add codes

README.md

@@ -1,10 +1,10 @@
 ---
 title: InstanceShadow
-emoji: 💻
-colorFrom: green
+emoji: 🔥
+colorFrom: red
 colorTo: gray
 sdk: gradio
-sdk_version: 4.7.1
+sdk_version: 3.50.2
 app_file: app.py
 pinned: false
 license: cc-by-4.0

app.py

@@ -0,0 +1,92 @@
+import gradio as gr
+
+from PIL import Image
+
+import torch
+from diffusion import DiffusionPipeline
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+pipe = DiffusionPipeline(device)
+
+def read_content(file_path: str) -> str:
+    """read the content of target file
+    """
+    with open(file_path, 'r', encoding='utf-8') as f:
+        content = f.read()
+
+    return content
+
+def predict(input, dkernel, diffusion_step, q=False):
+    lq = input["image"].convert("RGB")
+    mask = input["mask"].convert("RGB")
+    mask = mask.resize(lq.size, resample=Image.NEAREST)
+    output = pipe(lq=lq, mask=mask, dkernel=dkernel, diffusion_step=diffusion_step)
+    return output
+
+def qpredict(input, dkernel, diffusion_step, q=False):
+    lq = input["image"].convert("RGB")
+    mask = input["mask"].convert("RGB")
+    mask = mask.resize(lq.size, resample=Image.NEAREST)
+    for output in pipe.quick_solve(lq=lq, mask=mask, dkernel=dkernel, diffusion_step=diffusion_step):
+        yield output
+
+
+css = '''
+.container {max-width: 1150px;margin: auto;padding-top: 1.5rem}
+#image_upload{min-height:400px}
+#image_upload [data-testid="image"], #image_upload [data-testid="image"] > div{min-height: 400px}
+#mask_radio .gr-form{background:transparent; border: none}
+#word_mask{margin-top: .75em !important}
+#word_mask textarea:disabled{opacity: 0.3}
+.footer {margin-bottom: 45px;margin-top: 35px;text-align: center;border-bottom: 1px solid #e5e5e5}
+.footer>p {font-size: .8rem; display: inline-block; padding: 0 10px;transform: translateY(10px);background: white}
+.dark .footer {border-color: #303030}
+.dark .footer>p {background: #0b0f19}
+.acknowledgments h4{margin: 1.25em 0 .25em 0;font-weight: bold;font-size: 115%}
+#image_upload .touch-none{display: flex}
+@keyframes spin {
+    from {
+        transform: rotate(0deg);
+    }
+    to {
+        transform: rotate(360deg);
+    }
+}
+'''
+
+image_blocks = gr.Blocks(css=css)
+with image_blocks as demo:
+    gr.HTML(read_content("header.html"))
+    with gr.Group():
+        with gr.Box():
+            with gr.Row():
+                with gr.Column():
+                    image = gr.Image(source='upload', tool='sketch', elem_id="image_upload", type="pil", label="Shadow Image").style(height=400)
+                    dkernel = gr.Slider(minimum=11, maximum=55, step=2, value=25, label="Dilation Kernel Size")
+                    diffusion_step = gr.Slider(minimum=10, maximum=200, step=5, value=50, label="Diffusion Time Step")
+                    with gr.Row(elem_id="prompt-container").style(mobile_collapse=False, equal_height=True):
+                        with gr.Column():
+                            btn = gr.Button("Removal").style(
+                                margin=False,
+                                full_width=True,
+                            )
+                        with gr.Column():
+                            qbtn = gr.Button("Quick Removal").style(
+                                margin=False,
+                                full_width=True,
+                            )
+
+                with gr.Column():
+                    image_out = gr.Image(label="Removal Result", elem_id="output-img").style(height=400)
+            with gr.Row():
+                gr.Examples(examples=[
+                    'examples/web-shadow0243.jpg',
+                    'examples/web-shadow0248.jpg',
+                    'examples/lssd2025.jpg'
+                ], inputs=[image])
+
+            btn.click(fn=predict, inputs=[image, dkernel, diffusion_step], outputs=[image_out])
+            qbtn.click(fn=qpredict, inputs=[image, dkernel, diffusion_step], outputs=[image_out])
+
+image_blocks.launch(enable_queue=True, share=False, debug=False, server_port=10011)

diffusion.py

@@ -0,0 +1,185 @@
+import inspect
+from tkinter import Image
+from typing import List, Optional, Union
+
+import numpy as np
+import torch
+
+import PIL
+from PIL import Image
+from tqdm.auto import tqdm
+
+from diffusion_arch import DensePosteriorConditionalUNet
+from guided_diffusion.script_util import create_gaussian_diffusion
+
+import torch.nn.functional as F
+import torchvision.transforms.functional as TF
+
+from einops import rearrange
+from kornia.morphology import dilation
+
+from tqdm import tqdm
+
+def preprocess_image(image):
+    w, h = image.size
+    w, h = map(lambda x: x - x % 32, (w, h))  # resize to integer multiple of 32
+    image = image.resize((w, h), resample=PIL.Image.LANCZOS)
+    image = np.array(image).astype(np.float32) / 255.0
+    image = torch.from_numpy(image.transpose(2,0,1)).unsqueeze(0)
+    return 2.0 * image - 1.0
+
+def preprocess_mask(mask):
+    mask = mask.convert("L")
+    w, h = mask.size
+    w, h = map(lambda x: x - x % 32, (w, h))  # resize to integer multiple of 32
+    mask = mask.resize((w, h), resample=PIL.Image.NEAREST)
+    mask = np.array(mask).astype(np.float32) / 255.0
+    mask = torch.from_numpy(np.repeat(mask[None, ...], 3, axis=0)).unsqueeze(0)
+    mask[mask > 0] = 1
+    return mask
+
+
+class DiffusionPipeline():
+    def __init__(self, device):
+        super().__init__()
+        self.device = device
+        self.model = DensePosteriorConditionalUNet(
+            in_channels=9,
+            model_channels=256,
+            out_channels=6,
+            num_res_blocks=2,
+            attention_resolutions=[8, 16, 32],
+            dropout=0.0,
+            channel_mult=(1, 1, 2, 2, 4, 4),
+            num_classes=None,
+            use_checkpoint=False,
+            use_fp16=False,
+            num_heads=4,
+            num_head_channels=64,
+            num_heads_upsample=-1,
+            use_scale_shift_norm=True,
+            resblock_updown=True,
+            use_new_attention_order=True
+        )
+        self.model.eval()
+        self.model.to(self.device)
+        self.model.load_state_dict(torch.load('net_g_400000.pth', map_location='cpu')["params_ema"], strict=True)
+
+
+    @torch.no_grad()
+    def __call__(self, lq, mask, dkernel, diffusion_step):
+        self.eval_gaussian_diffusion = create_gaussian_diffusion(
+            steps=1000,
+            learn_sigma=True,
+            noise_schedule='linear',
+            use_kl=False,
+            timestep_respacing="ddim" + str(diffusion_step),
+            predict_xstart=False,
+            rescale_timesteps=False,
+            rescale_learned_sigmas=False,
+            p2_gamma=1,
+            p2_k=1,
+        )
+
+        ow, oh = lq.size
+
+        # preprocess image
+        lq = preprocess_image(lq).to(self.device)
+
+        # preprocess mask
+        mask = preprocess_mask(mask).to(self.device)
+        mask = dilation(mask, torch.ones(dkernel, dkernel, device=self.device))
+
+        # return Image.fromarray(np.uint8(torch.cat(((lq / 2 + 0.5).clamp(0, 1), mask), dim=2).cpu().permute(0, 2, 3, 1).numpy()[0] * 255.))
