, num_frames, *frames.shape[1:]).permute(0, 2, 1, 3, 4)
# we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
frames = frames.float()
return frames
def check_inputs(self, image, height, width):
if (
not isinstance(image, torch.Tensor)
and not isinstance(image, PIL.Image.Image)
and not isinstance(image, list)
):
raise ValueError(
"`image` has to be of type `torch.FloatTensor` or `PIL.Image.Image` or `List[PIL.Image.Image]` but is"
f" {type(image)}"
)
if height % 8 != 0 or width % 8 != 0:
raise ValueError(f"`height` and `width` have to be divisible by 8 but are {height} and {width}.")
def prepare_latents(
self,
batch_size,
num_frames,
num_channels_latents,
height,
width,
dtype,
device,
generator,
latents=None,
):
shape = (
batch_size,
num_frames,
num_channels_latents // 2,
height // self.vae_scale_factor,
width // self.vae_scale_factor,
)
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."
)
if latents is None:
latents = randn_tensor(shape, generator=generator, device=device, dtype=dtype)
else:
latents = latents.to(device)
# scale the initial noise by the standard deviation required by the scheduler
latents = latents * self.scheduler.init_noise_sigma
return latents
@property
def guidance_scale(self):
return self._guidance_scale
# here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
# of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1`
# corresponds to doing no classifier free guidance.
@property
def do_classifier_free_guidance(self):
return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None
@property
def num_timesteps(self):
return self._num_timesteps
@torch.no_grad()
def __call__(
self,
image: Union[PIL.Image.Image, List[PIL.Image.Image], torch.FloatTensor],
height: int = 576,
width: int = 1024,
num_frames: Optional[int] = None,
num_inference_steps: int = 25,
min_guidance_scale: float = 1.0,
max_guidance_scale: float = 3.0,
fps: int = 7,
motion_bucket_id: int = 127,
noise_aug_strength: int = 0.02,
decode_chunk_size: Optional[int] = None,
num_videos_per_prompt: Optional[int] = 1,
generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
latents: Optional[torch.FloatTensor] = None,
output_type: Optional[str] = "pil",
callback_on_step_end: Optional[Callable[[int, int, Dict], None]] = None,
callback_on_step_end_tensor_inputs: List[str] = ["latents"],
cache_interval: Optional[int] = 1,
cache_branch: Optional[int] = None,
return_dict: bool = True,
):
r"""
The call function to the pipeline for generation.
Args:
image (`PIL.Image.Image` or `List[PIL.Image.Image]` or `torch.FloatTensor`):
Image or images to guide image generation. If you provide a tensor, it needs to be compatible with
[`CLIPImageProcessor`](https://huggingface.co/lambdalabs/sd-image-variations-diffusers/blob/main/feature_extractor/preprocessor_config.json).
height (`int`, *optional*, defaults to `self.unet.config.sample_size * self.vae_scale_factor`):
The height in pixels of the generated image.
width (`int`, *optional*, defaults to `self.unet.config.sample_size * self.vae_scale_factor`):
The width in pixels of the generated image.
num_frames (`int`, *optional*):
The number of video frames to generate. Defaults to 14 for `stable-video-diffusion-img2vid` and to 25 for `stable-video-diffusion-img2vid-xt`
num_inference_steps (`int`, *optional*, defaults to 25):
The number of denoising steps. More denoising steps usually lead to a higher quality image at the
expense of slower inference. This parameter is modulated by `strength`.
min_guidance_scale (`float`, *optional*, defaults to 1.0):
The minimum guidance scale. Used for the classifier free guidance with first frame.
max_guidance_scale (`float`, *optional*, defaults to 3.0):
The maximum guidance scale. Used for the classifier free guidance with last frame.
fps (`int`, *optional*, defaults to 7):
Frames per second. The rate at which the generated images shall be exported to a video after generation.
Note that Stable Diffusion Video's UNet was micro-conditioned on fps-1 during training.
motion_bucket_id (`int`, *optional*, defaults to 127):
The motion bucket ID. Used as conditioning for the generation. The higher the number the more motion will be in the video.
noise_aug_strength (`int`, *optional*, defaults to 0.02):
The amount of noise added to the init image, the higher it is the less the video will look like the init image. Increase it for more motion.
decode_chunk_size (`int`, *optional*):
The number of frames to decode at a time. The higher the chunk size, the higher the temporal consistency
between frames, but also the higher the memory consumption. By default, the decoder will decode all frames at once
for maximal quality. Reduce `decode_chunk_size` to reduce memory usage.
