Rectified_Noise/VP-Disp/models.py

# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
# --------------------------------------------------------
# References:
# GLIDE: https://github.com/openai/glide-text2im
# MAE: https://github.com/facebookresearch/mae/blob/main/models_mae.py
# --------------------------------------------------------

import torch
import torch.nn as nn
import numpy as np
import math
from timm.models.vision_transformer import PatchEmbed, Attention, Mlp


def modulate(x, shift, scale):
    return x * (1 + scale.unsqueeze(1)) + shift.unsqueeze(1)


#################################################################################
#               Embedding Layers for Timesteps and Class Labels                 #
#################################################################################

class TimestepEmbedder(nn.Module):
    """
    Embeds scalar timesteps into vector representations.
    """
    def __init__(self, hidden_size, frequency_embedding_size=256):
        super().__init__()
        self.mlp = nn.Sequential(
            nn.Linear(frequency_embedding_size, hidden_size, bias=True),
            nn.SiLU(),
            nn.Linear(hidden_size, hidden_size, bias=True),
        )
        self.frequency_embedding_size = frequency_embedding_size

    @staticmethod
    def timestep_embedding(t, dim, max_period=10000):
        """
        Create sinusoidal timestep embeddings.
        :param t: a 1-D Tensor of N indices, one per batch element.
                          These may be fractional.
        :param dim: the dimension of the output.
        :param max_period: controls the minimum frequency of the embeddings.
        :return: an (N, D) Tensor of positional embeddings.
        """
        # https://github.com/openai/glide-text2im/blob/main/glide_text2im/nn.py
        half = dim // 2
        freqs = torch.exp(
            -math.log(max_period) * torch.arange(start=0, end=half, dtype=torch.float32) / half
        ).to(device=t.device)
        args = t[:, None].float() * freqs[None]
        embedding = torch.cat([torch.cos(args), torch.sin(args)], dim=-1)
        if dim % 2:
            embedding = torch.cat([embedding, torch.zeros_like(embedding[:, :1])], dim=-1)
        return embedding

    def forward(self, t):
        t_freq = self.timestep_embedding(t, self.frequency_embedding_size)
        t_emb = self.mlp(t_freq)
        return t_emb


class LabelEmbedder(nn.Module):
    """
    Embeds class labels into vector representations. Also handles label dropout for classifier-free guidance.
    """
    def __init__(self, num_classes, hidden_size, dropout_prob):
        super().__init__()
        use_cfg_embedding = dropout_prob > 0
        self.embedding_table = nn.Embedding(num_classes + use_cfg_embedding, hidden_size)
        self.num_classes = num_classes
        self.dropout_prob = dropout_prob

    def token_drop(self, labels, force_drop_ids=None):
        """
        Drops labels to enable classifier-free guidance.
        """
        if force_drop_ids is None:
            drop_ids = torch.rand(labels.shape[0], device=labels.device) < self.dropout_prob
        else:
            drop_ids = force_drop_ids == 1
        labels = torch.where(drop_ids, self.num_classes, labels)
        return labels

    def forward(self, labels, train, force_drop_ids=None):
        use_dropout = self.dropout_prob > 0
        if (train and use_dropout) or (force_drop_ids is not None):
            labels = self.token_drop(labels, force_drop_ids)
        embeddings = self.embedding_table(labels)
        return embeddings


#################################################################################
#                                 Core SiT Model                                #
#################################################################################

class SiTBlock(nn.Module):
    """
    A SiT block with adaptive layer norm zero (adaLN-Zero) conditioning.
    """
    def __init__(self, hidden_size, num_heads, mlp_ratio=4.0, **block_kwargs):
        super().__init__()
        self.norm1 = nn.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6)
        self.attn = Attention(hidden_size, num_heads=num_heads, qkv_bias=True, **block_kwargs)
        self.norm2 = nn.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6)
        mlp_hidden_dim = int(hidden_size * mlp_ratio)
        approx_gelu = lambda: nn.GELU(approximate="tanh")
        self.mlp = Mlp(in_features=hidden_size, hidden_features=mlp_hidden_dim, act_layer=approx_gelu, drop=0)
        self.adaLN_modulation = nn.Sequential(
            nn.SiLU(),
            nn.Linear(hidden_size, 6 * hidden_size, bias=True)
        )

