wanx tiled encode

diffsynth/configs/model_config.py

@@ -54,6 +54,7 @@ from ..models.hunyuan_video_dit import HunyuanVideoDiT
 from ..models.stepvideo_vae import StepVideoVAE
 from ..models.stepvideo_dit import StepVideoModel
 
+from ..models.wanx_vae import WanXVideoVAE
 
 model_loader_configs = [
     # These configs are provided for detecting model type automatically.
@@ -108,6 +109,7 @@ model_loader_configs = [
     (None, "84ef4bd4757f60e906b54aa6a7815dc6", ["hunyuan_video_dit"], [HunyuanVideoDiT], "civitai"),
     (None, "68beaf8429b7c11aa8ca05b1bd0058bd", ["stepvideo_vae"], [StepVideoVAE], "civitai"),
     (None, "5c0216a2132b082c10cb7a0e0377e681", ["stepvideo_dit"], [StepVideoModel], "civitai"),
+    (None, "1378ea763357eea97acdef78e65d6d96", ["wanxvideo_vae"], [WanXVideoVAE], "civitai")
 ]
 huggingface_model_loader_configs = [
     # These configs are provided for detecting model type automatically.

diffsynth/models/wanx_text_encoder.py

@@ -0,0 +1,254 @@
+import math
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+
+def fp16_clamp(x):
+    if x.dtype == torch.float16 and torch.isinf(x).any():
+        clamp = torch.finfo(x.dtype).max - 1000
+        x = torch.clamp(x, min=-clamp, max=clamp)
+    return x
+
+
+class GELU(nn.Module):
+
+    def forward(self, x):
+        return 0.5 * x * (1.0 + torch.tanh(
+            math.sqrt(2.0 / math.pi) * (x + 0.044715 * torch.pow(x, 3.0))))
+
+
+class T5LayerNorm(nn.Module):
+
+    def __init__(self, dim, eps=1e-6):
+        super(T5LayerNorm, self).__init__()
+        self.dim = dim
+        self.eps = eps
+        self.weight = nn.Parameter(torch.ones(dim))
+
+    def forward(self, x):
+        x = x * torch.rsqrt(x.float().pow(2).mean(dim=-1, keepdim=True) +
+                            self.eps)
+        if self.weight.dtype in [torch.float16, torch.bfloat16]:
+            x = x.type_as(self.weight)
+        return self.weight * x
+
+
+class T5Attention(nn.Module):
+
+    def __init__(self, dim, dim_attn, num_heads, dropout=0.1):
+        assert dim_attn % num_heads == 0
+        super(T5Attention, self).__init__()
+        self.dim = dim
+        self.dim_attn = dim_attn
+        self.num_heads = num_heads
+        self.head_dim = dim_attn // num_heads
+
+        # layers
+        self.q = nn.Linear(dim, dim_attn, bias=False)
+        self.k = nn.Linear(dim, dim_attn, bias=False)
+        self.v = nn.Linear(dim, dim_attn, bias=False)
+        self.o = nn.Linear(dim_attn, dim, bias=False)
+        self.dropout = nn.Dropout(dropout)
+
+    def forward(self, x, context=None, mask=None, pos_bias=None):
+        """
+        x:          [B, L1, C].
+        context:    [B, L2, C] or None.
+        mask:       [B, L2] or [B, L1, L2] or None.
+        """
+        # check inputs
+        context = x if context is None else context
+        b, n, c = x.size(0), self.num_heads, self.head_dim
+
+        # compute query, key, value
+        q = self.q(x).view(b, -1, n, c)
+        k = self.k(context).view(b, -1, n, c)
+        v = self.v(context).view(b, -1, n, c)
+
+        # attention bias
+        attn_bias = x.new_zeros(b, n, q.size(1), k.size(1))
+        if pos_bias is not None:
+            attn_bias += pos_bias
+        if mask is not None:
+            assert mask.ndim in [2, 3]
+            mask = mask.view(b, 1, 1,
+                             -1) if mask.ndim == 2 else mask.unsqueeze(1)
+            attn_bias.masked_fill_(mask == 0, torch.finfo(x.dtype).min)
+
+        # compute attention (T5 does not use scaling)
+        attn = torch.einsum('binc,bjnc->bnij', q, k) + attn_bias
+        attn = F.softmax(attn.float(), dim=-1).type_as(attn)
+        x = torch.einsum('bnij,bjnc->binc', attn, v)
+
+        # output
+        x = x.reshape(b, -1, n * c)
+        x = self.o(x)
+        x = self.dropout(x)
+        return x
+
+
+class T5FeedForward(nn.Module):
+
+    def __init__(self, dim, dim_ffn, dropout=0.1):
+        super(T5FeedForward, self).__init__()
+        self.dim = dim
+        self.dim_ffn = dim_ffn
+
+        # layers
+        self.gate = nn.Sequential(nn.Linear(dim, dim_ffn, bias=False), GELU())
+        self.fc1 = nn.Linear(dim, dim_ffn, bias=False)
+        self.fc2 = nn.Linear(dim_ffn, dim, bias=False)
+        self.dropout = nn.Dropout(dropout)
+
+    def forward(self, x):
+        x = self.fc1(x) * self.gate(x)
+        x = self.dropout(x)
+        x = self.fc2(x)
+        x = self.dropout(x)
+        return x
+
+
+class T5SelfAttention(nn.Module):
+
+    def __init__(self,
+                 dim,
+                 dim_attn,
+                 dim_ffn,
+                 num_heads,
+                 num_buckets,
+                 shared_pos=True,
+                 dropout=0.1):
+        super(T5SelfAttention, self).__init__()
+        self.dim = dim
+        self.dim_attn = dim_attn
+        self.dim_ffn = dim_ffn
+        self.num_heads = num_heads
+        self.num_buckets = num_buckets
+        self.shared_pos = shared_pos
+
+        # layers
