cogvideox_transformer_3d.py

# Copyright 2024 The CogVideoX team, Tsinghua University & ZhipuAI and The HuggingFace Team.
# All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

from typing import Any, Dict, Optional, Union

import torch
from torch import nn

from ...configuration_utils import ConfigMixin, register_to_config
from ...utils import is_torch_version, logging
from ...utils.torch_utils import maybe_allow_in_graph
from ..attention import Attention, FeedForward
from ..embeddings import CogVideoXPatchEmbed, TimestepEmbedding, Timesteps, get_3d_sincos_pos_embed
from ..modeling_outputs import Transformer2DModelOutput
from ..modeling_utils import ModelMixin
from ..normalization import AdaLayerNorm, CogVideoXLayerNormZero


logger = logging.get_logger(__name__)  # pylint: disable=invalid-name


@maybe_allow_in_graph
class CogVideoXBlock(nn.Module):
    r"""
    Transformer block used in [CogVideoX](https://github.com/THUDM/CogVideo) model.

    Parameters:
        dim (`int`):
            The number of channels in the input and output.
        num_attention_heads (`int`):
            The number of heads to use for multi-head attention.
        attention_head_dim (`int`):
            The number of channels in each head.
        time_embed_dim (`int`):
            The number of channels in timestep embedding.
        dropout (`float`, defaults to `0.0`):
            The dropout probability to use.
        activation_fn (`str`, defaults to `"gelu-approximate"`):
            Activation function to be used in feed-forward.
        attention_bias (`bool`, defaults to `False`):
            Whether or not to use bias in attention projection layers.
        qk_norm (`bool`, defaults to `True`):
            Whether or not to use normalization after query and key projections in Attention.
        norm_elementwise_affine (`bool`, defaults to `True`):
            Whether to use learnable elementwise affine parameters for normalization.
        norm_eps (`float`, defaults to `1e-5`):
            Epsilon value for normalization layers.
        final_dropout (`bool` defaults to `False`):
            Whether to apply a final dropout after the last feed-forward layer.
        ff_inner_dim (`int`, *optional*, defaults to `None`):
            Custom hidden dimension of Feed-forward layer. If not provided, `4 * dim` is used.
        ff_bias (`bool`, defaults to `True`):
            Whether or not to use bias in Feed-forward layer.
        attention_out_bias (`bool`, defaults to `True`):
            Whether or not to use bias in Attention output projection layer.
    """

    def __init__(
        self,
        dim: int,
        num_attention_heads: int,
        attention_head_dim: int,
        time_embed_dim: int,
        dropout: float = 0.0,
        activation_fn: str = "gelu-approximate",
        attention_bias: bool = False,
        qk_norm: bool = True,
        norm_elementwise_affine: bool = True,
        norm_eps: float = 1e-5,
        final_dropout: bool = True,
        ff_inner_dim: Optional[int] = None,
        ff_bias: bool = True,
        attention_out_bias: bool = True,
    ):
        super().__init__()

        # 1. Self Attention
        self.norm1 = CogVideoXLayerNormZero(time_embed_dim, dim, norm_elementwise_affine, norm_eps, bias=True)

        self.attn1 = Attention(
            query_dim=dim,
            dim_head=attention_head_dim,
            heads=num_attention_heads,
            qk_norm="layer_norm" if qk_norm else None,
            eps=1e-6,
            bias=attention_bias,
            out_bias=attention_out_bias,
        )

        # 2. Feed Forward
        self.norm2 = CogVideoXLayerNormZero(time_embed_dim, dim, norm_elementwise_affine, norm_eps, bias=True)

        self.ff = FeedForward(
            dim,
            dropout=dropout,
            activation_fn=activation_fn,
            final_dropout=final_dropout,
            inner_dim=ff_inner_dim,
            bias=ff_bias,
        )

    def forward(
        self,
        hidden_states: torch.Tensor,
        encoder_hidden_states: torch.Tensor,
        temb: torch.Tensor,
    ) -> torch.Tensor:
        norm_hidden_states, norm_encoder_hidden_states, gate_msa, enc_gate_msa = self.norm1(
            hidden_states, encoder_hidden_states, temb
        )