    def forward(self, x, c):
        shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = self.adaLN_modulation(c).chunk(6, dim=1)
        x = x + gate_msa.unsqueeze(1) * self.attn(modulate(self.norm1(x), shift_msa, scale_msa))
        x = x + gate_mlp.unsqueeze(1) * self.mlp(modulate(self.norm2(x), shift_mlp, scale_mlp))
        return x


class FinalLayer(nn.Module):
    """
    The final layer of SiT.
    """
    def __init__(self, hidden_size, patch_size, out_channels):
        super().__init__()
        self.norm_final = nn.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6)
        self.linear = nn.Linear(hidden_size, patch_size * patch_size * out_channels, bias=True)
        self.adaLN_modulation = nn.Sequential(
            nn.SiLU(),
            nn.Linear(hidden_size, 2 * hidden_size, bias=True)
        )

    def forward(self, x, c):
        shift, scale = self.adaLN_modulation(c).chunk(2, dim=1)
        x = modulate(self.norm_final(x), shift, scale)
        x = self.linear(x)
        return x


class SiT(nn.Module):
    """
    Diffusion model with a Transformer backbone.
    """
    def __init__(
        self,
        input_size=32,
        patch_size=2,
        in_channels=4,
        hidden_size=1152,
        depth=28,
        num_heads=16,
        mlp_ratio=4.0,
        class_dropout_prob=0.1,
        num_classes=1000,
        learn_sigma=True,
    ):
        super().__init__()
        self.learn_sigma = learn_sigma
        self.learn_sigma = True
        self.in_channels = in_channels
        self.out_channels = in_channels * 2
        self.patch_size = patch_size
        self.num_heads = num_heads

        self.x_embedder = PatchEmbed(input_size, patch_size, in_channels, hidden_size, bias=True)
        self.t_embedder = TimestepEmbedder(hidden_size)
        self.y_embedder = LabelEmbedder(num_classes, hidden_size, class_dropout_prob)
        num_patches = self.x_embedder.num_patches
        # Will use fixed sin-cos embedding:
        self.pos_embed = nn.Parameter(torch.zeros(1, num_patches, hidden_size), requires_grad=False)

        self.blocks = nn.ModuleList([
            SiTBlock(hidden_size, num_heads, mlp_ratio=mlp_ratio) for _ in range(depth)
        ])
        self.final_layer = FinalLayer(hidden_size, patch_size, self.out_channels)
        self.initialize_weights()

    def initialize_weights(self):
        # Initialize transformer layers:
        def _basic_init(module):
            if isinstance(module, nn.Linear):
                torch.nn.init.xavier_uniform_(module.weight)
                if module.bias is not None:
                    nn.init.constant_(module.bias, 0)
        self.apply(_basic_init)

        # Initialize (and freeze) pos_embed by sin-cos embedding:
        pos_embed = get_2d_sincos_pos_embed(self.pos_embed.shape[-1], int(self.x_embedder.num_patches ** 0.5))
        self.pos_embed.data.copy_(torch.from_numpy(pos_embed).float().unsqueeze(0))

        # Initialize patch_embed like nn.Linear (instead of nn.Conv2d):
        w = self.x_embedder.proj.weight.data
        nn.init.xavier_uniform_(w.view([w.shape[0], -1]))
        nn.init.constant_(self.x_embedder.proj.bias, 0)

        # Initialize label embedding table:
        nn.init.normal_(self.y_embedder.embedding_table.weight, std=0.02)

        # Initialize timestep embedding MLP:
        nn.init.normal_(self.t_embedder.mlp[0].weight, std=0.02)
        nn.init.normal_(self.t_embedder.mlp[2].weight, std=0.02)