+
+        #======== PADDING FORWARDING ============
+        stride = 64
+        kernel_size = 256
+
+        _, _, h, w = mask.shape
+        mask = F.unfold(mask, kernel_size=kernel_size, stride=stride)
+        lq = F.unfold(lq, kernel_size=kernel_size, stride=stride)
+
+        n, c, l = mask.shape
+        mask = rearrange(mask, 'n (c3 h w) l -> (n l) c3 h w', h=kernel_size, w=kernel_size)
+        lq = rearrange(lq, 'n (c3 h w) l -> (n l) c3 h w', h=kernel_size, w=kernel_size)
+
+        #======== PADDING END ============
+
+        #======== FORWARDING ============
+        sub_imgs = []
+        for (sub_lq, sub_mask) in zip(lq.unsqueeze(1), mask.unsqueeze(1)):
+            if torch.sum(sub_mask) > 1:
+                img = torch.randn_like(sub_lq, device=self.device)
+                indices = list(range(self.eval_gaussian_diffusion.num_timesteps))[::-1]
+                for i in indices:
+                    t = torch.tensor([i] * img.size(0), device=self.device)
+                    img = img * sub_mask + self.eval_gaussian_diffusion.q_sample(sub_lq, t) * (1 - sub_mask)
+                    out = self.eval_gaussian_diffusion.p_mean_variance(self.model, img.contiguous(), t, model_kwargs={'latent': torch.cat((sub_lq, sub_mask), dim=1)})
+                    nonzero_mask = (
+                        (t != 0).float().view(-1, *([1] * (len(img.shape) - 1)))
+                    )  # no noise when t == 0
+                    img = out["mean"] + nonzero_mask * torch.exp(0.5 * out["log_variance"]) * torch.randn_like(img, device=self.device)
+                sub_imgs.append(img)
+            else:
+                sub_imgs.append(sub_lq)
+        img = torch.cat(sub_imgs, dim=0)
+
+        #======== PADDING BACKWARDING ============
+        img = rearrange(img, '(n l) c3 h w -> n (c3 h w) l', h=kernel_size, w=kernel_size, l=l)
+        img = F.fold(img, (h, w), kernel_size=kernel_size, stride=stride)
+        norm_map = F.fold(F.unfold(torch.ones_like(img), kernel_size, stride=stride), (h, w), kernel_size, stride=stride)
+        img /= norm_map
+
+        img = (img / 2 + 0.5).clamp(0, 1)
+        img = img.cpu().permute(0, 2, 3, 1).numpy()[0]
+        img = Image.fromarray(np.uint8(img * 255.))
+        img = img.resize((ow, oh), resample=PIL.Image.LANCZOS)
+
+        return img
+
+
+
+    @torch.no_grad()
+    def quick_solve(self, lq, mask, dkernel, diffusion_step):
+        self.eval_gaussian_diffusion = create_gaussian_diffusion(
+            steps=1000,
+            learn_sigma=True,
+            noise_schedule='linear',
+            use_kl=False,
+            timestep_respacing="ddim" + str(diffusion_step),
+            predict_xstart=False,
+            rescale_timesteps=False,
+            rescale_learned_sigmas=False,
+            p2_gamma=1,
+            p2_k=1,
+        )
+
+        ow, oh = lq.size
+
+        lq = lq.resize((512, 512), resample=Image.LANCZOS)
+        mask = mask.resize((512, 512), resample=Image.NEAREST)
+
+        # preprocess image
+        lq = preprocess_image(lq).to(self.device)
+
+        # preprocess mask
+        mask = preprocess_mask(mask).to(self.device)
+        mask = dilation(mask, torch.ones(dkernel, dkernel, device=self.device))
+
+        # return Image.fromarray(np.uint8(torch.cat(((lq / 2 + 0.5).clamp(0, 1), mask), dim=2).cpu().permute(0, 2, 3, 1).numpy()[0] * 255.))
+
+        img = torch.randn_like(lq, device=self.device)
+        indices = list(range(self.eval_gaussian_diffusion.num_timesteps))[::-1]
+        for i in indices:
+            t = torch.tensor([i] * img.size(0), device=self.device)
+            img = img * mask + self.eval_gaussian_diffusion.q_sample(lq, t) * (1 - mask)
+            out = self.eval_gaussian_diffusion.p_mean_variance(self.model, img.contiguous(), t, model_kwargs={'latent': torch.cat((lq, mask), dim=1)})
+            nonzero_mask = (
+                (t != 0).float().view(-1, *([1] * (len(img.shape) - 1)))
+            )  # no noise when t == 0
+            img = out["mean"] + nonzero_mask * torch.exp(0.5 * out["log_variance"]) * torch.randn_like(img, device=self.device)
+
+            yield Image.fromarray(np.uint8((out["pred_xstart"] / 2 + 0.5).clamp(0, 1).cpu().permute(0, 2, 3, 1).numpy()[0] * 255.)).resize((ow, oh), resample=Image.LANCZOS)
+
+        yield Image.fromarray(np.uint8((img / 2 + 0.5).clamp(0, 1).cpu().permute(0, 2, 3, 1).numpy()[0] * 255.)).resize((ow, oh), resample=Image.LANCZOS)

diffusion_arch.py

@@ -0,0 +1,1208 @@
+from abc import abstractmethod
+
+import math
+import numpy as np
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+import torchvision
+import torchvision.transforms.functional as TF
+
+import numpy as np
+
+from guided_diffusion.fp16_util import convert_module_to_f16, convert_module_to_f32
+from guided_diffusion.nn import (
+    checkpoint,
+    conv_nd,
+    linear,
+    avg_pool_nd,
+    zero_module,
+    normalization,
+    timestep_embedding,
+)
+
+class AttentionPool2d(nn.Module):
+    """
+    Adapted from CLIP: https://github.com/openai/CLIP/blob/main/clip/model.py
+    """
+
+    def __init__(
+        self,
+        spacial_dim: int,
+        embed_dim: int,
+        num_heads_channels: int,
+        output_dim: int = None,
+    ):
+        super().__init__()
+        self.positional_embedding = nn.Parameter(
+            torch.randn(embed_dim, spacial_dim ** 2 + 1) / embed_dim ** 0.5
+        )
+        self.qkv_proj = conv_nd(1, embed_dim, 3 * embed_dim, 1)
+        self.c_proj = conv_nd(1, embed_dim, output_dim or embed_dim, 1)
+        self.num_heads = embed_dim // num_heads_channels
+        self.attention = QKVAttention(self.num_heads)
+
+    def forward(self, x):
+        b, c, *_spatial = x.shape
+        x = x.reshape(b, c, -1)  # NC(HW)
+        x = torch.cat([x.mean(dim=-1, keepdim=True), x], dim=-1)  # NC(HW+1)
+        x = x + self.positional_embedding[None, :, :].to(x.dtype)  # NC(HW+1)
+        x = self.qkv_proj(x)
+        x = self.attention(x)
+        x = self.c_proj(x)
+        return x[:, :, 0]
+
+
+class TimestepBlock(nn.Module):
+    """
+    Any module where forward() takes timestep embeddings as a second argument.