+        self.norm1 = T5LayerNorm(dim)
+        self.attn = T5Attention(dim, dim_attn, num_heads, dropout)
+        self.norm2 = T5LayerNorm(dim)
+        self.ffn = T5FeedForward(dim, dim_ffn, dropout)
+        self.pos_embedding = None if shared_pos else T5RelativeEmbedding(
+            num_buckets, num_heads, bidirectional=True)
+
+    def forward(self, x, mask=None, pos_bias=None):
+        e = pos_bias if self.shared_pos else self.pos_embedding(
+            x.size(1), x.size(1))
+        x = fp16_clamp(x + self.attn(self.norm1(x), mask=mask, pos_bias=e))
+        x = fp16_clamp(x + self.ffn(self.norm2(x)))
+        return x
+
+
+class T5RelativeEmbedding(nn.Module):
+
+    def __init__(self, num_buckets, num_heads, bidirectional, max_dist=128):
+        super(T5RelativeEmbedding, self).__init__()
+        self.num_buckets = num_buckets
+        self.num_heads = num_heads
+        self.bidirectional = bidirectional
+        self.max_dist = max_dist
+
+        # layers
+        self.embedding = nn.Embedding(num_buckets, num_heads)
+
+    def forward(self, lq, lk):
+        device = self.embedding.weight.device
+        # rel_pos = torch.arange(lk).unsqueeze(0).to(device) - \
+        #     torch.arange(lq).unsqueeze(1).to(device)
+        rel_pos = torch.arange(lk, device=device).unsqueeze(0) - \
+            torch.arange(lq, device=device).unsqueeze(1)
+        rel_pos = self._relative_position_bucket(rel_pos)
+        rel_pos_embeds = self.embedding(rel_pos)
+        rel_pos_embeds = rel_pos_embeds.permute(2, 0, 1).unsqueeze(
+            0)  # [1, N, Lq, Lk]
+        return rel_pos_embeds.contiguous()
+
+    def _relative_position_bucket(self, rel_pos):
+        # preprocess
+        if self.bidirectional:
+            num_buckets = self.num_buckets // 2
+            rel_buckets = (rel_pos > 0).long() * num_buckets
+            rel_pos = torch.abs(rel_pos)
+        else:
+            num_buckets = self.num_buckets
+            rel_buckets = 0
+            rel_pos = -torch.min(rel_pos, torch.zeros_like(rel_pos))
+
+        # embeddings for small and large positions
+        max_exact = num_buckets // 2
+        rel_pos_large = max_exact + (torch.log(rel_pos.float() / max_exact) /
+                                     math.log(self.max_dist / max_exact) *
+                                     (num_buckets - max_exact)).long()
+        rel_pos_large = torch.min(
+            rel_pos_large, torch.full_like(rel_pos_large, num_buckets - 1))
+        rel_buckets += torch.where(rel_pos < max_exact, rel_pos, rel_pos_large)
+        return rel_buckets
+
+def init_weights(m):
+    if isinstance(m, T5LayerNorm):
+        nn.init.ones_(m.weight)
+    elif isinstance(m, T5FeedForward):
+        nn.init.normal_(m.gate[0].weight, std=m.dim**-0.5)
+        nn.init.normal_(m.fc1.weight, std=m.dim**-0.5)
+        nn.init.normal_(m.fc2.weight, std=m.dim_ffn**-0.5)
+    elif isinstance(m, T5Attention):
+        nn.init.normal_(m.q.weight, std=(m.dim * m.dim_attn)**-0.5)
+        nn.init.normal_(m.k.weight, std=m.dim**-0.5)
+        nn.init.normal_(m.v.weight, std=m.dim**-0.5)
+        nn.init.normal_(m.o.weight, std=(m.num_heads * m.dim_attn)**-0.5)
+    elif isinstance(m, T5RelativeEmbedding):
+        nn.init.normal_(
+            m.embedding.weight, std=(2 * m.num_buckets * m.num_heads)**-0.5)
+
+
+class WanXTextEncoder(torch.nn.Module):
+
+    def __init__(self,
+                 vocab=256384,
+                 dim=4096,
+                 dim_attn=4096,
+                 dim_ffn=10240,
+                 num_heads=64,
+                 num_layers=24,
+                 num_buckets=32,
+                 shared_pos=False,
+                 dropout=0.1):
+        super(WanXTextEncoder, self).__init__()
+        self.dim = dim
+        self.dim_attn = dim_attn
+        self.dim_ffn = dim_ffn
+        self.num_heads = num_heads
+        self.num_layers = num_layers
+        self.num_buckets = num_buckets
+        self.shared_pos = shared_pos
+
+        # layers
+        self.token_embedding = vocab if isinstance(vocab, nn.Embedding) \
+            else nn.Embedding(vocab, dim)
+        self.pos_embedding = T5RelativeEmbedding(
+            num_buckets, num_heads, bidirectional=True) if shared_pos else None
+        self.dropout = nn.Dropout(dropout)
+        self.blocks = nn.ModuleList([
+            T5SelfAttention(dim, dim_attn, dim_ffn, num_heads, num_buckets,
+                            shared_pos, dropout) for _ in range(num_layers)
+        ])
+        self.norm = T5LayerNorm(dim)
+
+        # initialize weights
+        self.apply(init_weights)
+
+    def forward(self, ids, mask=None):
+        x = self.token_embedding(ids)
+        x = self.dropout(x)
+        e = self.pos_embedding(x.size(1),
+                               x.size(1)) if self.shared_pos else None
+        for block in self.blocks:
+            x = block(x, mask, pos_bias=e)
+        x = self.norm(x)
+        x = self.dropout(x)
+        return x