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.FloatTensor`, *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`.
output_type (`str`, *optional*, defaults to `"pil"`):
The output format of the generated image. Choose between `PIL.Image` or `np.array`.
callback_on_step_end (`Callable`, *optional*):
A function that calls at the end of each denoising steps during the inference. The function is called
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.
return_dict (`bool`, *optional*, defaults to `True`):
Whether or not to return a [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] instead of a
plain tuple.
Returns:
[`~pipelines.stable_diffusion.StableVideoDiffusionPipelineOutput`] or `tuple`:
If `return_dict` is `True`, [`~pipelines.stable_diffusion.StableVideoDiffusionPipelineOutput`] is returned,
otherwise a `tuple` is returned where the first element is a list of list with the generated frames.
Examples:
```py
from diffusers import StableVideoDiffusionPipeline
from diffusers.utils import load_image, export_to_video
pipe = StableVideoDiffusionPipeline.from_pretrained("stabilityai/stable-video-diffusion-img2vid-xt", torch_dtype=torch.float16, variant="fp16")
pipe.to("cuda")
image = load_image("https://lh3.googleusercontent.com/y-iFOHfLTwkuQSUegpwDdgKmOjRSTvPxat63dQLB25xkTs4lhIbRUFeNBWZzYf370g=s1200")
image = image.resize((1024, 576))
frames = pipe(image, num_frames=25, decode_chunk_size=8).frames[0]
export_to_video(frames, "generated.mp4", fps=7)
```
"""
# 0. Default height and width to unet
height = height or self.unet.config.sample_size * self.vae_scale_factor
width = width or self.unet.config.sample_size * self.vae_scale_factor
num_frames = num_frames if num_frames is not None else self.unet.config.num_frames
decode_chunk_size = decode_chunk_size if decode_chunk_size is not None else num_frames
# 1. Check inputs. Raise error if not correct
self.check_inputs(image, height, width)
# 2. Define call parameters
if isinstance(image, PIL.Image.Image):
batch_size = 1
elif isinstance(image, list):
batch_size = len(image)
else:
batch_size = image.shape[0]
device = self._execution_device
# here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
# of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1`
# corresponds to doing no classifier free guidance.
do_classifier_free_guidance = max_guidance_scale > 1.0
# 3. Encode input image
image_embeddings = self._encode_image(image, device, num_videos_per_prompt, do_classifier_free_guidance)
# NOTE: Stable Diffusion Video was conditioned on fps - 1, which
# is why it is reduced here.
# See: https://github.com/Stability-AI/generative-models/blob/ed0997173f98eaf8f4edf7ba5fe8f15c6b877fd3/scripts/sampling/simple_video_sample.py#L188
fps = fps - 1
# 4. Encode input image using VAE
image = self.image_processor.preprocess(image, height=height, width=width)
noise = randn_tensor(image.shape, generator=generator, device=image.device, dtype=image.dtype)
image = image + noise_aug_strength * noise
needs_upcasting = self.vae.dtype == torch.float16 and self.vae.config.force_upcast
if needs_upcasting:
self.vae.to(dtype=torch.float32)
image_latents = self._encode_vae_image(image, device, num_videos_per_prompt, do_classifier_free_guidance)
image_latents = image_latents.to(image_embeddings.dtype)
# cast back to fp16 if needed
if needs_upcasting:
self.vae.to(dtype=torch.float16)
# Repeat the image latents for each frame so we can concatenate them with the noise
# image_latents [batch, channels, height, width] ->[batch, num_frames, channels, height, width]
image_latents = image_latents.unsqueeze(1).repeat(1, num_frames, 1, 1, 1)
# 5. Get Added Time IDs
added_time_ids = self._get_add_time_ids(
fps,
motion_bucket_id,
noise_aug_strength,
image_embeddings.dtype,
batch_size,
num_videos_per_prompt,
do_classifier_free_guidance,
)
added_time_ids = added_time_ids.to(device)
# 4. Prepare timesteps
self.scheduler.set_timesteps(num_inference_steps, device=device)
timesteps = self.scheduler.timesteps
# 5. Prepare latent variables
num_channels_latents = self.unet.config.in_channels
latents = self.prepare_latents(
batch_size * num_videos_per_prompt,
num_frames,
num_channels_latents,
height,
width,
image_embeddings.dtype,
device,
generator,
latents,
)