        # Zero-out adaLN modulation layers in SiT blocks:
        for block in self.blocks:
            nn.init.constant_(block.adaLN_modulation[-1].weight, 0)
            nn.init.constant_(block.adaLN_modulation[-1].bias, 0)

        # Zero-out output layers:
        nn.init.constant_(self.final_layer.adaLN_modulation[-1].weight, 0)
        nn.init.constant_(self.final_layer.adaLN_modulation[-1].bias, 0)
        nn.init.constant_(self.final_layer.linear.weight, 0)
        nn.init.constant_(self.final_layer.linear.bias, 0)

    def unpatchify(self, x):
        """
        x: (N, T, patch_size**2 * C)
        imgs: (N, H, W, C)
        """
        c = self.out_channels
        p = self.x_embedder.patch_size[0]
        h = w = int(x.shape[1] ** 0.5)
        assert h * w == x.shape[1]

        x = x.reshape(shape=(x.shape[0], h, w, p, p, c))
        x = torch.einsum('nhwpqc->nchpwq', x)
        imgs = x.reshape(shape=(x.shape[0], c, h * p, h * p))
        return imgs

    def forward(self, x, t, y, return_act=False):
        """
        Forward pass of SiT.
        x: (N, C, H, W) tensor of spatial inputs (images or latent representations of images)
        t: (N,) tensor of diffusion timesteps
        y: (N,) tensor of class labels
        return_act: if True, return activations from transformer blocks
        """
        act = []
        x = self.x_embedder(x) + self.pos_embed  # (N, T, D), where T = H * W / patch_size ** 2
        t = self.t_embedder(t)                   # (N, D)
        y = self.y_embedder(y, self.training)    # (N, D)
        c = t + y                                # (N, D)
        for block in self.blocks:
            x = block(x, c)                      # (N, T, D)
            if return_act:
                act.append(x)
        x = self.final_layer(x, c)                # (N, T, patch_size ** 2 * out_channels)
        x = self.unpatchify(x)                   # (N, out_channels, H, W)
        if self.learn_sigma:
            x, _ = x.chunk(2, dim=1)
        if return_act:
            return x, act
        return x

    def forward_with_cfg(self, x, t, y, cfg_scale):
        """
        Forward pass of SiT, but also batches the unconSiTional forward pass for classifier-free guidance.
        """
        # https://github.com/openai/glide-text2im/blob/main/notebooks/text2im.ipynb
        half = x[: len(x) // 2]
        combined = torch.cat([half, half], dim=0)
        model_out = self.forward(combined, t, y)
        # For exact reproducibility reasons, we apply classifier-free guidance on only
        # three channels by default. The standard approach to cfg applies it to all channels.
        # This can be done by uncommenting the following line and commenting-out the line following that.
        # eps, rest = model_out[:, :self.in_channels], model_out[:, self.in_channels:]
        eps, rest = model_out[:, :3], model_out[:, 3:]
        cond_eps, uncond_eps = torch.split(eps, len(eps) // 2, dim=0)
        half_eps = uncond_eps + cfg_scale * (cond_eps - uncond_eps)
        eps = torch.cat([half_eps, half_eps], dim=0)
        return torch.cat([eps, rest], dim=1)


#################################################################################
#                   Sine/Cosine Positional Embedding Functions                  #
#################################################################################
# https://github.com/facebookresearch/mae/blob/main/util/pos_embed.py

def get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False, extra_tokens=0):
    """
    grid_size: int of the grid height and width
    return:
    pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or w/o cls_token)
    """
    grid_h = np.arange(grid_size, dtype=np.float32)
    grid_w = np.arange(grid_size, dtype=np.float32)
    grid = np.meshgrid(grid_w, grid_h)  # here w goes first
    grid = np.stack(grid, axis=0)

    grid = grid.reshape([2, 1, grid_size, grid_size])
    pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)
    if cls_token and extra_tokens > 0:
        pos_embed = np.concatenate([np.zeros([extra_tokens, embed_dim]), pos_embed], axis=0)
    return pos_embed


def get_2d_sincos_pos_embed_from_grid(embed_dim, grid):
    assert embed_dim % 2 == 0