+    """
+
+    @abstractmethod
+    def forward(self, x, emb):
+        """
+        Apply the module to `x` given `emb` timestep embeddings.
+        """
+
+
+class TimestepEmbedSequential(nn.Sequential, TimestepBlock):
+    """
+    A sequential module that passes timestep embeddings to the children that
+    support it as an extra input.
+    """
+
+    def forward(self, x, emb):
+        for layer in self:
+            if isinstance(layer, TimestepBlock):
+                x = layer(x, emb)
+            else:
+                x = layer(x)
+        return x
+
+class Upsample(nn.Module):
+    """
+    An upsampling layer with an optional convolution.
+
+    :param channels: channels in the inputs and outputs.
+    :param use_conv: a bool determining if a convolution is applied.
+    :param dims: determines if the signal is 1D, 2D, or 3D. If 3D, then
+                 upsampling occurs in the inner-two dimensions.
+    """
+
+    def __init__(self, channels, use_conv, dims=2, out_channels=None):
+        super().__init__()
+        self.channels = channels
+        self.out_channels = out_channels or channels
+        self.use_conv = use_conv
+        self.dims = dims
+        if use_conv:
+            self.conv = conv_nd(dims, self.channels, self.out_channels, 3, padding=1)
+
+    def forward(self, x):
+        assert x.shape[1] == self.channels
+        if self.dims == 3:
+            x = F.interpolate(
+                x, (x.shape[2], x.shape[3] * 2, x.shape[4] * 2), mode="nearest"
+            )
+        else:
+            x = F.interpolate(x, scale_factor=2, mode="nearest")
+        if self.use_conv:
+            x = self.conv(x)
+        return x
+
+
+class Downsample(nn.Module):
+    """
+    A downsampling layer with an optional convolution.
+
+    :param channels: channels in the inputs and outputs.
+    :param use_conv: a bool determining if a convolution is applied.
+    :param dims: determines if the signal is 1D, 2D, or 3D. If 3D, then
+                 downsampling occurs in the inner-two dimensions.
+    """
+
+    def __init__(self, channels, use_conv, dims=2, out_channels=None):
+        super().__init__()
+        self.channels = channels
+        self.out_channels = out_channels or channels
+        self.use_conv = use_conv
+        self.dims = dims
+        stride = 2 if dims != 3 else (1, 2, 2)
+        if use_conv:
+            self.op = conv_nd(
+                dims, self.channels, self.out_channels, 3, stride=stride, padding=1
+            )
+        else:
+            assert self.channels == self.out_channels
+            self.op = avg_pool_nd(dims, kernel_size=stride, stride=stride)
+
+    def forward(self, x):
+        assert x.shape[1] == self.channels
+        return self.op(x)
+
+
+class ResBlock(TimestepBlock):
+    """
+    A residual block that can optionally change the number of channels.
+
+    :param channels: the number of input channels.
+    :param emb_channels: the number of timestep embedding channels.
+    :param dropout: the rate of dropout.
+    :param out_channels: if specified, the number of out channels.
+    :param use_conv: if True and out_channels is specified, use a spatial
+        convolution instead of a smaller 1x1 convolution to change the
+        channels in the skip connection.
+    :param dims: determines if the signal is 1D, 2D, or 3D.
+    :param use_checkpoint: if True, use gradient checkpointing on this module.
+    :param up: if True, use this block for upsampling.
+    :param down: if True, use this block for downsampling.
+    """
+
+    def __init__(
+        self,
+        channels,
+        emb_channels,
+        dropout,
+        out_channels=None,
+        use_conv=False,
+        use_scale_shift_norm=False,
+        dims=2,
+        use_checkpoint=False,
+        up=False,
+        down=False,
+    ):
+        super().__init__()
+        self.channels = channels
+        self.emb_channels = emb_channels
+        self.dropout = dropout
+        self.out_channels = out_channels or channels
+        self.use_conv = use_conv
+        self.use_checkpoint = use_checkpoint
+        self.use_scale_shift_norm = use_scale_shift_norm
+
+        self.in_layers = nn.Sequential(
+            normalization(channels),
+            nn.SiLU(),
+            conv_nd(dims, channels, self.out_channels, 3, padding=1),
+        )
+
+        self.updown = up or down
+
+        if up:
+            self.h_upd = Upsample(channels, False, dims)
+            self.x_upd = Upsample(channels, False, dims)
+        elif down:
+            self.h_upd = Downsample(channels, False, dims)
+            self.x_upd = Downsample(channels, False, dims)
+        else:
+            self.h_upd = self.x_upd = nn.Identity()
+
+        self.emb_layers = nn.Sequential(
+            nn.SiLU(),
+            linear(
+                emb_channels,
+                2 * self.out_channels if use_scale_shift_norm else self.out_channels,
+            ),
+        )
+        self.out_layers = nn.Sequential(
+            normalization(self.out_channels),
+            nn.SiLU(),
+            nn.Dropout(p=dropout),
+            zero_module(
+                conv_nd(dims, self.out_channels, self.out_channels, 3, padding=1)
+            ),
+        )
+
+        if self.out_channels == channels:
+            self.skip_connection = nn.Identity()
+        elif use_conv:
+            self.skip_connection = conv_nd(
+                dims, channels, self.out_channels, 3, padding=1
+            )
+        else:
+            self.skip_connection = conv_nd(dims, channels, self.out_channels, 1)
+
+    def forward(self, x, emb):
+        """
+        Apply the block to a Tensor, conditioned on a timestep embedding.
+
+        :param x: an [N x C x ...] Tensor of features.
+        :param emb: an [N x emb_channels] Tensor of timestep embeddings.
+        :return: an [N x C x ...] Tensor of outputs.