# 7. Prepare guidance scale
guidance_scale = torch.linspace(min_guidance_scale, max_guidance_scale, num_frames).unsqueeze(0)
guidance_scale = guidance_scale.to(device, latents.dtype)
guidance_scale = guidance_scale.repeat(batch_size * num_videos_per_prompt, 1)
guidance_scale = _append_dims(guidance_scale, latents.ndim)
self._guidance_scale = guidance_scale
cache_features = None
interval_seq = list(range(0, num_inference_steps, cache_interval))
interval_seq = sorted(interval_seq)
# 8. 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):
# expand the latents if we are doing classifier free guidance
latent_model_input = torch.cat([latents] * 2) if do_classifier_free_guidance else latents
latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)
# Concatenate image_latents over channels dimention
latent_model_input = torch.cat([latent_model_input, image_latents], dim=2)
if i in interval_seq:
cache_features = None
# predict the noise residual
noise_pred, cache_features = self.unet(
latent_model_input,
t,
encoder_hidden_states=image_embeddings,
added_time_ids=added_time_ids,
cache_features=cache_features,
cache_branch=cache_branch,
return_dict=False,
)
# perform guidance
if do_classifier_free_guidance:
noise_pred_uncond, noise_pred_cond = noise_pred.chunk(2)
noise_pred = noise_pred_uncond + self.guidance_scale * (noise_pred_cond - noise_pred_uncond)
# compute the previous noisy sample x_t -> x_t-1
latents = self.scheduler.step(noise_pred, t, latents).prev_sample
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)
if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):
progress_bar.update()
if not output_type == "latent":
# cast back to fp16 if needed
if needs_upcasting:
self.vae.to(dtype=torch.float16)
frames = self.decode_latents(latents, num_frames, decode_chunk_size)
frames = tensor2vid(frames, self.image_processor, output_type=output_type)
else:
frames = latents
self.maybe_free_model_hooks()
if not return_dict:
return frames
return StableVideoDiffusionPipelineOutput(frames=frames)
# resizing utils
# TODO: clean up later
def _resize_with_antialiasing(input, size, interpolation="bicubic", align_corners=True):
h, w = input.shape[-2:]
factors = (h / size[0], w / size[1])
# First, we have to determine sigma
# Taken from skimage: https://github.com/scikit-image/scikit-image/blob/v0.19.2/skimage/transform/_warps.py#L171
sigmas = (
max((factors[0] - 1.0) / 2.0, 0.001),
max((factors[1] - 1.0) / 2.0, 0.001),
)
# Now kernel size. Good results are for 3 sigma, but that is kind of slow. Pillow uses 1 sigma
# https://github.com/python-pillow/Pillow/blob/master/src/libImaging/Resample.c#L206
# But they do it in the 2 passes, which gives better results. Let's try 2 sigmas for now
ks = int(max(2.0 * 2 * sigmas[0], 3)), int(max(2.0 * 2 * sigmas[1], 3))
# Make sure it is odd
if (ks[0] % 2) == 0:
ks = ks[0] + 1, ks[1]
if (ks[1] % 2) == 0:
ks = ks[0], ks[1] + 1
input = _gaussian_blur2d(input, ks, sigmas)
output = torch.nn.functional.interpolate(input, size=size, mode=interpolation, align_corners=align_corners)
return output
def _compute_padding(kernel_size):
"""Compute padding tuple."""
# 4 or 6 ints: (padding_left, padding_right,padding_top,padding_bottom)
# https://pytorch.org/docs/stable/nn.html#torch.nn.functional.pad
if len(kernel_size) < 2:
raise AssertionError(kernel_size)
computed = [k - 1 for k in kernel_size]
# for even kernels we need to do asymmetric padding :(
out_padding = 2 * len(kernel_size) * [0]
for i in range(len(kernel_size)):
computed_tmp = computed[-(i + 1)]
pad_front = computed_tmp // 2
pad_rear = computed_tmp - pad_front
out_padding[2 * i + 0] = pad_front
out_padding[2 * i + 1] = pad_rear
return out_padding
def _filter2d(input, kernel):
# prepare kernel
b, c, h, w = input.shape
tmp_kernel = kernel[:, None, ...].to(device=input.device, dtype=input.dtype)
tmp_kernel = tmp_kernel.expand(-1, c, -1, -1)
height, width = tmp_kernel.shape[-2:]
padding_shape: list[int] = _compute_padding([height, width])
input = torch.nn.functional.pad(input, padding_shape, mode="reflect")
# kernel and input tensor reshape to align element-wise or batch-wise params
tmp_kernel = tmp_kernel.reshape(-1, 1, height, width)
input = input.view(-1, tmp_kernel.size(0), input.size(-2), input.size(-1))