diffsynth/models/wanx_vae.py

@@ -0,0 +1,794 @@
+from einops import rearrange, repeat
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from tqdm import tqdm
+
+CACHE_T = 2
+
+def block_causal_mask(x, block_size):
+    # params
+    b, n, s, _, device = *x.size(), x.device
+    assert s % block_size == 0
+    num_blocks = s // block_size
+
+    # build mask
+    mask = torch.zeros(b, n, s, s, dtype=torch.bool, device=device)
+    for i in range(num_blocks):
+        mask[:, :,
+             i * block_size:(i + 1) * block_size, :(i + 1) * block_size] = 1
+    return mask
+
+
+class CausalConv3d(nn.Conv3d):
+    """
+    Causal 3d convolusion.
+    """
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self._padding = (self.padding[2], self.padding[2], self.padding[1],
+                         self.padding[1], 2 * self.padding[0], 0)
+        self.padding = (0, 0, 0)
+
+    def forward(self, x, cache_x=None):
+        padding = list(self._padding)
+        if cache_x is not None and self._padding[4] > 0:
+            cache_x = cache_x.to(x.device)
+            x = torch.cat([cache_x, x], dim=2)
+            padding[4] -= cache_x.shape[2]
+        x = F.pad(x, padding)
+
+        return super().forward(x)
+
+
+class RMS_norm(nn.Module):
+
+    def __init__(self, dim, channel_first=True, images=True, bias=False):
+        super().__init__()
+        broadcastable_dims = (1, 1, 1) if not images else (1, 1)
+        shape = (dim, *broadcastable_dims) if channel_first else (dim,)
+
+        self.channel_first = channel_first
+        self.scale = dim**0.5
+        self.gamma = nn.Parameter(torch.ones(shape))
+        self.bias = nn.Parameter(torch.zeros(shape)) if bias else 0.
+
+    def forward(self, x):
+        return F.normalize(
+            x, dim=(1 if self.channel_first else
+                    -1)) * self.scale * self.gamma + self.bias
+
+
+class Upsample(nn.Upsample):
+
+    def forward(self, x):
+        """
+        Fix bfloat16 support for nearest neighbor interpolation.
+        """
+        return super().forward(x.float()).type_as(x)
+
+
+class Resample(nn.Module):
+
+    def __init__(self, dim, mode):
+        assert mode in ('none', 'upsample2d', 'upsample3d', 'downsample2d',
+                        'downsample3d')
+        super().__init__()
+        self.dim = dim
+        self.mode = mode
+
+        # layers
+        if mode == 'upsample2d':
+            self.resample = nn.Sequential(
+                Upsample(scale_factor=(2., 2.), mode='nearest-exact'),
+                nn.Conv2d(dim, dim // 2, 3, padding=1))
+        elif mode == 'upsample3d':
+            self.resample = nn.Sequential(
+                Upsample(scale_factor=(2., 2.), mode='nearest-exact'),
+                nn.Conv2d(dim, dim // 2, 3, padding=1))
+            self.time_conv = CausalConv3d(dim,
+                                          dim * 2, (3, 1, 1),
+                                          padding=(1, 0, 0))
+
+        elif mode == 'downsample2d':
+            self.resample = nn.Sequential(
+                nn.ZeroPad2d((0, 1, 0, 1)),
+                nn.Conv2d(dim, dim, 3, stride=(2, 2)))
+        elif mode == 'downsample3d':
+            self.resample = nn.Sequential(
+                nn.ZeroPad2d((0, 1, 0, 1)),
+                nn.Conv2d(dim, dim, 3, stride=(2, 2)))
+            self.time_conv = CausalConv3d(dim,
+                                          dim, (3, 1, 1),
+                                          stride=(2, 1, 1),
+                                          padding=(0, 0, 0))
+
+        else:
+            self.resample = nn.Identity()
+
+    def forward(self, x, feat_cache=None, feat_idx=[0]):
+        b, c, t, h, w = x.size()
+        if self.mode == 'upsample3d':
+            if feat_cache is not None:
+                idx = feat_idx[0]
+                if feat_cache[idx] is None:
+                    feat_cache[idx] = 'Rep'
+                    feat_idx[0] += 1
+                else:
+
+                    cache_x = x[:, :, -CACHE_T:, :, :].clone()
+                    if cache_x.shape[2] < 2 and feat_cache[
+                            idx] is not None and feat_cache[idx] != 'Rep':
+                        # cache last frame of last two chunk
+                        cache_x = torch.cat([
+                            feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(
+                                cache_x.device), cache_x
+                        ],
+                                            dim=2)
+                    if cache_x.shape[2] < 2 and feat_cache[
+                            idx] is not None and feat_cache[idx] == 'Rep':
+                        cache_x = torch.cat([
+                            torch.zeros_like(cache_x).to(cache_x.device),
+                            cache_x
+                        ],
+                                            dim=2)
+                    if feat_cache[idx] == 'Rep':
+                        x = self.time_conv(x)
+                    else:
+                        x = self.time_conv(x, feat_cache[idx])
+                    feat_cache[idx] = cache_x
+                    feat_idx[0] += 1
+
+                    x = x.reshape(b, 2, c, t, h, w)
+                    x = torch.stack((x[:, 0, :, :, :, :], x[:, 1, :, :, :, :]),
+                                    3)
+                    x = x.reshape(b, c, t * 2, h, w)
+        t = x.shape[2]
+        x = rearrange(x, 'b c t h w -> (b t) c h w')
+        x = self.resample(x)
+        x = rearrange(x, '(b t) c h w -> b c t h w', t=t)
+
+        if self.mode == 'downsample3d':
+            if feat_cache is not None:
+                idx = feat_idx[0]
+                if feat_cache[idx] is None:
+                    feat_cache[idx] = x.clone()
+                    feat_idx[0] += 1
+                else:
+                    cache_x = x[:, :, -1:, :, :].clone()
+                    x = self.time_conv(
+                        torch.cat([feat_cache[idx][:, :, -1:, :, :], x], 2))