+        """
+        return checkpoint(
+            self._forward, (x, emb), self.parameters(), self.use_checkpoint
+        )
+
+    def _forward(self, x, emb):
+        if self.updown:
+            in_rest, in_conv = self.in_layers[:-1], self.in_layers[-1]
+            h = in_rest(x)
+            h = self.h_upd(h)
+            x = self.x_upd(x)
+            h = in_conv(h)
+        else:
+            h = self.in_layers(x)
+        emb_out = self.emb_layers(emb).type(h.dtype)
+        while len(emb_out.shape) < len(h.shape):
+            emb_out = emb_out[..., None]
+        if self.use_scale_shift_norm:
+            out_norm, out_rest = self.out_layers[0], self.out_layers[1:]
+            scale, shift = torch.chunk(emb_out, 2, dim=1)
+            h = out_norm(h) * (1 + scale) + shift
+            h = out_rest(h)
+        else:
+            h = h + emb_out
+            h = self.out_layers(h)
+        return self.skip_connection(x) + h
+
+
+class AttentionBlock(nn.Module):
+    """
+    An attention block that allows spatial positions to attend to each other.
+
+    Originally ported from here, but adapted to the N-d case.
+    https://github.com/hojonathanho/diffusion/blob/1e0dceb3b3495bbe19116a5e1b3596cd0706c543/diffusion_tf/models/unet.py#L66.
+    """
+
+    def __init__(
+        self,
+        channels,
+        num_heads=1,
+        num_head_channels=-1,
+        use_checkpoint=False,
+        use_new_attention_order=False,
+    ):
+        super().__init__()
+        self.channels = channels
+        if num_head_channels == -1:
+            self.num_heads = num_heads
+        else:
+            assert (
+                channels % num_head_channels == 0
+            ), f"q,k,v channels {channels} is not divisible by num_head_channels {num_head_channels}"
+            self.num_heads = channels // num_head_channels
+        self.use_checkpoint = use_checkpoint
+        self.norm = normalization(channels)
+        self.qkv = conv_nd(1, channels, channels * 3, 1)
+        if use_new_attention_order:
+            # split qkv before split heads
+            self.attention = QKVAttention(self.num_heads)
+        else:
+            # split heads before split qkv
+            self.attention = QKVAttentionLegacy(self.num_heads)
+
+        self.proj_out = zero_module(conv_nd(1, channels, channels, 1))
+
+    def forward(self, x):
+        return checkpoint(self._forward, (x,), self.parameters(), True)
+
+    def _forward(self, x):
+        b, c, *spatial = x.shape
+        x = x.reshape(b, c, -1)
+        qkv = self.qkv(self.norm(x))
+        h = self.attention(qkv)
+        h = self.proj_out(h)
+        return (x + h).reshape(b, c, *spatial)
+
+
+def count_flops_attn(model, _x, y):
+    """
+    A counter for the `thop` package to count the operations in an
+    attention operation.
+    Meant to be used like:
+        macs, params = thop.profile(
+            model,
+            inputs=(inputs, timestamps),
+            custom_ops={QKVAttention: QKVAttention.count_flops},
+        )
+    """
+    b, c, *spatial = y[0].shape
+    num_spatial = int(np.prod(spatial))
+    # We perform two matmuls with the same number of ops.
+    # The first computes the weight matrix, the second computes
+    # the combination of the value vectors.
+    matmul_ops = 2 * b * (num_spatial ** 2) * c
+    model.total_ops += torch.DoubleTensor([matmul_ops])
+
+
+class QKVAttentionLegacy(nn.Module):
+    """
+    A module which performs QKV attention. Matches legacy QKVAttention + input/ouput heads shaping
+    """
+
+    def __init__(self, n_heads):
+        super().__init__()
+        self.n_heads = n_heads
+
+    def forward(self, qkv):
+        """
+        Apply QKV attention.
+
+        :param qkv: an [N x (H * 3 * C) x T] tensor of Qs, Ks, and Vs.
+        :return: an [N x (H * C) x T] tensor after attention.
+        """
+        bs, width, length = qkv.shape
+        assert width % (3 * self.n_heads) == 0
+        ch = width // (3 * self.n_heads)
+        q, k, v = qkv.reshape(bs * self.n_heads, ch * 3, length).split(ch, dim=1)
+        scale = 1 / math.sqrt(math.sqrt(ch))
+        weight = torch.einsum(
+            "bct,bcs->bts", q * scale, k * scale
+        )  # More stable with f16 than dividing afterwards
+        weight = torch.softmax(weight.float(), dim=-1).type(weight.dtype)
+        a = torch.einsum("bts,bcs->bct", weight, v)
+        return a.reshape(bs, -1, length)
+
+    @staticmethod
+    def count_flops(model, _x, y):
+        return count_flops_attn(model, _x, y)
+
+
+class QKVAttention(nn.Module):
+    """
+    A module which performs QKV attention and splits in a different order.
+    """
+
+    def __init__(self, n_heads):
+        super().__init__()
+        self.n_heads = n_heads
+
+    def forward(self, qkv):
+        """
+        Apply QKV attention.
+
+        :param qkv: an [N x (3 * H * C) x T] tensor of Qs, Ks, and Vs.
+        :return: an [N x (H * C) x T] tensor after attention.
+        """
+        bs, width, length = qkv.shape
+        assert width % (3 * self.n_heads) == 0
+        ch = width // (3 * self.n_heads)
+        q, k, v = qkv.chunk(3, dim=1)
+        scale = 1 / math.sqrt(math.sqrt(ch))
+        weight = torch.einsum(
+            "bct,bcs->bts",
+            (q * scale).view(bs * self.n_heads, ch, length),
+            (k * scale).view(bs * self.n_heads, ch, length),
+        )  # More stable with f16 than dividing afterwards
+        weight = torch.softmax(weight.float(), dim=-1).type(weight.dtype)
+        a = torch.einsum("bts,bcs->bct", weight, v.reshape(bs * self.n_heads, ch, length))
+        return a.reshape(bs, -1, length)
+
+    @staticmethod
+    def count_flops(model, _x, y):
+        return count_flops_attn(model, _x, y)
+
+class UNetModel(nn.Module):
+    """
+    The full UNet model with attention and timestep embedding.
+
+    :param in_channels: channels in the input Tensor.
+    :param model_channels: base channel count for the model.
+    :param out_channels: channels in the output Tensor.
+    :param num_res_blocks: number of residual blocks per downsample.
+    :param attention_resolutions: a collection of downsample rates at which
+        attention will take place. May be a set, list, or tuple.
+        For example, if this contains 4, then at 4x downsampling, attention
+        will be used.
+    :param dropout: the dropout probability.
+    :param channel_mult: channel multiplier for each level of the UNet.
+    :param conv_resample: if True, use learned convolutions for upsampling and
+        downsampling.
+    :param dims: determines if the signal is 1D, 2D, or 3D.
+    :param num_classes: if specified (as an int), then this model will be
+        class-conditional with `num_classes` classes.
+    :param use_checkpoint: use gradient checkpointing to reduce memory usage.
+    :param num_heads: the number of attention heads in each attention layer.
+    :param num_heads_channels: if specified, ignore num_heads and instead use
+                               a fixed channel width per attention head.