# convolve the tensor with the kernel.
output = torch.nn.functional.conv2d(input, tmp_kernel, groups=tmp_kernel.size(0), padding=0, stride=1)
out = output.view(b, c, h, w)
return out
def _gaussian(window_size: int, sigma):
if isinstance(sigma, float):
sigma = torch.tensor([[sigma]])
batch_size = sigma.shape[0]
x = (torch.arange(window_size, device=sigma.device, dtype=sigma.dtype) - window_size // 2).expand(batch_size, -1)
if window_size % 2 == 0:
x = x + 0.5
gauss = torch.exp(-x.pow(2.0) / (2 * sigma.pow(2.0)))
return gauss / gauss.sum(-1, keepdim=True)
def _gaussian_blur2d(input, kernel_size, sigma):
if isinstance(sigma, tuple):
sigma = torch.tensor([sigma], dtype=input.dtype)
else:
sigma = sigma.to(dtype=input.dtype)
ky, kx = int(kernel_size[0]), int(kernel_size[1])
bs = sigma.shape[0]
kernel_x = _gaussian(kx, sigma[:, 1].view(bs, 1))
kernel_y = _gaussian(ky, sigma[:, 0].view(bs, 1))
out_x = _filter2d(input, kernel_x[..., None, :])
out = _filter2d(out_x, kernel_y[..., None])
return out

2108
ixformer_sdk/contrib/DeepCache/svd/pipeline_utils.py Normal file
View File

File diff suppressed because it is too large Load Diff

2412
ixformer_sdk/contrib/DeepCache/svd/unet_3d_blocks.py Normal file
View File

File diff suppressed because it is too large Load Diff

566
ixformer_sdk/contrib/DeepCache/svd/unet_spatio_temporal_condition.py Normal file
View File

@@ -0,0 +1,566 @@
from dataclasses import dataclass
from typing import Dict, Optional, Tuple, Union
import torch
import torch.nn as nn
from diffusers.configuration_utils import ConfigMixin, register_to_config
from diffusers.loaders import UNet2DConditionLoadersMixin
from diffusers.utils import BaseOutput, logging
from diffusers.models.attention_processor import CROSS_ATTENTION_PROCESSORS, AttentionProcessor, AttnProcessor
from diffusers.models.embeddings import TimestepEmbedding, Timesteps
from diffusers.models.modeling_utils import ModelMixin
from .unet_3d_blocks import UNetMidBlockSpatioTemporal, get_down_block, get_up_block
logger = logging.get_logger(__name__) # pylint: disable=invalid-name
@dataclass
class UNetSpatioTemporalConditionOutput(BaseOutput):
"""
The output of [`UNetSpatioTemporalConditionModel`].
Args:
sample (`torch.FloatTensor` of shape `(batch_size, num_frames, num_channels, height, width)`):
The hidden states output conditioned on `encoder_hidden_states` input. Output of last layer of model.
"""
sample: torch.FloatTensor = None
class UNetSpatioTemporalConditionModel(ModelMixin, ConfigMixin, UNet2DConditionLoadersMixin):
r"""
A conditional Spatio-Temporal UNet model that takes a noisy video frames, conditional state, and a timestep and returns a sample
shaped output.
This model inherits from [`ModelMixin`]. Check the superclass documentation for it's generic methods implemented
for all models (such as downloading or saving).
Parameters:
sample_size (`int` or `Tuple[int, int]`, *optional*, defaults to `None`):
Height and width of input/output sample.
in_channels (`int`, *optional*, defaults to 8): Number of channels in the input sample.
out_channels (`int`, *optional*, defaults to 4): Number of channels in the output.
down_block_types (`Tuple[str]`, *optional*, defaults to `("CrossAttnDownBlockSpatioTemporal", "CrossAttnDownBlockSpatioTemporal", "CrossAttnDownBlockSpatioTemporal", "DownBlockSpatioTemporal")`):
The tuple of downsample blocks to use.
up_block_types (`Tuple[str]`, *optional*, defaults to `("UpBlockSpatioTemporal", "CrossAttnUpBlockSpatioTemporal", "CrossAttnUpBlockSpatioTemporal", "CrossAttnUpBlockSpatioTemporal")`):
The tuple of upsample blocks to use.
block_out_channels (`Tuple[int]`, *optional*, defaults to `(320, 640, 1280, 1280)`):
The tuple of output channels for each block.
addition_time_embed_dim: (`int`, defaults to 256):
Dimension to to encode the additional time ids.
projection_class_embeddings_input_dim (`int`, defaults to 768):
The dimension of the projection of encoded `added_time_ids`.
layers_per_block (`int`, *optional*, defaults to 2): The number of layers per block.
cross_attention_dim (`int` or `Tuple[int]`, *optional*, defaults to 1280):
The dimension of the cross attention features.
transformer_layers_per_block (`int`, `Tuple[int]`, or `Tuple[Tuple]` , *optional*, defaults to 1):
The number of transformer blocks of type [`~models.attention.BasicTransformerBlock`]. Only relevant for
[`~models.unet_3d_blocks.CrossAttnDownBlockSpatioTemporal`], [`~models.unet_3d_blocks.CrossAttnUpBlockSpatioTemporal`],
[`~models.unet_3d_blocks.UNetMidBlockSpatioTemporal`].
num_attention_heads (`int`, `Tuple[int]`, defaults to `(5, 10, 10, 20)`):
The number of attention heads.
dropout (`float`, *optional*, defaults to 0.0): The dropout probability to use.