        # attention
        text_length = norm_encoder_hidden_states.size(1)

        # CogVideoX uses concatenated text + video embeddings with self-attention instead of using
        # them in cross-attention individually
        norm_hidden_states = torch.cat([norm_encoder_hidden_states, norm_hidden_states], dim=1)
        attn_output = self.attn1(
            hidden_states=norm_hidden_states,
            encoder_hidden_states=None,
        )

        hidden_states = hidden_states + gate_msa * attn_output[:, text_length:]
        encoder_hidden_states = encoder_hidden_states + enc_gate_msa * attn_output[:, :text_length]

        # norm & modulate
        norm_hidden_states, norm_encoder_hidden_states, gate_ff, enc_gate_ff = self.norm2(
            hidden_states, encoder_hidden_states, temb
        )

        # feed-forward
        norm_hidden_states = torch.cat([norm_encoder_hidden_states, norm_hidden_states], dim=1)
        ff_output = self.ff(norm_hidden_states)

        hidden_states = hidden_states + gate_ff * ff_output[:, text_length:]
        encoder_hidden_states = encoder_hidden_states + enc_gate_ff * ff_output[:, :text_length]
        return hidden_states, encoder_hidden_states


class CogVideoXTransformer3DModel(ModelMixin, ConfigMixin):
    """
    A Transformer model for video-like data in [CogVideoX](https://github.com/THUDM/CogVideo).

    Parameters:
        num_attention_heads (`int`, defaults to `30`):
            The number of heads to use for multi-head attention.
        attention_head_dim (`int`, defaults to `64`):
            The number of channels in each head.
        in_channels (`int`, defaults to `16`):
            The number of channels in the input.
        out_channels (`int`, *optional*, defaults to `16`):
            The number of channels in the output.
        flip_sin_to_cos (`bool`, defaults to `True`):
            Whether to flip the sin to cos in the time embedding.
        time_embed_dim (`int`, defaults to `512`):
            Output dimension of timestep embeddings.
        text_embed_dim (`int`, defaults to `4096`):
            Input dimension of text embeddings from the text encoder.
        num_layers (`int`, defaults to `30`):
            The number of layers of Transformer blocks to use.
        dropout (`float`, defaults to `0.0`):
            The dropout probability to use.
        attention_bias (`bool`, defaults to `True`):
            Whether or not to use bias in the attention projection layers.
        sample_width (`int`, defaults to `90`):
            The width of the input latents.
        sample_height (`int`, defaults to `60`):
            The height of the input latents.
        sample_frames (`int`, defaults to `49`):
            The number of frames in the input latents. Note that this parameter was incorrectly initialized to 49
            instead of 13 because CogVideoX processed 13 latent frames at once in its default and recommended settings,
            but cannot be changed to the correct value to ensure backwards compatibility. To create a transformer with
            K latent frames, the correct value to pass here would be: ((K - 1) * temporal_compression_ratio + 1).
        patch_size (`int`, defaults to `2`):
            The size of the patches to use in the patch embedding layer.
        temporal_compression_ratio (`int`, defaults to `4`):
            The compression ratio across the temporal dimension. See documentation for `sample_frames`.
        max_text_seq_length (`int`, defaults to `226`):
            The maximum sequence length of the input text embeddings.
        activation_fn (`str`, defaults to `"gelu-approximate"`):
            Activation function to use in feed-forward.
        timestep_activation_fn (`str`, defaults to `"silu"`):
            Activation function to use when generating the timestep embeddings.
        norm_elementwise_affine (`bool`, defaults to `True`):
            Whether or not to use elementwise affine in normalization layers.
        norm_eps (`float`, defaults to `1e-5`):
            The epsilon value to use in normalization layers.
        spatial_interpolation_scale (`float`, defaults to `1.875`):
            Scaling factor to apply in 3D positional embeddings across spatial dimensions.
        temporal_interpolation_scale (`float`, defaults to `1.0`):
            Scaling factor to apply in 3D positional embeddings across temporal dimensions.
    """