+    :param num_heads_upsample: works with num_heads to set a different number
+                               of heads for upsampling. Deprecated.
+    :param use_scale_shift_norm: use a FiLM-like conditioning mechanism.
+    :param resblock_updown: use residual blocks for up/downsampling.
+    :param use_new_attention_order: use a different attention pattern for potentially
+                                    increased efficiency.
+    """
+
+    def __init__(
+        self,
+        in_channels,
+        out_channels,
+        model_channels,
+        num_res_blocks,
+        attention_resolutions,
+        dropout=0,
+        channel_mult=(1, 2, 4, 8),
+        conv_resample=True,
+        dims=2,
+        num_classes=None,
+        use_checkpoint=False,
+        use_fp16=False,
+        num_heads=1,
+        num_head_channels=-1,
+        num_heads_upsample=-1,
+        use_scale_shift_norm=False,
+        resblock_updown=False,
+        use_new_attention_order=False,
+    ):
+        super().__init__()
+
+        if num_heads_upsample == -1:
+            num_heads_upsample = num_heads
+
+        self.in_channels = in_channels
+        self.model_channels = model_channels
+        self.out_channels = out_channels
+        self.num_res_blocks = num_res_blocks
+        self.attention_resolutions = attention_resolutions
+        self.dropout = dropout
+        self.channel_mult = channel_mult
+        self.conv_resample = conv_resample
+        self.num_classes = num_classes
+        self.use_checkpoint = use_checkpoint
+        self.dtype = torch.float16 if use_fp16 else torch.float32
+        self.num_heads = num_heads
+        self.num_head_channels = num_head_channels
+        self.num_heads_upsample = num_heads_upsample
+
+        time_embed_dim = model_channels * 4
+        self.dt_embed = nn.Sequential(
+            linear(model_channels, time_embed_dim),
+            nn.SiLU(),
+            linear(time_embed_dim, time_embed_dim),
+        )
+
+        if self.num_classes is not None:
+            self.label_emb = nn.Embedding(num_classes, time_embed_dim)
+
+        ch = input_ch = int(channel_mult[0] * model_channels)
+        self.input_blocks = nn.ModuleList(
+            [TimestepEmbedSequential(conv_nd(dims, in_channels, ch, 3, padding=1))]
+        )
+        self._feature_size = ch
+        input_block_chans = [ch]
+        ds = 1
+        for level, mult in enumerate(channel_mult):
+            for _ in range(num_res_blocks):
+                layers = [
+                    ResBlock(
+                        ch,
+                        time_embed_dim,
+                        dropout,
+                        out_channels=int(mult * model_channels),
+                        dims=dims,
+                        use_checkpoint=use_checkpoint,
+                        use_scale_shift_norm=use_scale_shift_norm,
+                    )
+                ]
+                ch = int(mult * model_channels)
+                if ds in attention_resolutions:
+                    layers.append(
+                        AttentionBlock(
+                            ch,
+                            use_checkpoint=use_checkpoint,
+                            num_heads=num_heads,
+                            num_head_channels=num_head_channels,
+                            use_new_attention_order=use_new_attention_order,
+                        )
+                    )
+                self.input_blocks.append(TimestepEmbedSequential(*layers))
+                self._feature_size += ch
+                input_block_chans.append(ch)
+            if level != len(channel_mult) - 1:
+                out_ch = ch
+                self.input_blocks.append(
+                    TimestepEmbedSequential(
+                        ResBlock(
+                            ch,
+                            time_embed_dim,
+                            dropout,
+                            out_channels=out_ch,
+                            dims=dims,
+                            use_checkpoint=use_checkpoint,
+                            use_scale_shift_norm=use_scale_shift_norm,
+                            down=True,
+                        )
+                        if resblock_updown
+                        else Downsample(
+                            ch, conv_resample, dims=dims, out_channels=out_ch
+                        )
+                    )
+                )
+                ch = out_ch
+                input_block_chans.append(ch)
+                ds *= 2
+                self._feature_size += ch
+
+        self.middle_block = TimestepEmbedSequential(
+            ResBlock(
+                ch,
+                time_embed_dim,
+                dropout,
+                dims=dims,
+                use_checkpoint=use_checkpoint,
+                use_scale_shift_norm=use_scale_shift_norm,
+            ),
+            AttentionBlock(
+                ch,
+                use_checkpoint=use_checkpoint,
+                num_heads=num_heads,
+                num_head_channels=num_head_channels,
+                use_new_attention_order=use_new_attention_order,
+            ),
+            ResBlock(
+                ch,
+                time_embed_dim,
+                dropout,
+                dims=dims,
+                use_checkpoint=use_checkpoint,
+                use_scale_shift_norm=use_scale_shift_norm,
+            ),
+        )
+        self._feature_size += ch
+
+        self.output_blocks = nn.ModuleList([])
+        for level, mult in list(enumerate(channel_mult))[::-1]:
+            for i in range(num_res_blocks + 1):
+                ich = input_block_chans.pop()
+                layers = [
+                    ResBlock(
+                        ch + ich,
+                        time_embed_dim,
+                        dropout,
+                        out_channels=int(model_channels * mult),
+                        dims=dims,
+                        use_checkpoint=use_checkpoint,
+                        use_scale_shift_norm=use_scale_shift_norm,
+                    )
+                ]
+                ch = int(model_channels * mult)
+                if ds in attention_resolutions:
+                    layers.append(
+                        AttentionBlock(
+                            ch,
+                            use_checkpoint=use_checkpoint,
+                            num_heads=num_heads_upsample,
+                            num_head_channels=num_head_channels,
+                            use_new_attention_order=use_new_attention_order,
+                        )
+                    )
+                if level and i == num_res_blocks:
+                    out_ch = ch
+                    layers.append(
+                        ResBlock(
+                            ch,
+                            time_embed_dim,
+                            dropout,
+                            out_channels=out_ch,
+                            dims=dims,
+                            use_checkpoint=use_checkpoint,
+                            use_scale_shift_norm=use_scale_shift_norm,
+                            up=True,
+                        )
+                        if resblock_updown
+                        else Upsample(ch, conv_resample, dims=dims, out_channels=out_ch)
+                    )
+                    ds //= 2
+                self.output_blocks.append(TimestepEmbedSequential(*layers))
+                self._feature_size += ch
+
+        self.out = nn.Sequential(
+            normalization(ch),
+            nn.SiLU(),
+            zero_module(conv_nd(dims, input_ch, out_channels, 3, padding=1)),
+        )
+
+    def convert_to_fp16(self):
+        """
+        Convert the torso of the model to float16.
+        """
+        self.input_blocks.apply(convert_module_to_f16)
+        self.middle_block.apply(convert_module_to_f16)
+        self.output_blocks.apply(convert_module_to_f16)
+
+    def convert_to_fp32(self):
+        """
+        Convert the torso of the model to float32.