"""
_supports_gradient_checkpointing = True
@register_to_config
def __init__(
self,
sample_size: Optional[int] = None,
in_channels: int = 8,
out_channels: int = 4,
down_block_types: Tuple[str] = (
"CrossAttnDownBlockSpatioTemporal",
"CrossAttnDownBlockSpatioTemporal",
"CrossAttnDownBlockSpatioTemporal",
"DownBlockSpatioTemporal",
),
up_block_types: Tuple[str] = (
"UpBlockSpatioTemporal",
"CrossAttnUpBlockSpatioTemporal",
"CrossAttnUpBlockSpatioTemporal",
"CrossAttnUpBlockSpatioTemporal",
),
block_out_channels: Tuple[int] = (320, 640, 1280, 1280),
addition_time_embed_dim: int = 256,
projection_class_embeddings_input_dim: int = 768,
layers_per_block: Union[int, Tuple[int]] = 2,
cross_attention_dim: Union[int, Tuple[int]] = 1024,
transformer_layers_per_block: Union[int, Tuple[int], Tuple[Tuple]] = 1,
num_attention_heads: Union[int, Tuple[int]] = (5, 10, 10, 20),
num_frames: int = 25,
):
super().__init__()
self.sample_size = sample_size
# Check inputs
if len(down_block_types) != len(up_block_types):
raise ValueError(
f"Must provide the same number of `down_block_types` as `up_block_types`. `down_block_types`: {down_block_types}. `up_block_types`: {up_block_types}."
)
if len(block_out_channels) != len(down_block_types):
raise ValueError(
f"Must provide the same number of `block_out_channels` as `down_block_types`. `block_out_channels`: {block_out_channels}. `down_block_types`: {down_block_types}."
)
if not isinstance(num_attention_heads, int) and len(num_attention_heads) != len(down_block_types):
raise ValueError(
f"Must provide the same number of `num_attention_heads` as `down_block_types`. `num_attention_heads`: {num_attention_heads}. `down_block_types`: {down_block_types}."
)
if isinstance(cross_attention_dim, list) and len(cross_attention_dim) != len(down_block_types):
raise ValueError(
f"Must provide the same number of `cross_attention_dim` as `down_block_types`. `cross_attention_dim`: {cross_attention_dim}. `down_block_types`: {down_block_types}."
)
if not isinstance(layers_per_block, int) and len(layers_per_block) != len(down_block_types):
raise ValueError(
f"Must provide the same number of `layers_per_block` as `down_block_types`. `layers_per_block`: {layers_per_block}. `down_block_types`: {down_block_types}."
)
# input
self.conv_in = nn.Conv2d(
in_channels,
block_out_channels[0],
kernel_size=3,
padding=1,
)
# time
time_embed_dim = block_out_channels[0] * 4
self.time_proj = Timesteps(block_out_channels[0], True, downscale_freq_shift=0)
timestep_input_dim = block_out_channels[0]
self.time_embedding = TimestepEmbedding(timestep_input_dim, time_embed_dim)
self.add_time_proj = Timesteps(addition_time_embed_dim, True, downscale_freq_shift=0)
self.add_embedding = TimestepEmbedding(projection_class_embeddings_input_dim, time_embed_dim)
self.down_blocks = nn.ModuleList([])
self.up_blocks = nn.ModuleList([])
if isinstance(num_attention_heads, int):
num_attention_heads = (num_attention_heads,) * len(down_block_types)
if isinstance(cross_attention_dim, int):
cross_attention_dim = (cross_attention_dim,) * len(down_block_types)
if isinstance(layers_per_block, int):
layers_per_block = [layers_per_block] * len(down_block_types)
if isinstance(transformer_layers_per_block, int):
transformer_layers_per_block = [transformer_layers_per_block] * len(down_block_types)
blocks_time_embed_dim = time_embed_dim
# down
output_channel = block_out_channels[0]
for i, down_block_type in enumerate(down_block_types):
input_channel = output_channel
output_channel = block_out_channels[i]
is_final_block = i == len(block_out_channels) - 1
down_block = get_down_block(
down_block_type,
num_layers=layers_per_block[i],
transformer_layers_per_block=transformer_layers_per_block[i],
in_channels=input_channel,
out_channels=output_channel,
temb_channels=blocks_time_embed_dim,
add_downsample=not is_final_block,
resnet_eps=1e-5,
cross_attention_dim=cross_attention_dim[i],
num_attention_heads=num_attention_heads[i],
resnet_act_fn="silu",
)
self.down_blocks.append(down_block)
# mid
self.mid_block = UNetMidBlockSpatioTemporal(
block_out_channels[-1],
temb_channels=blocks_time_embed_dim,
transformer_layers_per_block=transformer_layers_per_block[-1],
cross_attention_dim=cross_attention_dim[-1],
num_attention_heads=num_attention_heads[-1],
)
# count how many layers upsample the images
self.num_upsamplers = 0
# up
reversed_block_out_channels = list(reversed(block_out_channels))
reversed_num_attention_heads = list(reversed(num_attention_heads))
