[refactor] move positional embeddings to patch embed layer for CogVideoX (#9263)

* remove frame limit in cogvideox

* remove debug prints

* Update src/diffusers/models/transformers/cogvideox_transformer_3d.py

* revert pipeline; remove frame limitation

* revert transformer changes

* address review comments

* add error message

* apply suggestions from review

Author: a-r-r-o-w

src/diffusers/models/embeddings.py

@@ -342,15 +342,58 @@ def __init__(
         embed_dim: int = 1920,
         text_embed_dim: int = 4096,
         bias: bool = True,
+        sample_width: int = 90,
+        sample_height: int = 60,
+        sample_frames: int = 49,
+        temporal_compression_ratio: int = 4,
+        max_text_seq_length: int = 226,
+        spatial_interpolation_scale: float = 1.875,
+        temporal_interpolation_scale: float = 1.0,
+        use_positional_embeddings: bool = True,
     ) -> None:
         super().__init__()
+
         self.patch_size = patch_size
+        self.embed_dim = embed_dim
+        self.sample_height = sample_height
+        self.sample_width = sample_width
+        self.sample_frames = sample_frames
+        self.temporal_compression_ratio = temporal_compression_ratio
+        self.max_text_seq_length = max_text_seq_length
+        self.spatial_interpolation_scale = spatial_interpolation_scale
+        self.temporal_interpolation_scale = temporal_interpolation_scale
+        self.use_positional_embeddings = use_positional_embeddings
 
         self.proj = nn.Conv2d(
             in_channels, embed_dim, kernel_size=(patch_size, patch_size), stride=patch_size, bias=bias
         )
         self.text_proj = nn.Linear(text_embed_dim, embed_dim)
 
+        if use_positional_embeddings:
+            pos_embedding = self._get_positional_embeddings(sample_height, sample_width, sample_frames)
+            self.register_buffer("pos_embedding", pos_embedding, persistent=False)
+
+    def _get_positional_embeddings(self, sample_height: int, sample_width: int, sample_frames: int) -> torch.Tensor:
+        post_patch_height = sample_height // self.patch_size
+        post_patch_width = sample_width // self.patch_size
+        post_time_compression_frames = (sample_frames - 1) // self.temporal_compression_ratio + 1
+        num_patches = post_patch_height * post_patch_width * post_time_compression_frames
+
+        pos_embedding = get_3d_sincos_pos_embed(
+            self.embed_dim,
+            (post_patch_width, post_patch_height),
+            post_time_compression_frames,
+            self.spatial_interpolation_scale,
+            self.temporal_interpolation_scale,
+        )
+        pos_embedding = torch.from_numpy(pos_embedding).flatten(0, 1)
+        joint_pos_embedding = torch.zeros(
+            1, self.max_text_seq_length + num_patches, self.embed_dim, requires_grad=False
+        )
+        joint_pos_embedding.data[:, self.max_text_seq_length :].copy_(pos_embedding)
+
+        return joint_pos_embedding
+
     def forward(self, text_embeds: torch.Tensor, image_embeds: torch.Tensor):
         r"""
         Args:
@@ -371,6 +414,21 @@ def forward(self, text_embeds: torch.Tensor, image_embeds: torch.Tensor):
         embeds = torch.cat(
             [text_embeds, image_embeds], dim=1
         ).contiguous()  # [batch, seq_length + num_frames x height x width, channels]
+
+        if self.use_positional_embeddings:
+            pre_time_compression_frames = (num_frames - 1) * self.temporal_compression_ratio + 1
+            if (
+                self.sample_height != height
+                or self.sample_width != width
+                or self.sample_frames != pre_time_compression_frames
+            ):
+                pos_embedding = self._get_positional_embeddings(height, width, pre_time_compression_frames)
+                pos_embedding = pos_embedding.to(embeds.device, dtype=embeds.dtype)
+            else:
+                pos_embedding = self.pos_embedding
+
+            embeds = embeds + pos_embedding
+
         return embeds
 
 

src/diffusers/models/transformers/cogvideox_transformer_3d.py

@@ -23,7 +23,7 @@
 from ...utils.torch_utils import maybe_allow_in_graph
 from ..attention import Attention, FeedForward
 from ..attention_processor import AttentionProcessor, CogVideoXAttnProcessor2_0, FusedCogVideoXAttnProcessor2_0
-from ..embeddings import CogVideoXPatchEmbed, TimestepEmbedding, Timesteps, get_3d_sincos_pos_embed
+from ..embeddings import CogVideoXPatchEmbed, TimestepEmbedding, Timesteps
 from ..modeling_outputs import Transformer2DModelOutput
 from ..modeling_utils import ModelMixin
 from ..normalization import AdaLayerNorm, CogVideoXLayerNormZero
@@ -239,33 +239,29 @@ def __init__(
         super().__init__()
         inner_dim = num_attention_heads * attention_head_dim
 