+        """
+        self.input_blocks.apply(convert_module_to_f32)
+        self.middle_block.apply(convert_module_to_f32)
+        self.output_blocks.apply(convert_module_to_f32)
+
+    def forward(self, x, dt, y=None):
+        """
+        Apply the model to an input batch.
+
+        :param x: an [N x C x ...] Tensor of inputs.
+        :param timesteps: a 1-D batch of timesteps.
+        :param y: an [N] Tensor of labels, if class-conditional.
+        :return: an [N x C x ...] Tensor of outputs.
+        """
+        assert (y is not None) == (
+            self.num_classes is not None
+        ), "must specify y if and only if the model is class-conditional"
+
+        hs = []
+        emb = self.dt_embed(timestep_embedding(dt, self.model_channels))
+
+        if self.num_classes is not None:
+            assert y.shape == (x.shape[0],)
+            emb = emb + self.label_emb(y)
+
+        h = x.type(self.dtype)
+        for module in self.input_blocks:
+            h = module(h, emb)
+            hs.append(h)
+        h = self.middle_block(h, emb)
+        for module in self.output_blocks:
+            h = torch.cat([h, hs.pop()], dim=1)
+            h = module(h, emb)
+        h = h.type(x.dtype)
+        return self.out(h)
+
+
+
+class SkippedUNetModel(nn.Module):
+    """
+    The full UNet model with attention and timestep embedding.
+
+    :param in_channels: channels in the input Tensor.
+    :param model_channels: base channel count for the model.
+    :param out_channels: channels in the output Tensor.
+    :param num_res_blocks: number of residual blocks per downsample.
+    :param attention_resolutions: a collection of downsample rates at which
+        attention will take place. May be a set, list, or tuple.
+        For example, if this contains 4, then at 4x downsampling, attention
+        will be used.
+    :param dropout: the dropout probability.
+    :param channel_mult: channel multiplier for each level of the UNet.
+    :param conv_resample: if True, use learned convolutions for upsampling and
+        downsampling.
+    :param dims: determines if the signal is 1D, 2D, or 3D.
+    :param num_classes: if specified (as an int), then this model will be
+        class-conditional with `num_classes` classes.
+    :param use_checkpoint: use gradient checkpointing to reduce memory usage.
+    :param num_heads: the number of attention heads in each attention layer.
+    :param num_heads_channels: if specified, ignore num_heads and instead use
+                               a fixed channel width per attention head.
+    :param num_heads_upsample: works with num_heads to set a different number
+                               of heads for upsampling. Deprecated.
+    :param use_scale_shift_norm: use a FiLM-like conditioning mechanism.
+    :param resblock_updown: use residual blocks for up/downsampling.
+    :param use_new_attention_order: use a different attention pattern for potentially
+                                    increased efficiency.
+    """
+
+    def __init__(
+        self,
+        in_channels,
+        out_channels,
+        model_channels,
+        num_res_blocks,
+        attention_resolutions,
+        dropout=0,
+        channel_mult=(1, 2, 4, 8),
+        conv_resample=True,
+        dims=2,
+        num_classes=None,
+        use_checkpoint=False,
+        use_fp16=False,
+        num_heads=1,
+        num_head_channels=-1,
+        num_heads_upsample=-1,
+        use_scale_shift_norm=False,
+        resblock_updown=False,
+        use_new_attention_order=False,
+    ):
+        super().__init__()
+
+        if num_heads_upsample == -1:
+            num_heads_upsample = num_heads
+
+        self.in_channels = in_channels
+        self.model_channels = model_channels
+        self.out_channels = out_channels
+        self.num_res_blocks = num_res_blocks
+        self.attention_resolutions = attention_resolutions
+        self.dropout = dropout
+        self.channel_mult = channel_mult
+        self.conv_resample = conv_resample
+        self.num_classes = num_classes
+        self.use_checkpoint = use_checkpoint
+        self.dtype = torch.float16 if use_fp16 else torch.float32
+        self.num_heads = num_heads
+        self.num_head_channels = num_head_channels
+        self.num_heads_upsample = num_heads_upsample
+
+        time_embed_dim = model_channels * 4
+        self.dt_embed = nn.Sequential(
+            linear(model_channels, time_embed_dim),
+            nn.SiLU(),
+            linear(time_embed_dim, time_embed_dim),
+        )
+
+        if self.num_classes is not None:
+            self.label_emb = nn.Embedding(num_classes, time_embed_dim)
+
+        ch = input_ch = int(channel_mult[0] * model_channels)
+        self.input_blocks = nn.ModuleList(
+            [TimestepEmbedSequential(conv_nd(dims, in_channels, ch, 3, padding=1))]
+        )
+        self._feature_size = ch
+        input_block_chans = [ch]
+        ds = 1
+        for level, mult in enumerate(channel_mult):
+            for _ in range(num_res_blocks):
+                layers = [
+                    ResBlock(
+                        ch,
+                        time_embed_dim,
+                        dropout,
+                        out_channels=int(mult * model_channels),
+                        dims=dims,
+                        use_checkpoint=use_checkpoint,
+                        use_scale_shift_norm=use_scale_shift_norm,
+                    )
+                ]
+                ch = int(mult * model_channels)
+                if ds in attention_resolutions:
+                    layers.append(
+                        AttentionBlock(
+                            ch,
+                            use_checkpoint=use_checkpoint,
+                            num_heads=num_heads,
+                            num_head_channels=num_head_channels,
+                            use_new_attention_order=use_new_attention_order,
+                        )
+                    )
+                self.input_blocks.append(TimestepEmbedSequential(*layers))
+                self._feature_size += ch
+                input_block_chans.append(ch)
+            if level != len(channel_mult) - 1:
+                out_ch = ch
+                self.input_blocks.append(
+                    TimestepEmbedSequential(
+                        ResBlock(
+                            ch,
+                            time_embed_dim,
+                            dropout,
+                            out_channels=out_ch,
+                            dims=dims,
+                            use_checkpoint=use_checkpoint,
+                            use_scale_shift_norm=use_scale_shift_norm,
+                            down=True,
+                        )
+                        if resblock_updown
+                        else Downsample(
+                            ch, conv_resample, dims=dims, out_channels=out_ch
+                        )
+                    )
+                )
+                ch = out_ch
+                input_block_chans.append(ch)
+                ds *= 2
+                self._feature_size += ch
+
+        self.middle_block = TimestepEmbedSequential(
+            ResBlock(
+                ch,
+                time_embed_dim,
+                dropout,
+                dims=dims,
+                use_checkpoint=use_checkpoint,
+                use_scale_shift_norm=use_scale_shift_norm,
+            ),
+            AttentionBlock(
+                ch,
+                use_checkpoint=use_checkpoint,
+                num_heads=num_heads,
+                num_head_channels=num_head_channels,