+                    feat_cache[idx] = cache_x
+                    feat_idx[0] += 1
+        return x
+
+    def init_weight(self, conv):
+        conv_weight = conv.weight
+        nn.init.zeros_(conv_weight)
+        c1, c2, t, h, w = conv_weight.size()
+        one_matrix = torch.eye(c1, c2)
+        init_matrix = one_matrix
+        nn.init.zeros_(conv_weight)
+        conv_weight.data[:, :, 1, 0, 0] = init_matrix
+        conv.weight.data.copy_(conv_weight)
+        nn.init.zeros_(conv.bias.data)
+
+    def init_weight2(self, conv):
+        conv_weight = conv.weight.data
+        nn.init.zeros_(conv_weight)
+        c1, c2, t, h, w = conv_weight.size()
+        init_matrix = torch.eye(c1 // 2, c2)
+        conv_weight[:c1 // 2, :, -1, 0, 0] = init_matrix
+        conv_weight[c1 // 2:, :, -1, 0, 0] = init_matrix
+        conv.weight.data.copy_(conv_weight)
+        nn.init.zeros_(conv.bias.data)
+
+
+class ResidualBlock(nn.Module):
+
+    def __init__(self, in_dim, out_dim, dropout=0.0):
+        super().__init__()
+        self.in_dim = in_dim
+        self.out_dim = out_dim
+
+        # layers
+        self.residual = nn.Sequential(
+            RMS_norm(in_dim, images=False), nn.SiLU(),
+            CausalConv3d(in_dim, out_dim, 3, padding=1),
+            RMS_norm(out_dim, images=False), nn.SiLU(), nn.Dropout(dropout),
+            CausalConv3d(out_dim, out_dim, 3, padding=1))
+        self.shortcut = CausalConv3d(in_dim, out_dim, 1) \
+            if in_dim != out_dim else nn.Identity()
+
+    def forward(self, x, feat_cache=None, feat_idx=[0]):
+        h = self.shortcut(x)
+        for layer in self.residual:
+            if isinstance(layer, CausalConv3d) and feat_cache is not None:
+                idx = feat_idx[0]
+                cache_x = x[:, :, -CACHE_T:, :, :].clone()
+                if cache_x.shape[2] < 2 and feat_cache[idx] is not None:
+                    # cache last frame of last two chunk
+                    cache_x = torch.cat([
+                        feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(
+                            cache_x.device), cache_x
+                    ],
+                                        dim=2)
+                x = layer(x, feat_cache[idx])
+                feat_cache[idx] = cache_x
+                feat_idx[0] += 1
+            else:
+                x = layer(x)
+        return x + h
+
+
+class AttentionBlock(nn.Module):
+    """
+    Causal self-attention with a single head.
+    """
+
+    def __init__(self, dim):
+        super().__init__()
+        self.dim = dim
+
+        # layers
+        self.norm = RMS_norm(dim)
+        self.to_qkv = nn.Conv2d(dim, dim * 3, 1)
+        self.proj = nn.Conv2d(dim, dim, 1)
+
+        # zero out the last layer params
+        nn.init.zeros_(self.proj.weight)
+
+    def forward(self, x):
+        identity = x
+        b, c, t, h, w = x.size()
+        x = rearrange(x, 'b c t h w -> (b t) c h w')
+        x = self.norm(x)
+        # compute query, key, value
+        q, k, v = self.to_qkv(x).reshape(b * t, 1, c * 3, -1).permute(
+            0, 1, 3, 2).contiguous().chunk(3, dim=-1)
+
+        # apply attention
+        x = F.scaled_dot_product_attention(
+            q,
+            k,
+            v,
+            #attn_mask=block_causal_mask(q, block_size=h * w)
+        )
+        x = x.squeeze(1).permute(0, 2, 1).reshape(b * t, c, h, w)
+
+        # output
+        x = self.proj(x)
+        x = rearrange(x, '(b t) c h w-> b c t h w', t=t)
+        return x + identity
+
+
+class Encoder3d(nn.Module):
+
+    def __init__(self,
+                 dim=128,
+                 z_dim=4,
+                 dim_mult=[1, 2, 4, 4],
+                 num_res_blocks=2,
+                 attn_scales=[],
+                 temperal_downsample=[True, True, False],
+                 dropout=0.0):
+        super().__init__()
+        self.dim = dim
+        self.z_dim = z_dim
+        self.dim_mult = dim_mult
+        self.num_res_blocks = num_res_blocks
+        self.attn_scales = attn_scales
+        self.temperal_downsample = temperal_downsample
+
+        # dimensions
+        dims = [dim * u for u in [1] + dim_mult]
+        scale = 1.0
+
+        # init block
+        self.conv1 = CausalConv3d(3, dims[0], 3, padding=1)
+
+        # downsample blocks
+        downsamples = []
+        for i, (in_dim, out_dim) in enumerate(zip(dims[:-1], dims[1:])):
+            # residual (+attention) blocks
+            for _ in range(num_res_blocks):
+                downsamples.append(ResidualBlock(in_dim, out_dim, dropout))
+                if scale in attn_scales:
+                    downsamples.append(AttentionBlock(out_dim))
+                in_dim = out_dim
+
+            # downsample block
+            if i != len(dim_mult) - 1:
+                mode = 'downsample3d' if temperal_downsample[
+                    i] else 'downsample2d'
+                downsamples.append(Resample(out_dim, mode=mode))
+                scale /= 2.0
+        self.downsamples = nn.Sequential(*downsamples)
+
+        # middle blocks
+        self.middle = nn.Sequential(ResidualBlock(out_dim, out_dim, dropout),
+                                    AttentionBlock(out_dim),
+                                    ResidualBlock(out_dim, out_dim, dropout))
+
+        # output blocks
+        self.head = nn.Sequential(RMS_norm(out_dim, images=False), nn.SiLU(),
+                                  CausalConv3d(out_dim, z_dim, 3, padding=1))
+
+    def forward(self, x, feat_cache=None, feat_idx=[0]):
+        if feat_cache is not None:
+            idx = feat_idx[0]
+            cache_x = x[:, :, -CACHE_T:, :, :].clone()
+            if cache_x.shape[2] < 2 and feat_cache[idx] is not None:
+                # cache last frame of last two chunk
+                cache_x = torch.cat([
+                    feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(
+                        cache_x.device), cache_x
+                ],