reversed_layers_per_block = list(reversed(layers_per_block))
reversed_cross_attention_dim = list(reversed(cross_attention_dim))
reversed_transformer_layers_per_block = list(reversed(transformer_layers_per_block))
output_channel = reversed_block_out_channels[0]
for i, up_block_type in enumerate(up_block_types):
is_final_block = i == len(block_out_channels) - 1
prev_output_channel = output_channel
output_channel = reversed_block_out_channels[i]
input_channel = reversed_block_out_channels[min(i + 1, len(block_out_channels) - 1)]
# add upsample block for all BUT final layer
if not is_final_block:
add_upsample = True
self.num_upsamplers += 1
else:
add_upsample = False
up_block = get_up_block(
up_block_type,
num_layers=reversed_layers_per_block[i] + 1,
transformer_layers_per_block=reversed_transformer_layers_per_block[i],
in_channels=input_channel,
out_channels=output_channel,
prev_output_channel=prev_output_channel,
temb_channels=blocks_time_embed_dim,
add_upsample=add_upsample,
resnet_eps=1e-5,
resolution_idx=i,
cross_attention_dim=reversed_cross_attention_dim[i],
num_attention_heads=reversed_num_attention_heads[i],
resnet_act_fn="silu",
)
self.up_blocks.append(up_block)
prev_output_channel = output_channel
# out
self.conv_norm_out = nn.GroupNorm(num_channels=block_out_channels[0], num_groups=32, eps=1e-5)
self.conv_act = nn.SiLU()
self.conv_out = nn.Conv2d(
block_out_channels[0],
out_channels,
kernel_size=3,
padding=1,
)
@property
def attn_processors(self) -> Dict[str, AttentionProcessor]:
r"""
Returns:
`dict` of attention processors: A dictionary containing all attention processors used in the model with
indexed by its weight name.
"""
# set recursively
processors = {}
def fn_recursive_add_processors(
name: str,
module: torch.nn.Module,
processors: Dict[str, AttentionProcessor],
):
if hasattr(module, "get_processor"):
processors[f"{name}.processor"] = module.get_processor(return_deprecated_lora=True)
for sub_name, child in module.named_children():
fn_recursive_add_processors(f"{name}.{sub_name}", child, processors)
return processors
for name, module in self.named_children():
fn_recursive_add_processors(name, module, processors)
return processors
def set_attn_processor(self, processor: Union[AttentionProcessor, Dict[str, AttentionProcessor]]):
r"""
Sets the attention processor to use to compute attention.
Parameters:
processor (`dict` of `AttentionProcessor` or only `AttentionProcessor`):
The instantiated processor class or a dictionary of processor classes that will be set as the processor
for **all** `Attention` layers.
If `processor` is a dict, the key needs to define the path to the corresponding cross attention
processor. This is strongly recommended when setting trainable attention processors.
"""
count = len(self.attn_processors.keys())
if isinstance(processor, dict) and len(processor) != count:
raise ValueError(
f"A dict of processors was passed, but the number of processors {len(processor)} does not match the"
f" number of attention layers: {count}. Please make sure to pass {count} processor classes."
)
def fn_recursive_attn_processor(name: str, module: torch.nn.Module, processor):
if hasattr(module, "set_processor"):
if not isinstance(processor, dict):
module.set_processor(processor)
else:
module.set_processor(processor.pop(f"{name}.processor"))
for sub_name, child in module.named_children():
fn_recursive_attn_processor(f"{name}.{sub_name}", child, processor)
for name, module in self.named_children():
fn_recursive_attn_processor(name, module, processor)
def set_default_attn_processor(self):
"""
Disables custom attention processors and sets the default attention implementation.
"""
if all(proc.__class__ in CROSS_ATTENTION_PROCESSORS for proc in self.attn_processors.values()):
processor = AttnProcessor()
else:
raise ValueError(
f"Cannot call `set_default_attn_processor` when attention processors are of type {next(iter(self.attn_processors.values()))}"
)
self.set_attn_processor(processor)
def _set_gradient_checkpointing(self, module, value=False):
if hasattr(module, "gradient_checkpointing"):
module.gradient_checkpointing = value
# Copied from diffusers.models.unet_3d_condition.UNet3DConditionModel.enable_forward_chunking
def enable_forward_chunking(self, chunk_size: Optional[int] = None, dim: int = 0) -> None:
"""
Sets the attention processor to use [feed forward
chunking](https://huggingface.co/blog/reformer#2-chunked-feed-forward-layers).