+                use_new_attention_order=use_new_attention_order,
+            ),
+            ResBlock(
+                ch,
+                time_embed_dim,
+                dropout,
+                dims=dims,
+                use_checkpoint=use_checkpoint,
+                use_scale_shift_norm=use_scale_shift_norm,
+            ),
+        )
+        self._feature_size += ch
+
+        self.output_blocks = nn.ModuleList([])
+        for level, mult in list(enumerate(channel_mult))[::-1]:
+            for i in range(num_res_blocks + 1):
+                ich = input_block_chans.pop()
+                layers = [
+                    ResBlock(
+                        ch + ich,
+                        time_embed_dim,
+                        dropout,
+                        out_channels=int(model_channels * mult),
+                        dims=dims,
+                        use_checkpoint=use_checkpoint,
+                        use_scale_shift_norm=use_scale_shift_norm,
+                    )
+                ]
+                ch = int(model_channels * mult)
+                if ds in attention_resolutions:
+                    layers.append(
+                        AttentionBlock(
+                            ch,
+                            use_checkpoint=use_checkpoint,
+                            num_heads=num_heads_upsample,
+                            num_head_channels=num_head_channels,
+                            use_new_attention_order=use_new_attention_order,
+                        )
+                    )
+                if level and i == num_res_blocks:
+                    out_ch = ch
+                    layers.append(
+                        ResBlock(
+                            ch,
+                            time_embed_dim,
+                            dropout,
+                            out_channels=out_ch,
+                            dims=dims,
+                            use_checkpoint=use_checkpoint,
+                            use_scale_shift_norm=use_scale_shift_norm,
+                            up=True,
+                        )
+                        if resblock_updown
+                        else Upsample(ch, conv_resample, dims=dims, out_channels=out_ch)
+                    )
+                    ds //= 2
+                self.output_blocks.append(TimestepEmbedSequential(*layers))
+                self._feature_size += ch
+
+        self.out = nn.Sequential(
+            normalization(ch),
+            nn.SiLU(),
+            zero_module(conv_nd(dims, input_ch, out_channels, 3, padding=1)),
+        )
+
+        self.emb_h_layer = nn.Sequential(
+            linear(64 * 64 * 3, time_embed_dim),
+            nn.SiLU(),
+            linear(time_embed_dim, time_embed_dim),
+        )
+
+    def convert_to_fp16(self):
+        """
+        Convert the torso of the model to float16.
+        """
+        self.input_blocks.apply(convert_module_to_f16)
+        self.middle_block.apply(convert_module_to_f16)
+        self.output_blocks.apply(convert_module_to_f16)
+
+    def convert_to_fp32(self):
+        """
+        Convert the torso of the model to float32.
+        """
+        self.input_blocks.apply(convert_module_to_f32)
+        self.middle_block.apply(convert_module_to_f32)
+        self.output_blocks.apply(convert_module_to_f32)
+
+    def forward(self, x, dt, shadow=None, **kwargs):
+        """
+        Apply the model to an input batch.
+
+        :param x: an [N x C x ...] Tensor of inputs.
+        :param timesteps: a 1-D batch of timesteps.
+        :param y: an [N] Tensor of labels, if class-conditional.
+        :return: an [N x C x ...] Tensor of outputs.
+        """
+
+        hs = []
+        emb = self.dt_embed(timestep_embedding(dt, self.model_channels))
+        emb_h = F.adaptive_avg_pool2d(x, (64, 64)).view(x.shape[0], -1)
+        emb_h = self.emb_h_layer(emb_h)
+        emb = emb + emb_h
+
+        x = torch.cat([x, shadow], dim=1)
+
+        h = x.type(self.dtype)
+        for module in self.input_blocks:
+            h = module(h, emb)
+            hs.append(h)
+        h = self.middle_block(h, emb)
+        for module in self.output_blocks:
+            h = torch.cat([h, hs.pop()], dim=1)
+            h = module(h, emb)
+        h = h.type(x.dtype)
+        return self.out(h)
+
+
+class ConditionalUNetModel(UNetModel):
+    def __init__(self, in_channels, *args, **kwargs):
+        super().__init__(in_channels * 2, *args, **kwargs)
+
+    def forward(self, x, timesteps, shadow=None, **kwargs):
+        x = torch.cat([x, shadow], dim=1)
+        return super().forward(x, timesteps, **kwargs)
+
+
+class ConditionalMaskUNetModel(UNetModel):
+    def __init__(self, in_channels, out_channels, *args, **kwargs):
+        super().__init__(in_channels * 3, out_channels, *args, **kwargs)
+
+    def forward(self, x, timesteps, shadow=None, mask=None, **kwargs):
+        x = torch.cat([x, shadow, mask], dim=1)
+        x = super().forward(x, timesteps, **kwargs)
+        return x
+
+
+class ConditionalLatentUNetModel(UNetModel):
+    def __init__(self, in_channels, out_channels, *args, **kwargs):
+        super().__init__(in_channels, out_channels, *args, **kwargs)
+
+    def forward(self, x, timesteps, latent=None, **kwargs):
+        x = torch.cat([x, latent], dim=1)
+        x = super().forward(x, timesteps, **kwargs)
+        return x
+
+
+
+class DensePosteriorConditionalUNet(nn.Module):
+    def __init__(
+        self,
+        in_channels,
+        out_channels,
+        model_channels,
+        num_res_blocks,
+        attention_resolutions,
+        dropout=0,
+        channel_mult=(1, 2, 4, 8),
+        conv_resample=True,
+        dims=2,
+        num_classes=None,
+        use_checkpoint=False,
+        use_fp16=False,
+        num_heads=1,
+        num_head_channels=-1,
+        num_heads_upsample=-1,
+        use_scale_shift_norm=False,
+        resblock_updown=False,
+        use_new_attention_order=False,
+    ):
+        super().__init__()
+
+        if num_heads_upsample == -1:
+            num_heads_upsample = num_heads
+
+        self.in_channels = in_channels
+        self.model_channels = model_channels
+        self.out_channels = out_channels
+        self.num_res_blocks = num_res_blocks
+        self.attention_resolutions = attention_resolutions
+        self.dropout = dropout
+        self.channel_mult = channel_mult
+        self.conv_resample = conv_resample
+        self.num_classes = num_classes
+        self.use_checkpoint = use_checkpoint
+        self.dtype = torch.float16 if use_fp16 else torch.float32
+        self.num_heads = num_heads
+        self.num_head_channels = num_head_channels
+        self.num_heads_upsample = num_heads_upsample
+
+        time_embed_dim = model_channels * 4
+        self.dt_embed = nn.Sequential(
+            linear(model_channels, time_embed_dim),
+            nn.SiLU(),
+            linear(time_embed_dim, time_embed_dim),
+        )
+
+        if self.num_classes is not None:
+            self.label_emb = nn.Embedding(num_classes, time_embed_dim)
+
+        ch = input_ch = int(channel_mult[0] * model_channels)
+        self.input_blocks = nn.ModuleList(
+            [TimestepEmbedSequential(conv_nd(dims, in_channels, ch, 3, padding=1))]
+        )
+        self._feature_size = ch
+        input_block_chans = [ch]
+        ds = 1
+        for level, mult in enumerate(channel_mult):
+            for _ in range(num_res_blocks):
+                layers = [
+                    ResBlock(
+                        ch,
+                        time_embed_dim,