+                                    dim=2)
+            x = self.conv1(x, feat_cache[idx])
+            feat_cache[idx] = cache_x
+            feat_idx[0] += 1
+        else:
+            x = self.conv1(x)
+
+        ## downsamples
+        for layer in self.downsamples:
+            if feat_cache is not None:
+                x = layer(x, feat_cache, feat_idx)
+            else:
+                x = layer(x)
+
+        ## middle
+        for layer in self.middle:
+            if isinstance(layer, ResidualBlock) and feat_cache is not None:
+                x = layer(x, feat_cache, feat_idx)
+            else:
+                x = layer(x)
+
+        ## head
+        for layer in self.head:
+            if isinstance(layer, CausalConv3d) and feat_cache is not None:
+                idx = feat_idx[0]
+                cache_x = x[:, :, -CACHE_T:, :, :].clone()
+                if cache_x.shape[2] < 2 and feat_cache[idx] is not None:
+                    # cache last frame of last two chunk
+                    cache_x = torch.cat([
+                        feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(
+                            cache_x.device), cache_x
+                    ],
+                                        dim=2)
+                x = layer(x, feat_cache[idx])
+                feat_cache[idx] = cache_x
+                feat_idx[0] += 1
+            else:
+                x = layer(x)
+        return x
+
+
+class Decoder3d(nn.Module):
+
+    def __init__(self,
+                 dim=128,
+                 z_dim=4,
+                 dim_mult=[1, 2, 4, 4],
+                 num_res_blocks=2,
+                 attn_scales=[],
+                 temperal_upsample=[False, True, True],
+                 dropout=0.0):
+        super().__init__()
+        self.dim = dim
+        self.z_dim = z_dim
+        self.dim_mult = dim_mult
+        self.num_res_blocks = num_res_blocks
+        self.attn_scales = attn_scales
+        self.temperal_upsample = temperal_upsample
+
+        # dimensions
+        dims = [dim * u for u in [dim_mult[-1]] + dim_mult[::-1]]
+        scale = 1.0 / 2**(len(dim_mult) - 2)
+
+        # init block
+        self.conv1 = CausalConv3d(z_dim, dims[0], 3, padding=1)
+
+        # middle blocks
+        self.middle = nn.Sequential(ResidualBlock(dims[0], dims[0], dropout),
+                                    AttentionBlock(dims[0]),
+                                    ResidualBlock(dims[0], dims[0], dropout))
+
+        # upsample blocks
+        upsamples = []
+        for i, (in_dim, out_dim) in enumerate(zip(dims[:-1], dims[1:])):
+            # residual (+attention) blocks
+            if i == 1 or i == 2 or i == 3:
+                in_dim = in_dim // 2
+            for _ in range(num_res_blocks + 1):
+                upsamples.append(ResidualBlock(in_dim, out_dim, dropout))
+                if scale in attn_scales:
+                    upsamples.append(AttentionBlock(out_dim))
+                in_dim = out_dim
+
+            # upsample block
+            if i != len(dim_mult) - 1:
+                mode = 'upsample3d' if temperal_upsample[i] else 'upsample2d'
+                upsamples.append(Resample(out_dim, mode=mode))
+                scale *= 2.0
+        self.upsamples = nn.Sequential(*upsamples)
+
+        # output blocks
+        self.head = nn.Sequential(RMS_norm(out_dim, images=False), nn.SiLU(),
+                                  CausalConv3d(out_dim, 3, 3, padding=1))
+
+    def forward(self, x, feat_cache=None, feat_idx=[0]):
+        ## conv1
+        if feat_cache is not None:
+            idx = feat_idx[0]
+            cache_x = x[:, :, -CACHE_T:, :, :].clone()
+            if cache_x.shape[2] < 2 and feat_cache[idx] is not None:
+                # cache last frame of last two chunk
+                cache_x = torch.cat([
+                    feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(
+                        cache_x.device), cache_x
+                ],
+                                    dim=2)
+            x = self.conv1(x, feat_cache[idx])
+            feat_cache[idx] = cache_x
+            feat_idx[0] += 1
+        else:
+            x = self.conv1(x)
+
+        ## middle
+        for layer in self.middle:
+            if isinstance(layer, ResidualBlock) and feat_cache is not None:
+                x = layer(x, feat_cache, feat_idx)
+            else:
+                x = layer(x)
+
+        ## upsamples
+        for layer in self.upsamples:
+            if feat_cache is not None:
+                x = layer(x, feat_cache, feat_idx)
+            else:
+                x = layer(x)
+
+        ## head
+        for layer in self.head:
+            if isinstance(layer, CausalConv3d) and feat_cache is not None:
+                idx = feat_idx[0]
+                cache_x = x[:, :, -CACHE_T:, :, :].clone()
+                if cache_x.shape[2] < 2 and feat_cache[idx] is not None:
+                    # cache last frame of last two chunk
+                    cache_x = torch.cat([
+                        feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(
+                            cache_x.device), cache_x
+                    ],
+                                        dim=2)
+                x = layer(x, feat_cache[idx])
+                feat_cache[idx] = cache_x
+                feat_idx[0] += 1
+            else:
+                x = layer(x)
+        return x
+
+
+def count_conv3d(model):
+    count = 0
+    for m in model.modules():
+        if isinstance(m, CausalConv3d):
+            count += 1
+    return count
+
+
+class VideoVAE_(nn.Module):
+
+    def __init__(self,
+                 dim=96,
+                 z_dim=16,
+                 dim_mult=[1, 2, 4, 4],
+                 num_res_blocks=2,
+                 attn_scales=[],
+                 temperal_downsample=[False, True, True],
+                 dropout=0.0):
+        super().__init__()
+        self.dim = dim
+        self.z_dim = z_dim