Parameters:
chunk_size (`int`, *optional*):
The chunk size of the feed-forward layers. If not specified, will run feed-forward layer individually
over each tensor of dim=`dim`.
dim (`int`, *optional*, defaults to `0`):
The dimension over which the feed-forward computation should be chunked. Choose between dim=0 (batch)
or dim=1 (sequence length).
"""
if dim not in [0, 1]:
raise ValueError(f"Make sure to set `dim` to either 0 or 1, not {dim}")
# By default chunk size is 1
chunk_size = chunk_size or 1
def fn_recursive_feed_forward(module: torch.nn.Module, chunk_size: int, dim: int):
if hasattr(module, "set_chunk_feed_forward"):
module.set_chunk_feed_forward(chunk_size=chunk_size, dim=dim)
for child in module.children():
fn_recursive_feed_forward(child, chunk_size, dim)
for module in self.children():
fn_recursive_feed_forward(module, chunk_size, dim)
def forward(
self,
sample: torch.FloatTensor,
timestep: Union[torch.Tensor, float, int],
encoder_hidden_states: torch.Tensor,
added_time_ids: torch.Tensor,
cache_features: Optional[torch.Tensor] = None,
cache_branch: Optional[int] = None,
return_dict: bool = True,
) -> Union[UNetSpatioTemporalConditionOutput, Tuple]:
r"""
The [`UNetSpatioTemporalConditionModel`] forward method.
Args:
sample (`torch.FloatTensor`):
The noisy input tensor with the following shape `(batch, num_frames, channel, height, width)`.
timestep (`torch.FloatTensor` or `float` or `int`): The number of timesteps to denoise an input.
encoder_hidden_states (`torch.FloatTensor`):
The encoder hidden states with shape `(batch, sequence_length, cross_attention_dim)`.
added_time_ids: (`torch.FloatTensor`):
The additional time ids with shape `(batch, num_additional_ids)`. These are encoded with sinusoidal
embeddings and added to the time embeddings.
return_dict (`bool`, *optional*, defaults to `True`):
Whether or not to return a [`~models.unet_slatio_temporal.UNetSpatioTemporalConditionOutput`] instead of a plain
tuple.
Returns:
[`~models.unet_slatio_temporal.UNetSpatioTemporalConditionOutput`] or `tuple`:
If `return_dict` is True, an [`~models.unet_slatio_temporal.UNetSpatioTemporalConditionOutput`] is returned, otherwise
a `tuple` is returned where the first element is the sample tensor.
"""
# 1. time
timesteps = timestep
if not torch.is_tensor(timesteps):
# TODO: this requires sync between CPU and GPU. So try to pass timesteps as tensors if you can
# This would be a good case for the `match` statement (Python 3.10+)
is_mps = sample.device.type == "mps"
if isinstance(timestep, float):
dtype = torch.float32 if is_mps else torch.float64
else:
dtype = torch.int32 if is_mps else torch.int64
timesteps = torch.tensor([timesteps], dtype=dtype, device=sample.device)
elif len(timesteps.shape) == 0:
timesteps = timesteps[None].to(sample.device)
# broadcast to batch dimension in a way that's compatible with ONNX/Core ML
batch_size, num_frames = sample.shape[:2]
timesteps = timesteps.expand(batch_size)
t_emb = self.time_proj(timesteps)
# `Timesteps` does not contain any weights and will always return f32 tensors
# but time_embedding might actually be running in fp16. so we need to cast here.
# there might be better ways to encapsulate this.
t_emb = t_emb.to(dtype=sample.dtype)
emb = self.time_embedding(t_emb)
time_embeds = self.add_time_proj(added_time_ids.flatten())
time_embeds = time_embeds.reshape((batch_size, -1))
time_embeds = time_embeds.to(emb.dtype)
aug_emb = self.add_embedding(time_embeds)
emb = emb + aug_emb
# Flatten the batch and frames dimensions
# sample: [batch, frames, channels, height, width] -> [batch * frames, channels, height, width]
sample = sample.flatten(0, 1)
# Repeat the embeddings num_video_frames times
# emb: [batch, channels] -> [batch * frames, channels]
emb = emb.repeat_interleave(num_frames, dim=0)
# encoder_hidden_states: [batch, 1, channels] -> [batch * frames, 1, channels]
encoder_hidden_states = encoder_hidden_states.repeat_interleave(num_frames, dim=0)
# 2. pre-process
sample = self.conv_in(sample)
image_only_indicator = torch.zeros(batch_size, num_frames, dtype=sample.dtype, device=sample.device)
# Branch: 4 down_blocks, each with 3 skip connections. Here we ignore the first skip branch, whose computations only has up_blocks but without down_blocks.