    # use half of dimensions to encode grid_h
    emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0])  # (H*W, D/2)
    emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1])  # (H*W, D/2)

    emb = np.concatenate([emb_h, emb_w], axis=1) # (H*W, D)
    return emb


def get_1d_sincos_pos_embed_from_grid(embed_dim, pos):
    """
    embed_dim: output dimension for each position
    pos: a list of positions to be encoded: size (M,)
    out: (M, D)
    """
    assert embed_dim % 2 == 0
    omega = np.arange(embed_dim // 2, dtype=np.float64)
    omega /= embed_dim / 2.
    omega = 1. / 10000**omega  # (D/2,)

    pos = pos.reshape(-1)  # (M,)
    out = np.einsum('m,d->md', pos, omega)  # (M, D/2), outer product

    emb_sin = np.sin(out) # (M, D/2)
    emb_cos = np.cos(out) # (M, D/2)

    emb = np.concatenate([emb_sin, emb_cos], axis=1)  # (M, D)
    return emb


#################################################################################
#                                   SiT Configs                                  #
#################################################################################

def SiT_XL_2(**kwargs):
    return SiT(depth=28, hidden_size=1152, patch_size=2, num_heads=16, **kwargs)

def SiT_XL_4(**kwargs):
    return SiT(depth=28, hidden_size=1152, patch_size=4, num_heads=16, **kwargs)

def SiT_XL_8(**kwargs):
    return SiT(depth=28, hidden_size=1152, patch_size=8, num_heads=16, **kwargs)

def SiT_L_2(**kwargs):
    return SiT(depth=24, hidden_size=1024, patch_size=2, num_heads=16, **kwargs)

def SiT_L_4(**kwargs):
    return SiT(depth=24, hidden_size=1024, patch_size=4, num_heads=16, **kwargs)

def SiT_L_8(**kwargs):
    return SiT(depth=24, hidden_size=1024, patch_size=8, num_heads=16, **kwargs)

def SiT_B_2(**kwargs):
    return SiT(depth=12, hidden_size=768, patch_size=2, num_heads=12, **kwargs)

def SiT_B_4(**kwargs):
    return SiT(depth=12, hidden_size=768, patch_size=4, num_heads=12, **kwargs)

def SiT_B_8(**kwargs):
    return SiT(depth=12, hidden_size=768, patch_size=8, num_heads=12, **kwargs)

def SiT_S_2(**kwargs):
    return SiT(depth=12, hidden_size=384, patch_size=2, num_heads=6, **kwargs)

def SiT_S_4(**kwargs):
    return SiT(depth=12, hidden_size=384, patch_size=4, num_heads=6, **kwargs)

def SiT_S_8(**kwargs):
    return SiT(depth=12, hidden_size=384, patch_size=8, num_heads=6, **kwargs)


SiT_models = {
    'SiT-XL/2': SiT_XL_2,  'SiT-XL/4': SiT_XL_4,  'SiT-XL/8': SiT_XL_8,
    'SiT-L/2':  SiT_L_2,   'SiT-L/4':  SiT_L_4,   'SiT-L/8':  SiT_L_8,
    'SiT-B/2':  SiT_B_2,   'SiT-B/4':  SiT_B_4,   'SiT-B/8':  SiT_B_8,
    'SiT-S/2':  SiT_S_2,   'SiT-S/4':  SiT_S_4,   'SiT-S/8':  SiT_S_8,
}