+        self.dim_mult = dim_mult
+        self.num_res_blocks = num_res_blocks
+        self.attn_scales = attn_scales
+        self.temperal_downsample = temperal_downsample
+        self.temperal_upsample = temperal_downsample[::-1]
+
+        # modules
+        self.encoder = Encoder3d(dim, z_dim * 2, dim_mult, num_res_blocks,
+                                 attn_scales, self.temperal_downsample, dropout)
+        self.conv1 = CausalConv3d(z_dim * 2, z_dim * 2, 1)
+        self.conv2 = CausalConv3d(z_dim, z_dim, 1)
+        self.decoder = Decoder3d(dim, z_dim, dim_mult, num_res_blocks,
+                                 attn_scales, self.temperal_upsample, dropout)
+
+    def forward(self, x):
+        mu, log_var = self.encode(x)
+        z = self.reparameterize(mu, log_var)
+        x_recon = self.decode(z)
+        return x_recon, mu, log_var
+
+    def encode(self, x, scale):
+        self.clear_cache()
+        ## cache
+        t = x.shape[2]
+        iter_ = 1 + (t - 1) // 4
+
+        for i in range(iter_):
+            self._enc_conv_idx = [0]
+            if i == 0:
+                out = self.encoder(x[:, :, :1, :, :],
+                                   feat_cache=self._enc_feat_map,
+                                   feat_idx=self._enc_conv_idx)
+            else:
+                out_ = self.encoder(x[:, :, 1 + 4 * (i - 1):1 + 4 * i, :, :],
+                                    feat_cache=self._enc_feat_map,
+                                    feat_idx=self._enc_conv_idx)
+                out = torch.cat([out, out_], 2)
+        mu, log_var = self.conv1(out).chunk(2, dim=1)
+        if isinstance(scale[0], torch.Tensor):
+            scale = [s.to(dtype=mu.dtype, device=mu.device) for s in scale]
+            mu = (mu - scale[0].view(1, self.z_dim, 1, 1, 1)) * scale[1].view(
+                1, self.z_dim, 1, 1, 1)
+        else:
+            scale = scale.to(dtype=mu.dtype, device=mu.device)
+            mu = (mu - scale[0]) * scale[1]
+        return mu
+
+    def decode(self, z, scale):
+        self.clear_cache()
+        # z: [b,c,t,h,w]
+        if isinstance(scale[0], torch.Tensor):
+            scale = [s.to(dtype=z.dtype, device=z.device) for s in scale]
+            z = z / scale[1].view(1, self.z_dim, 1, 1, 1) + scale[0].view(
+                1, self.z_dim, 1, 1, 1)
+        else:
+            scale = scale.to(dtype=z.dtype, device=z.device)
+            z = z / scale[1] + scale[0]
+        iter_ = z.shape[2]
+        x = self.conv2(z)
+        for i in range(iter_):
+            self._conv_idx = [0]
+            if i == 0:
+                out = self.decoder(x[:, :, i:i + 1, :, :],
+                                   feat_cache=self._feat_map,
+                                   feat_idx=self._conv_idx)
+            else:
+                out_ = self.decoder(x[:, :, i:i + 1, :, :],
+                                    feat_cache=self._feat_map,
+                                    feat_idx=self._conv_idx)
+                out = torch.cat([out, out_], 2) # may add tensor offload
+        return out
+
+    def reparameterize(self, mu, log_var):
+        std = torch.exp(0.5 * log_var)
+        eps = torch.randn_like(std)
+        return eps * std + mu
+
+    def sample(self, imgs, deterministic=False):
+        mu, log_var = self.encode(imgs)
+        if deterministic:
+            return mu
+        std = torch.exp(0.5 * log_var.clamp(-30.0, 20.0))
+        return mu + std * torch.randn_like(std)
+
+    def clear_cache(self):
+        self._conv_num = count_conv3d(self.decoder)
+        self._conv_idx = [0]
+        self._feat_map = [None] * self._conv_num
+        # cache encode
+        self._enc_conv_num = count_conv3d(self.encoder)
+        self._enc_conv_idx = [0]
+        self._enc_feat_map = [None] * self._enc_conv_num
+
+
+class WanXVideoVAE(nn.Module):
+
+    def __init__(self, z_dim=16):
+        super().__init__()
+
+        mean = [
+            -0.7571, -0.7089, -0.9113, 0.1075, -0.1745, 0.9653, -0.1517, 1.5508,
+            0.4134, -0.0715, 0.5517, -0.3632, -0.1922, -0.9497, 0.2503, -0.2921
+        ]
+        std = [
+            2.8184, 1.4541, 2.3275, 2.6558, 1.2196, 1.7708, 2.6052, 2.0743,
+            3.2687, 2.1526, 2.8652, 1.5579, 1.6382, 1.1253, 2.8251, 1.9160
+        ]
+        self.mean = torch.tensor(mean)
+        self.std = torch.tensor(std)
+        self.scale = [self.mean, 1.0 / self.std]
+
+        # init model
+        self.model = VideoVAE_(z_dim=z_dim).eval().requires_grad_(False)
+        self.upsampling_factor = 8
+
+
+    def build_1d_mask(self, length, left_bound, right_bound, border_width):
+        x = torch.ones((length,))
+        if not left_bound:
+            x[:border_width] = (torch.arange(border_width) + 1) / border_width
+        if not right_bound:
+            x[-border_width:] = torch.flip((torch.arange(border_width) + 1) / border_width, dims=(0,))
+        return x
+
+
+    def build_mask(self, data, is_bound, border_width):
+        _, _, _, H, W = data.shape
+        h = self.build_1d_mask(H, is_bound[0], is_bound[1], border_width[0])
+        w = self.build_1d_mask(W, is_bound[2], is_bound[3], border_width[1])
+
+        h = repeat(h, "H -> H W", H=H, W=W)
+        w = repeat(w, "W -> H W", H=H, W=W)
+
+        mask = torch.stack([h, w]).min(dim=0).values
+        mask = rearrange(mask, "H W -> 1 1 1 H W")
+        return mask
+
+
+    def tiled_decode(self, hidden_states, device, tile_size, tile_stride):
+        _, _, T, H, W = hidden_states.shape
+        size_h, size_w = tile_size
+        stride_h, stride_w = tile_stride
+
+        # Split tasks
+        tasks = []
+        for h in range(0, H, stride_h):
+            if (h-stride_h >= 0 and h-stride_h+size_h >= H): continue
+            for w in range(0, W, stride_w):
+                if (w-stride_w >= 0 and w-stride_w+size_w >= W): continue
+                h_, w_ = h + size_h, w + size_w
+                tasks.append((h, h_, w, w_))
+
+        data_device = "cpu"