-        post_patch_height = sample_height // patch_size
-        post_patch_width = sample_width // patch_size
-        post_time_compression_frames = (sample_frames - 1) // temporal_compression_ratio + 1
-        self.num_patches = post_patch_height * post_patch_width * post_time_compression_frames
-
         # 1. Patch embedding
-        self.patch_embed = CogVideoXPatchEmbed(patch_size, in_channels, inner_dim, text_embed_dim, bias=True)
-        self.embedding_dropout = nn.Dropout(dropout)
-
-        # 2. 3D positional embeddings
-        spatial_pos_embedding = get_3d_sincos_pos_embed(
-            inner_dim,
-            (post_patch_width, post_patch_height),
-            post_time_compression_frames,
-            spatial_interpolation_scale,
-            temporal_interpolation_scale,
+        self.patch_embed = CogVideoXPatchEmbed(
+            patch_size=patch_size,
+            in_channels=in_channels,
+            embed_dim=inner_dim,
+            text_embed_dim=text_embed_dim,
+            bias=True,
+            sample_width=sample_width,
+            sample_height=sample_height,
+            sample_frames=sample_frames,
+            temporal_compression_ratio=temporal_compression_ratio,
+            max_text_seq_length=max_text_seq_length,
+            spatial_interpolation_scale=spatial_interpolation_scale,
+            temporal_interpolation_scale=temporal_interpolation_scale,
+            use_positional_embeddings=not use_rotary_positional_embeddings,
         )
-        spatial_pos_embedding = torch.from_numpy(spatial_pos_embedding).flatten(0, 1)
-        pos_embedding = torch.zeros(1, max_text_seq_length + self.num_patches, inner_dim, requires_grad=False)
-        pos_embedding.data[:, max_text_seq_length:].copy_(spatial_pos_embedding)
-        self.register_buffer("pos_embedding", pos_embedding, persistent=False)
+        self.embedding_dropout = nn.Dropout(dropout)
 
-        # 3. Time embeddings
+        # 2. Time embeddings
         self.time_proj = Timesteps(inner_dim, flip_sin_to_cos, freq_shift)
         self.time_embedding = TimestepEmbedding(inner_dim, time_embed_dim, timestep_activation_fn)
 
-        # 4. Define spatio-temporal transformers blocks
+        # 3. Define spatio-temporal transformers blocks
         self.transformer_blocks = nn.ModuleList(
             [
                 CogVideoXBlock(
@@ -284,7 +280,7 @@ def __init__(
         )
         self.norm_final = nn.LayerNorm(inner_dim, norm_eps, norm_elementwise_affine)
 
-        # 5. Output blocks
+        # 4. Output blocks
         self.norm_out = AdaLayerNorm(
             embedding_dim=time_embed_dim,
             output_dim=2 * inner_dim,
@@ -422,20 +418,13 @@ def forward(
 
         # 2. Patch embedding
         hidden_states = self.patch_embed(encoder_hidden_states, hidden_states)
+        hidden_states = self.embedding_dropout(hidden_states)
 
-        # 3. Position embedding
         text_seq_length = encoder_hidden_states.shape[1]
-        if not self.config.use_rotary_positional_embeddings:
-            seq_length = height * width * num_frames // (self.config.patch_size**2)
-
-            pos_embeds = self.pos_embedding[:, : text_seq_length + seq_length]
-            hidden_states = hidden_states + pos_embeds
-            hidden_states = self.embedding_dropout(hidden_states)
-
         encoder_hidden_states = hidden_states[:, :text_seq_length]
         hidden_states = hidden_states[:, text_seq_length:]
 
-        # 4. Transformer blocks
+        # 3. Transformer blocks
         for i, block in enumerate(self.transformer_blocks):
             if self.training and self.gradient_checkpointing:
 
@@ -471,11 +460,11 @@ def custom_forward(*inputs):
             hidden_states = self.norm_final(hidden_states)
             hidden_states = hidden_states[:, text_seq_length:]
 
-        # 5. Final block
+        # 4. Final block
         hidden_states = self.norm_out(hidden_states, temb=emb)
         hidden_states = self.proj_out(hidden_states)
 
-        # 6. Unpatchify
+        # 5. Unpatchify
         p = self.config.patch_size
         output = hidden_states.reshape(batch_size, num_frames, height // p, width // p, channels, p, p)
         output = output.permute(0, 1, 4, 2, 5, 3, 6).flatten(5, 6).flatten(3, 4)