+                        dropout,
+                        out_channels=int(mult * model_channels),
+                        dims=dims,
+                        use_checkpoint=use_checkpoint,
+                        use_scale_shift_norm=use_scale_shift_norm,
+                    )
+                ]
+                ch = int(mult * model_channels)
+                if ds in attention_resolutions:
+                    layers.append(
+                        AttentionBlock(
+                            ch,
+                            use_checkpoint=use_checkpoint,
+                            num_heads=num_heads,
+                            num_head_channels=num_head_channels,
+                            use_new_attention_order=use_new_attention_order,
+                        )
+                    )
+                self.input_blocks.append(TimestepEmbedSequential(*layers))
+                self._feature_size += ch
+                input_block_chans.append(ch)
+            if level != len(channel_mult) - 1:
+                out_ch = ch
+                self.input_blocks.append(
+                    TimestepEmbedSequential(
+                        ResBlock(
+                            ch,
+                            time_embed_dim,
+                            dropout,
+                            out_channels=out_ch,
+                            dims=dims,
+                            use_checkpoint=use_checkpoint,
+                            use_scale_shift_norm=use_scale_shift_norm,
+                            down=True,
+                        )
+                        if resblock_updown
+                        else Downsample(
+                            ch, conv_resample, dims=dims, out_channels=out_ch
+                        )
+                    )
+                )
+                ch = out_ch
+                input_block_chans.append(ch)
+                ds *= 2
+                self._feature_size += ch
+
+        self.middle_block = TimestepEmbedSequential(
+            ResBlock(
+                ch,
+                time_embed_dim,
+                dropout,
+                dims=dims,
+                use_checkpoint=use_checkpoint,
+                use_scale_shift_norm=use_scale_shift_norm,
+            ),
+            AttentionBlock(
+                ch,
+                use_checkpoint=use_checkpoint,
+                num_heads=num_heads,
+                num_head_channels=num_head_channels,
+                use_new_attention_order=use_new_attention_order,
+            ),
+            ResBlock(
+                ch,
+                time_embed_dim,
+                dropout,
+                dims=dims,
+                use_checkpoint=use_checkpoint,
+                use_scale_shift_norm=use_scale_shift_norm,
+            ),
+        )
+        self._feature_size += ch
+
+        self.output_blocks = nn.ModuleList([])
+        for level, mult in list(enumerate(channel_mult))[::-1]:
+            for i in range(num_res_blocks + 1):
+                ich = input_block_chans.pop()
+                layers = [
+                    ResBlock(
+                        ch + ich,
+                        time_embed_dim,
+                        dropout,
+                        out_channels=int(model_channels * mult),
+                        dims=dims,
+                        use_checkpoint=use_checkpoint,
+                        use_scale_shift_norm=use_scale_shift_norm,
+                    )
+                ]
+                ch = int(model_channels * mult)
+                if ds in attention_resolutions:
+                    layers.append(
+                        AttentionBlock(
+                            ch,
+                            use_checkpoint=use_checkpoint,
+                            num_heads=num_heads_upsample,
+                            num_head_channels=num_head_channels,
+                            use_new_attention_order=use_new_attention_order,
+                        )
+                    )
+                if level and i == num_res_blocks:
+                    out_ch = ch
+                    layers.append(
+                        ResBlock(
+                            ch,
+                            time_embed_dim,
+                            dropout,
+                            out_channels=out_ch,
+                            dims=dims,
+                            use_checkpoint=use_checkpoint,
+                            use_scale_shift_norm=use_scale_shift_norm,
+                            up=True,
+                        )
+                        if resblock_updown
+                        else Upsample(ch, conv_resample, dims=dims, out_channels=out_ch)
+                    )
+                    ds //= 2
+                self.output_blocks.append(TimestepEmbedSequential(*layers))
+                self._feature_size += ch
+
+        self.out = nn.Sequential(
+            normalization(ch),
+            nn.SiLU(),
+            zero_module(conv_nd(dims, input_ch, out_channels, 3, padding=1)),
+        )
+
+        self.emb_h_layer = nn.Sequential(
+            linear(64 * 64 * 3, time_embed_dim),
+            nn.SiLU(),
+            linear(time_embed_dim, time_embed_dim),
+        )
+
+    def convert_to_fp16(self):
+        """
+        Convert the torso of the model to float16.
+        """
+        self.input_blocks.apply(convert_module_to_f16)
+        self.middle_block.apply(convert_module_to_f16)
+        self.output_blocks.apply(convert_module_to_f16)
+
+    def convert_to_fp32(self):
+        """
+        Convert the torso of the model to float32.
+        """
+        self.input_blocks.apply(convert_module_to_f32)
+        self.middle_block.apply(convert_module_to_f32)
+        self.output_blocks.apply(convert_module_to_f32)
+
+    def forward(self, x, dt, latent, **kwargs):
+        """
+        Apply the model to an input batch.
+
+        :param x: an [N x C x ...] Tensor of inputs.
+        :param timesteps: a 1-D batch of timesteps.
+        :param y: an [N] Tensor of labels, if class-conditional.
+        :return: an [N x C x ...] Tensor of outputs.
+        """
+
+
+        hs = []
+        emb = self.dt_embed(timestep_embedding(dt, self.model_channels))
+        emb_h = F.adaptive_avg_pool2d(x, (64, 64)).view(x.shape[0], -1)
+        emb_h = self.emb_h_layer(emb_h)
+        emb = emb + emb_h
+
+        x = torch.cat([x, latent], dim=1)
+
+        h = x.type(self.dtype)
+        for module in self.input_blocks:
+            h = module(h, emb)
+            hs.append(h)
+        h = self.middle_block(h, emb)
+        for module in self.output_blocks:
+            h = torch.cat([h, hs.pop()], dim=1)
+            h = module(h, emb)
+        h = h.type(x.dtype)
+        return self.out(h)

header.html

@@ -0,0 +1,18 @@
+<div style="text-align: center; max-width: 650px; margin: 0 auto;">
+    <div style="
+          display: inline-flex;
+          gap: 0.8rem;
+          font-size: 1.75rem;
+          justify-content: center;
+          margin-bottom: 10px;
+        ">
+      <h1 style="font-weight: 900; align-items: center; margin-bottom: 7px; margin-top: 20px;">
+        Instance Shadow Removal
+      </h1>
+    </div>
+    <div>
+      <p style="align-items: center; margin-bottom: 7px;">
+        The new sota shadow removal model even for instnace case. <br> Add a mask for what shadow want to removal. Please draw as accurate as possible.
+      </p>
+    </div>
+  </div>

net_g_400000.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9eb122113d07cfb5321ea20f34884a739171e74c4d9cb54d8ee69ed485b8e3f
+size 4532130337