    _supports_gradient_checkpointing = True

    @register_to_config
    def __init__(
        self,
        num_attention_heads: int = 30,
        attention_head_dim: int = 64,
        in_channels: int = 16,
        out_channels: Optional[int] = 16,
        flip_sin_to_cos: bool = True,
        freq_shift: int = 0,
        time_embed_dim: int = 512,
        text_embed_dim: int = 4096,
        num_layers: int = 30,
        dropout: float = 0.0,
        attention_bias: bool = True,
        sample_width: int = 90,
        sample_height: int = 60,
        sample_frames: int = 49,
        patch_size: int = 2,
        temporal_compression_ratio: int = 4,
        max_text_seq_length: int = 226,
        activation_fn: str = "gelu-approximate",
        timestep_activation_fn: str = "silu",
        norm_elementwise_affine: bool = True,
        norm_eps: float = 1e-5,
        spatial_interpolation_scale: float = 1.875,
        temporal_interpolation_scale: float = 1.0,
    ):
        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,
        )
        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)

        # 3. 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
        self.transformer_blocks = nn.ModuleList(
            [
                CogVideoXBlock(
                    dim=inner_dim,
                    num_attention_heads=num_attention_heads,
                    attention_head_dim=attention_head_dim,
                    time_embed_dim=time_embed_dim,
                    dropout=dropout,
                    activation_fn=activation_fn,
                    attention_bias=attention_bias,
                    norm_elementwise_affine=norm_elementwise_affine,
                    norm_eps=norm_eps,
                )
                for _ in range(num_layers)
            ]
        )
        self.norm_final = nn.LayerNorm(inner_dim, norm_eps, norm_elementwise_affine)

        # 5. Output blocks
        self.norm_out = AdaLayerNorm(
            embedding_dim=time_embed_dim,
            output_dim=2 * inner_dim,
            norm_elementwise_affine=norm_elementwise_affine,
            norm_eps=norm_eps,
            chunk_dim=1,
        )
        self.proj_out = nn.Linear(inner_dim, patch_size * patch_size * out_channels)

        self.gradient_checkpointing = False

    def _set_gradient_checkpointing(self, module, value=False):
        self.gradient_checkpointing = value

    def forward(
        self,
        hidden_states: torch.Tensor,
        encoder_hidden_states: torch.Tensor,
        timestep: Union[int, float, torch.LongTensor],
        timestep_cond: Optional[torch.Tensor] = None,
        return_dict: bool = True,
    ):
        batch_size, num_frames, channels, height, width = hidden_states.shape

        # 1. Time embedding
        timesteps = timestep
        t_emb = self.time_proj(timesteps)

        # timesteps does not contain any weights and will always return f32 tensors
        # but time_embedding might actually be running in fp16. so we need to cast here.
        # there might be better ways to encapsulate this.
        t_emb = t_emb.to(dtype=hidden_states.dtype)
        emb = self.time_embedding(t_emb, timestep_cond)

        # 2. Patch embedding
        hidden_states = self.patch_embed(encoder_hidden_states, hidden_states)

        # 3. Position embedding
        seq_length = height * width * num_frames // (self.config.patch_size**2)

        pos_embeds = self.pos_embedding[:, : self.config.max_text_seq_length + seq_length]
        hidden_states = hidden_states + pos_embeds
        hidden_states = self.embedding_dropout(hidden_states)

        encoder_hidden_states = hidden_states[:, : self.config.max_text_seq_length]
        hidden_states = hidden_states[:, self.config.max_text_seq_length :]

        # 4. Transformer blocks
        for i, block in enumerate(self.transformer_blocks):
            if self.training and self.gradient_checkpointing:

                def create_custom_forward(module):
                    def custom_forward(*inputs):
                        return module(*inputs)

                    return custom_forward

                ckpt_kwargs: Dict[str, Any] = {"use_reentrant": False} if is_torch_version(">=", "1.11.0") else {}
                hidden_states, encoder_hidden_states = torch.utils.checkpoint.checkpoint(
                    create_custom_forward(block),
                    hidden_states,
                    encoder_hidden_states,
                    emb,
                    **ckpt_kwargs,
                )
            else:
                hidden_states, encoder_hidden_states = block(
                    hidden_states=hidden_states,
                    encoder_hidden_states=encoder_hidden_states,
                    temb=emb,
                )

        hidden_states = self.norm_final(hidden_states)

        # 5. Final block
        hidden_states = self.norm_out(hidden_states, temb=emb)
        hidden_states = self.proj_out(hidden_states)

        # 6. 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)

        if not return_dict:
            return (output,)
        return Transformer2DModelOutput(sample=output)