+        computation_device = device
+
+        out_T = T * 4 - 3
+        weight = torch.zeros((1, 1, out_T, H * self.upsampling_factor, W * self.upsampling_factor), dtype=hidden_states.dtype, device=data_device)
+        values = torch.zeros((1, 3, out_T, H * self.upsampling_factor, W * self.upsampling_factor), dtype=hidden_states.dtype, device=data_device)
+
+        for h, h_, w, w_ in tqdm(tasks, desc="VAE decoding"):
+            hidden_states_batch = hidden_states[:, :, :, h:h_, w:w_].to(computation_device)
+            hidden_states_batch = self.model.decode(hidden_states_batch, self.scale).to(data_device)
+
+            mask = self.build_mask(
+                hidden_states_batch,
+                is_bound=(h==0, h_>=H, w==0, w_>=W),
+                border_width=((size_h - stride_h) * self.upsampling_factor, (size_w - stride_w) * self.upsampling_factor)
+            ).to(dtype=hidden_states.dtype, device=data_device)
+
+            target_h = h * self.upsampling_factor
+            target_w = w * self.upsampling_factor
+            values[
+                :,
+                :,
+                :,
+                target_h:target_h + hidden_states_batch.shape[3],
+                target_w:target_w + hidden_states_batch.shape[4],
+            ] += hidden_states_batch * mask
+            weight[
+                :,
+                :,
+                :,
+                target_h: target_h + hidden_states_batch.shape[3],
+                target_w: target_w + hidden_states_batch.shape[4],
+            ] += mask
+        values = values / weight
+        values = values.float().clamp_(-1, 1)
+        return values
+
+
+    def tiled_encode(self, video, device, tile_size, tile_stride):
+        _, _, T, H, W = video.shape
+        size_h, size_w = tile_size
+        stride_h, stride_w = tile_stride
+
+        # Split tasks
+        tasks = []
+        for h in range(0, H, stride_h):
+            if (h-stride_h >= 0 and h-stride_h+size_h >= H): continue
+            for w in range(0, W, stride_w):
+                if (w-stride_w >= 0 and w-stride_w+size_w >= W): continue
+                h_, w_ = h + size_h, w + size_w
+                tasks.append((h, h_, w, w_))
+
+        data_device = "cpu"
+        computation_device = device
+
+        out_T = (T + 3) // 4
+        weight = torch.zeros((1, 1, out_T, H // self.upsampling_factor, W // self.upsampling_factor), dtype=video.dtype, device=data_device)
+        values = torch.zeros((1, 16, out_T, H // self.upsampling_factor, W // self.upsampling_factor), dtype=video.dtype, device=data_device)
+
+        for h, h_, w, w_ in tqdm(tasks, desc="VAE encoding"):
+            hidden_states_batch = video[:, :, :, h:h_, w:w_].to(computation_device)
+            hidden_states_batch = self.model.encode(hidden_states_batch, self.scale).to(data_device)
+
+            mask = self.build_mask(
+                hidden_states_batch,
+                is_bound=(h==0, h_>=H, w==0, w_>=W),
+                border_width=((size_h - stride_h) // self.upsampling_factor, (size_w - stride_w) // self.upsampling_factor)
+            ).to(dtype=video.dtype, device=data_device)
+
+            target_h = h // self.upsampling_factor
+            target_w = w // self.upsampling_factor
+            values[
+                :,
+                :,
+                :,
+                target_h:target_h + hidden_states_batch.shape[3],
+                target_w:target_w + hidden_states_batch.shape[4],
+            ] += hidden_states_batch * mask
+            weight[
+                :,
+                :,
+                :,
+                target_h: target_h + hidden_states_batch.shape[3],
+                target_w: target_w + hidden_states_batch.shape[4],
+            ] += mask
+        values = values / weight
+        values = values.float()
+        return values
+
+
+    def single_encode(self, video, device):
+        video = video.to(device)
+        x = self.model.encode(video, self.scale)
+        return x.float()
+
+
+    def single_decode(self, hidden_state, device):
+        hidden_state = hidden_state.to(device)
+        video = self.model.decode(hidden_state, self.scale)
+        return video.float().clamp_(-1, 1)
+
+
+    def encode(self, videos, device, tiled=False, tile_size=(272, 272), tile_stride=(144, 128)):
+
+        videos = [video.to("cpu") for video in videos]
+        hidden_states = []
+        for video in videos:
+            video = video.unsqueeze(0)
+            if tiled:
+                assert tile_size[0] % self.upsampling_factor == 0 and tile_size[1] % self.upsampling_factor == 0, f"tile_size must be devisible by {self.upsampling_factor}"
+                hidden_state = self.tiled_encode(video, device, tile_size, tile_stride)
+            else:
+                hidden_state = self.single_encode(video, device)
+            hidden_state = hidden_state.squeeze(0)
+            hidden_states.append(hidden_state)
+        return hidden_states
+
+
+    def decode(self, hidden_states, device, tiled=False, tile_size=(34, 34), tile_stride=(18, 16)):
+        hidden_states = [hidden_state.to("cpu") for hidden_state in hidden_states]
+        videos = []
+        for hidden_state in hidden_states:
+            hidden_state = hidden_state.unsqueeze(0)
+            if tiled:
+                video = self.tiled_decode(hidden_state, device, tile_size, tile_stride)
+            else:
+                video = self.single_decode(hidden_state, device)
+            video = video.squeeze(0)
+            videos.append(video)
+        return videos
+
+
+    @staticmethod
+    def state_dict_converter():
+        return WanXVideoVAEStateDictConverter()
+
+
+class WanXVideoVAEStateDictConverter:
+
+    def __init__(self):
+        pass
+
+    def from_civitai(self, state_dict):
+        state_dict_ = {}
+        for name in state_dict['model_state']:
+            state_dict_['model.' + name] = state_dict['model_state'][name]
+        return state_dict_