#################################################################################
#                                 SiTF1, SiTF2, CombinedModel                   #
#################################################################################

class SiTF1(nn.Module):
    """
    SiTF1 Model
    """
    def __init__(
        self,
        input_size=32,
        patch_size=2,
        in_channels=4,
        hidden_size=1152,
        depth=28,
        num_heads=16,
        mlp_ratio=4.0,
        class_dropout_prob=0.1,
        num_classes=1000,
        learn_sigma=True,
        final_layer=None,
    ):
        super().__init__()
        self.input_size = input_size
        self.patch_size= patch_size
        self.hidden_size = hidden_size
        self.in_channels = in_channels
        self.out_channels = in_channels * 2
        self.patch_size = patch_size
        self.num_heads = num_heads
        self.learn_sigma = learn_sigma
        self.x_embedder = PatchEmbed(input_size, patch_size, in_channels, hidden_size, bias=True)
        self.t_embedder = TimestepEmbedder(hidden_size)
        self.y_embedder = LabelEmbedder(num_classes, hidden_size, class_dropout_prob)
        num_patches = self.x_embedder.num_patches
        self.pos_embed = nn.Parameter(torch.zeros(1, num_patches, hidden_size), requires_grad=False)
        self.blocks = nn.ModuleList([
            SiTBlock(hidden_size, num_heads, mlp_ratio=mlp_ratio) for _ in range(depth)
        ])
        self.final_layer = FinalLayer(hidden_size, patch_size, self.out_channels)
        self.initialize_weights()
        
    def unpatchify(self, x):
        """
        x: (N, T, patch_size**2 * C)
        imgs: (N, H, W, C)
        """
        c = self.out_channels
        p = self.x_embedder.patch_size[0]
        h = w = int(x.shape[1] ** 0.5)
        assert h * w == x.shape[1]

        x = x.reshape(shape=(x.shape[0], h, w, p, p, c))
        x = torch.einsum('nhwpqc->nchpwq', x)
        imgs = x.reshape(shape=(x.shape[0], c, h * p, h * p))
        return imgs
    
    def initialize_weights(self):
        def _basic_init(module):
            if isinstance(module, nn.Linear):
                torch.nn.init.xavier_uniform_(module.weight)
                if module.bias is not None:
                    nn.init.constant_(module.bias, 0)
        self.apply(_basic_init)
        pos_embed = get_2d_sincos_pos_embed(self.pos_embed.shape[-1], int(self.x_embedder.num_patches ** 0.5))
        self.pos_embed.data.copy_(torch.from_numpy(pos_embed).float().unsqueeze(0))
        w = self.x_embedder.proj.weight.data
        nn.init.xavier_uniform_(w.view([w.shape[0], -1]))
        nn.init.constant_(self.x_embedder.proj.bias, 0)
        nn.init.normal_(self.y_embedder.embedding_table.weight, std=0.02)
        nn.init.normal_(self.t_embedder.mlp[0].weight, std=0.02)
        nn.init.normal_(self.t_embedder.mlp[2].weight, std=0.02)
        for block in self.blocks:
            nn.init.constant_(block.adaLN_modulation[-1].weight, 0)
            nn.init.constant_(block.adaLN_modulation[-1].bias, 0)

    def forward(self, x, t, y):
        x = self.x_embedder(x) + self.pos_embed
        t = self.t_embedder(t)
        y = self.y_embedder(y, self.training)
        c = t + y
        for block in self.blocks:
            x = block(x, c)
        x_now = self.final_layer(x, c)                # (N, T, patch_size ** 2 * out_channels)
        x_now = self.unpatchify(x_now)                   # (N, out_channels, H, W)
        x_now, _ = x_now.chunk(2, dim=1)
        return x,x_now  # patch token (N, T, D)

    def forward_with_cfg(self, x, t, y, cfg_scale):
        """
        Forward pass with classifier-free guidance for SiTF1.
        Applies guidance consistently to both patch tokens and image output (x_now).
        """
        # Take the first half (conditional inputs) and duplicate it so that
        # it can be paired with conditional and unconditional labels in `y`.
        half = x[: len(x) // 2]
        combined = torch.cat([half, half], dim=0)
        patch_tokens, x_now = self.forward(combined, t, y)