if cache_branch is not None:
each_module_num = len(self.down_blocks[0].resnets) + 1
down_cache_block_idx = cache_branch // each_module_num
down_cache_module_idx = cache_branch % each_module_num
up_cache_block_idx = len(self.up_blocks) - 1 - down_cache_block_idx
up_cache_module_idx = 1 - down_cache_module_idx
if down_cache_module_idx == each_module_num - 1:
up_cache_block_idx -= 1
up_cache_module_idx = 2
if cache_features is not None:
# 3. down
down_block_res_samples = (sample,)
for block_id, downsample_block in enumerate(self.down_blocks):
if hasattr(downsample_block, "has_cross_attention") and downsample_block.has_cross_attention:
sample, res_samples = downsample_block(
hidden_states=sample,
temb=emb,
encoder_hidden_states=encoder_hidden_states,
image_only_indicator=image_only_indicator,
exist_module_idx=down_cache_module_idx if down_cache_block_idx == block_id else None
)
else:
sample, res_samples = downsample_block(
hidden_states=sample,
temb=emb,
image_only_indicator=image_only_indicator,
exist_module_idx=down_cache_module_idx if down_cache_block_idx == block_id else None
)
down_block_res_samples += res_samples
if down_cache_block_idx == block_id:
break
# 4. no mid
sample = cache_features
# 5. up
for i, upsample_block in enumerate(self.up_blocks):
if i < up_cache_block_idx:
continue
if i == up_cache_block_idx:
trunc_res_samples_len = len(upsample_block.resnets) - up_cache_module_idx
else:
trunc_res_samples_len = len(upsample_block.resnets)
res_samples = down_block_res_samples[-trunc_res_samples_len :]
down_block_res_samples = down_block_res_samples[: -trunc_res_samples_len]
if hasattr(upsample_block, "has_cross_attention") and upsample_block.has_cross_attention:
sample, _ = upsample_block(
hidden_states=sample,
temb=emb,
res_hidden_states_tuple=res_samples,
encoder_hidden_states=encoder_hidden_states,
image_only_indicator=image_only_indicator,
enter_module_idx=up_cache_module_idx if i == up_cache_block_idx else None
)
else:
sample, _ = upsample_block(
hidden_states=sample,
temb=emb,
res_hidden_states_tuple=res_samples,
image_only_indicator=image_only_indicator,
enter_module_idx=up_cache_module_idx if i == up_cache_block_idx else None
)
else:
# 3. down
down_block_res_samples = (sample,)
for downsample_block in self.down_blocks:
if hasattr(downsample_block, "has_cross_attention") and downsample_block.has_cross_attention:
sample, res_samples = downsample_block(
hidden_states=sample,
temb=emb,
encoder_hidden_states=encoder_hidden_states,
image_only_indicator=image_only_indicator,
)
else:
sample, res_samples = downsample_block(
hidden_states=sample,
temb=emb,
image_only_indicator=image_only_indicator,
)
down_block_res_samples += res_samples
# 4. mid
sample = self.mid_block(
hidden_states=sample,
temb=emb,
encoder_hidden_states=encoder_hidden_states,
image_only_indicator=image_only_indicator,
)
# 5. up
for i, upsample_block in enumerate(self.up_blocks):
res_samples = down_block_res_samples[-len(upsample_block.resnets) :]
down_block_res_samples = down_block_res_samples[: -len(upsample_block.resnets)]
if hasattr(upsample_block, "has_cross_attention") and upsample_block.has_cross_attention:
sample, current_record_f = upsample_block(
hidden_states=sample,
temb=emb,
res_hidden_states_tuple=res_samples,
encoder_hidden_states=encoder_hidden_states,
image_only_indicator=image_only_indicator,
)
else:
sample, current_record_f = upsample_block(
hidden_states=sample,
temb=emb,
res_hidden_states_tuple=res_samples,
image_only_indicator=image_only_indicator,
)
if cache_branch is not None and i == up_cache_block_idx:
cache_features = current_record_f[up_cache_module_idx]
# 6. post-process
sample = self.conv_norm_out(sample)
sample = self.conv_act(sample)
sample = self.conv_out(sample)
# 7. Reshape back to original shape
sample = sample.reshape(batch_size, num_frames, *sample.shape[1:])
if not return_dict:
return (sample, cache_features)
return UNetSpatioTemporalConditionOutput(sample=sample)