diffsynth/prompters/__init__.py

@@ -9,3 +9,4 @@ from .omost import OmostPromter
 from .cog_prompter import CogPrompter
 from .hunyuan_video_prompter import HunyuanVideoPrompter
 from .stepvideo_prompter import StepVideoPrompter
+from .wanx_prompter import WanXPrompter

diffsynth/prompters/wanx_prompter.py

@@ -0,0 +1,103 @@
+from .base_prompter import BasePrompter
+from ..models.wanx_text_encoder import WanXTextEncoder
+from transformers import AutoTokenizer
+import os, torch
+import ftfy
+import html
+import string
+
+import regex as re
+
+
+def basic_clean(text):
+    text = ftfy.fix_text(text)
+    text = html.unescape(html.unescape(text))
+    return text.strip()
+
+def whitespace_clean(text):
+    text = re.sub(r'\s+', ' ', text)
+    text = text.strip()
+    return text
+
+def canonicalize(text, keep_punctuation_exact_string=None):
+    text = text.replace('_', ' ')
+    if keep_punctuation_exact_string:
+        text = keep_punctuation_exact_string.join(
+            part.translate(str.maketrans('', '', string.punctuation))
+            for part in text.split(keep_punctuation_exact_string))
+    else:
+        text = text.translate(str.maketrans('', '', string.punctuation))
+    text = text.lower()
+    text = re.sub(r'\s+', ' ', text)
+    return text.strip()
+
+class HuggingfaceTokenizer:
+
+    def __init__(self, name, seq_len=None, clean=None, **kwargs):
+        assert clean in (None, 'whitespace', 'lower', 'canonicalize')
+        self.name = name
+        self.seq_len = seq_len
+        self.clean = clean
+
+        # init tokenizer
+        self.tokenizer = AutoTokenizer.from_pretrained(name, **kwargs)
+        self.vocab_size = self.tokenizer.vocab_size
+
+    def __call__(self, sequence, **kwargs):
+        return_mask = kwargs.pop('return_mask', False)
+
+        # arguments
+        _kwargs = {'return_tensors': 'pt'}
+        if self.seq_len is not None:
+            _kwargs.update({
+                'padding': 'max_length',
+                'truncation': True,
+                'max_length': self.seq_len
+            })
+        _kwargs.update(**kwargs)
+
+        # tokenization
+        if isinstance(sequence, str):
+            sequence = [sequence]
+        if self.clean:
+            sequence = [self._clean(u) for u in sequence]
+        ids = self.tokenizer(sequence, **_kwargs)
+
+        # output
+        if return_mask:
+            return ids.input_ids, ids.attention_mask
+        else:
+            return ids.input_ids
+
+    def _clean(self, text):
+        if self.clean == 'whitespace':
+            text = whitespace_clean(basic_clean(text))
+        elif self.clean == 'lower':
+            text = whitespace_clean(basic_clean(text)).lower()
+        elif self.clean == 'canonicalize':
+            text = canonicalize(basic_clean(text))
+        return text
+
+class WanXPrompter(BasePrompter):
+
+    def __init__(self, tokenizer_path=None, text_len=512):
+        if tokenizer_path is None:
+            base_path = os.path.dirname(os.path.dirname(__file__))
+            tokenizer_path = os.path.join(
+                base_path, "tokenizer_configs/hunyuan_dit/tokenizer")
+        super().__init__()
+        self.tokenizer = HuggingfaceTokenizer(name=tokenizer_path, seq_len=text_len, clean='whitespace')
+        self.text_encoder = None
+        
+    def fetch_models(self, text_encoder: WanXTextEncoder = None):
+        self.text_encoder = text_encoder
+
+    def encode_prompt(self, prompt, device="cuda"):
+        ids, mask = self.tokenizer(prompt, return_mask=True, add_special_tokens=True)
+        ids = ids.to(device)
+        mask = mask.to(device)
+        seq_lens = mask.gt(0).sum(dim=1).long()
+        prompt_emb = self.text_encoder(ids, mask)
+        prompt_emb = [u[:v] for u, v in zip(prompt_emb, seq_lens)]
+        return prompt_emb
+

examples/WanX/test_prompter.py

@@ -0,0 +1,18 @@
+import torch
+from diffsynth.prompters import WanXPrompter
+from diffsynth.models.wanx_text_encoder import WanXTextEncoder
+
+prompter = WanXPrompter('models/WanX/google/umt5-xxl')
+text_encoder = WanXTextEncoder()
+text_encoder.load_state_dict(torch.load('models/WanX/models_t5_umt5-xxl-enc-bf16.pth', map_location='cpu'))
+text_encoder = text_encoder.eval().requires_grad_(False).to(dtype=torch.bfloat16, device='cuda')
+
+prompter.fetch_models(text_encoder)
+
+prompt = '维京战士双手挥舞着大斧，对抗猛犸象，黄昏，雪地中，漫天飞雪'
+neg_prompt = '色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走'
+
+prompt_emb = prompter.encode_prompt(prompt)
+neg_prompt_emb = prompter.encode_prompt(neg_prompt)
+print(prompt_emb[0]) # torch.Size([31, 4096])
+print(neg_prompt_emb[0]) # torch.Size([126, 4096])

examples/WanX/test_vae.py

@@ -0,0 +1,46 @@
+import torch
+import torchvision
+import imageio
+from diffsynth import ModelManager
+
+def save_video(tensor,
+                save_file=None,
+                fps=30,
+                nrow=8,
+                normalize=True,
+                value_range=(-1, 1)):
+
+    tensor = tensor.clamp(min(value_range), max(value_range))
+    tensor = torch.stack([
+        torchvision.utils.make_grid(
+            u, nrow=nrow, normalize=normalize, value_range=value_range)
+        for u in tensor.unbind(2)
+    ],
+                         dim=1).permute(1, 2, 3, 0) #frame, h, w, 3
+    tensor = (tensor * 255).type(torch.uint8).cpu()
+
+    # write video
+    writer = imageio.get_writer(
+        save_file, fps=fps, codec='libx264', quality=8)
+    for frame in tensor.numpy():
+        writer.append_data(frame)
+    writer.close()
+
+torch.cuda.memory._record_memory_history()
+
+model_manager = ModelManager(torch_dtype=torch.float, device="cuda")
+model_manager.load_models([
+    "models/WanX/vae.pth",
+])
+
+vae = model_manager.fetch_model('wanxvideo_vae')
+
+latents = [torch.load('sample.pt')]
+videos = vae.decode(latents, device=latents[0].device, tiled=True)
+back_encode = vae.encode(videos, device=latents[0].device, tiled=True)
+
+videos_back_encode = vae.decode(back_encode, device=latents[0].device, tiled=False)
+torch.cuda.memory._dump_snapshot("my_snapshot.pickle")
+
+save_video(videos[0][None], save_file='example.mp4', fps=16, nrow=1)
+save_video(videos_back_encode[0][None], save_file='example_backencode.mp4', fps=16, nrow=1)