        # Apply CFG on the image output channels (first 3 channels by default)
        eps, rest = x_now[:, :3, ...], x_now[:, 3:, ...]
        cond_eps, uncond_eps = torch.split(eps, len(eps) // 2, dim=0)
        half_eps = uncond_eps + cfg_scale * (cond_eps - uncond_eps)
        eps = torch.cat([half_eps, half_eps], dim=0)
        x_now = torch.cat([eps, rest], dim=1)

        # Apply same guidance logic to patch tokens so downstream modules see
        # a consistent guided representation.
        cond_tok, uncond_tok = torch.split(patch_tokens, len(patch_tokens) // 2, dim=0)
        half_tok = uncond_tok + cfg_scale * (cond_tok - uncond_tok)
        patch_tokens = torch.cat([half_tok, half_tok], dim=0)

        return patch_tokens, x_now


class SiTF2(nn.Module):
    """
    SiTF2:
    """
    def __init__(
        self,
        input_size=32,
        hidden_size=1152,
        out_channels=8,  
        patch_size=2,
        num_heads=16,
        mlp_ratio=4.0,
        depth=4,
        learn_sigma=True,
        final_layer=None,
        num_classes=1000,
        class_dropout_prob=0.1,
        learn_mu=False,
    ):
        super().__init__()
        self.learn_sigma = learn_sigma
        self.learn_mu = learn_mu
        self.out_channels = out_channels
        self.in_channels = 4
        self.patch_size = patch_size
        self.num_heads = num_heads
        self.blocks = nn.ModuleList([
            SiTBlock(hidden_size, num_heads, mlp_ratio=mlp_ratio) for _ in range(depth)
        ])
        self.x_embedder = PatchEmbed(input_size, patch_size, self.in_channels, hidden_size, bias=True)
        self.t_embedder = TimestepEmbedder(hidden_size)
        self.y_embedder = LabelEmbedder(num_classes, hidden_size, class_dropout_prob)
        num_patches = self.x_embedder.num_patches
        self.num_patches = num_patches  # Save original num_patches for unpatchify
        # pos_embed needs to support 2*num_patches for concatenated input
        self.pos_embed = nn.Parameter(torch.zeros(1, 2 * num_patches, hidden_size), requires_grad=False)
        # Initialize pos_embed with sin-cos embedding
        pos_embed = get_2d_sincos_pos_embed(hidden_size, int(num_patches ** 0.5))
        # Repeat the pos_embed for both halves (or could use different embeddings)
        pos_embed_full = np.concatenate([pos_embed, pos_embed], axis=0)
        self.pos_embed.data.copy_(torch.from_numpy(pos_embed_full).float().unsqueeze(0))

        if final_layer is not None:
            self.final_layer = final_layer
        else:
            self.final_layer = FinalLayer(hidden_size, patch_size, out_channels)
            # if depth !=0:
            #     for p in self.final_layer.parameters():
            #         if p is not None:
            #             torch.nn.init.constant_(p, 0)

    def unpatchify(self, x, patch_size, out_channels):
        c = out_channels
        p = patch_size
        # x.shape[1] might be 2*num_patches when using concatenated input
        # Use original num_patches to calculate h and w
        h = w = int(self.num_patches ** 0.5)
        # If input has 2*num_patches, we need to handle it
        if x.shape[1] == 2 * self.num_patches:
            # Take only the first half (or average, or other strategy)
            # For now, we'll take the first half
            x = x[:, :self.num_patches, :]
        assert h * w == x.shape[1], f"Expected {h * w} patches, got {x.shape[1]}"
        x = x.reshape(shape=(x.shape[0], h, w, p, p, c))
        x = torch.einsum('nhwpqc->nchpwq', x)
        imgs = x.reshape(shape=(x.shape[0], c, h * p, h * p))
        return imgs

    def forward(self, x, c, t, return_act=False):
        act = []
        for block in self.blocks:
            x = block(x, c)
            if return_act:
                act.append(x)
        x = self.final_layer(x, c)
        x = self.unpatchify(x, self.patch_size, self.out_channels)
        if self.learn_sigma:
            mean_pred, log_var_pred = x.chunk(2, dim=1)
            variance_pred = torch.exp(log_var_pred)
            std_dev_pred = torch.sqrt(variance_pred)
            noise = torch.randn_like(mean_pred)
            #uniform_noise = torch.rand_like(mean_pred)
            #uniform_noise = uniform_noise.clamp(min=1e-5, max=1-1e-5) 
            #gumbel_noise = -torch.log(-torch.log(uniform_noise))

            if self.learn_mu==True:
                resampled_x = mean_pred + std_dev_pred * noise     
            else:
                resampled_x = std_dev_pred * noise
            x = resampled_x
        else:
            x, _ = x.chunk(2, dim=1)
        if return_act:
            return x, act
        return x
    
    def forward_noise(self, x, c):
        for block in self.blocks:
            x = block(x, c)
        x = self.final_layer(x, c)
        x = self.unpatchify(x, self.patch_size, self.out_channels)
        if self.learn_sigma:
            mean_pred, log_var_pred = x.chunk(2, dim=1)
            variance_pred = torch.exp(log_var_pred)
            std_dev_pred = torch.sqrt(variance_pred)
            noise = torch.randn_like(mean_pred)
            if self.learn_mu==True:
                resampled_x = mean_pred + std_dev_pred * noise     
            else:
                resampled_x = std_dev_pred * noise
            x = resampled_x
        else:
            x, _ = x.chunk(2, dim=1)
        return x

#有两种写法，一种是拿理想的，一种是拿真实的，一种是拼接，一种是加和
class CombinedModel(nn.Module):
    """
    CombinedModel。
    """
    def __init__(self, sitf1: SiTF1, sitf2: SiTF2):
        super().__init__()
        self.sitf1 = sitf1
        self.sitf2 = sitf2
        input_size=self.sitf1.input_size
        patch_size=self.sitf1.patch_size
        hidden_size=self.sitf1.hidden_size
        self.x_embedder = PatchEmbed(input_size, patch_size, 4, hidden_size, bias=True)
        num_patches = self.x_embedder.num_patches
        # pos_embed needs to support 2*num_patches for concatenated input
        self.pos_embed = nn.Parameter(torch.zeros(1, 2 * num_patches, hidden_size), requires_grad=False)
        # Initialize pos_embed with sin-cos embedding
        pos_embed = get_2d_sincos_pos_embed(hidden_size, int(num_patches ** 0.5))
        # Repeat the pos_embed for both halves (or could use different embeddings)
        pos_embed_full = np.concatenate([pos_embed, pos_embed], axis=0)
        self.pos_embed.data.copy_(torch.from_numpy(pos_embed_full).float().unsqueeze(0))
        
    def forward(self, x, t, y, return_act=False):
        patch_tokens,x_now = self.sitf1(x, t, y)
        # Interpolate between x_now and x using timestep t: (1-t)*x_now + t*x
        # t shape is (N,), need to broadcast to (N, 1, 1, 1) for broadcasting with image (N, C, H, W)
        t_broadcast = t.view(-1, 1, 1, 1)  # (N, 1, 1, 1)
        # Compute interpolated input: (1-t)*x_now + t*x
        x_interpolated = (1 - t_broadcast) * x_now +  x
        # Convert interpolated input (image format) back to patch token format (without pos_embed, will add later)
        x_now_patches = self.x_embedder(x_interpolated)
        # Concatenate patch_tokens and x_now_patches along the sequence dimension
        concatenated_input = torch.cat([patch_tokens, x_now_patches], dim=1)  # (N, 2*T, D)
        # Add position embedding for the concatenated input
        # Use the same pos_embed for both halves (or could use different embeddings)
        concatenated_input = concatenated_input + self.pos_embed
        t_emb = self.sitf1.t_embedder(t)
        y_emb = self.sitf1.y_embedder(y, self.training)
        c = t_emb + y_emb
        return self.sitf2(concatenated_input, c, t, return_act=return_act)