SD3 Transformer Model

The Transformer model introduced in Stable Diffusion 3. Its novelty lies in the MMDiT transformer block.

mindone.diffusers.SD3Transformer2DModel

Bases: ModelMixin, ConfigMixin, PeftAdapterMixin, FromOriginalModelMixin

The Transformer model introduced in Stable Diffusion 3.

Reference: https://arxiv.org/abs/2403.03206

PARAMETER	DESCRIPTION
sample_size	The width of the latent images. This is fixed during training since it is used to learn a number of position embeddings. TYPE: `int` DEFAULT: 128
patch_size	Patch size to turn the input data into small patches. TYPE: `int` DEFAULT: 2
in_channels	The number of channels in the input. TYPE: `int`, *optional*, defaults to 16 DEFAULT: 16
num_layers	The number of layers of Transformer blocks to use. TYPE: `int`, *optional*, defaults to 18 DEFAULT: 18
attention_head_dim	The number of channels in each head. TYPE: `int`, *optional*, defaults to 64 DEFAULT: 64
num_attention_heads	The number of heads to use for multi-head attention. TYPE: `int`, *optional*, defaults to 18 DEFAULT: 18
cross_attention_dim	The number of encoder_hidden_states dimensions to use. TYPE: `int`, *optional*
caption_projection_dim	Number of dimensions to use when projecting the encoder_hidden_states. TYPE: `int` DEFAULT: 1152
pooled_projection_dim	Number of dimensions to use when projecting the pooled_projections. TYPE: `int` DEFAULT: 2048
out_channels	Number of output channels. TYPE: `int`, defaults to 16 DEFAULT: 16

Source code in mindone/diffusers/models/transformers/transformer_sd3.py

class SD3Transformer2DModel(ModelMixin, ConfigMixin, PeftAdapterMixin, FromOriginalModelMixin):
    """
    The Transformer model introduced in Stable Diffusion 3.

    Reference: https://arxiv.org/abs/2403.03206

    Parameters:
        sample_size (`int`): The width of the latent images. This is fixed during training since
            it is used to learn a number of position embeddings.
        patch_size (`int`): Patch size to turn the input data into small patches.
        in_channels (`int`, *optional*, defaults to 16): The number of channels in the input.
        num_layers (`int`, *optional*, defaults to 18): The number of layers of Transformer blocks to use.
        attention_head_dim (`int`, *optional*, defaults to 64): The number of channels in each head.
        num_attention_heads (`int`, *optional*, defaults to 18): The number of heads to use for multi-head attention.
        cross_attention_dim (`int`, *optional*): The number of `encoder_hidden_states` dimensions to use.
        caption_projection_dim (`int`): Number of dimensions to use when projecting the `encoder_hidden_states`.
        pooled_projection_dim (`int`): Number of dimensions to use when projecting the `pooled_projections`.
        out_channels (`int`, defaults to 16): Number of output channels.

    """

    _supports_gradient_checkpointing = True

    @register_to_config
    def __init__(
        self,
        sample_size: int = 128,
        patch_size: int = 2,
        in_channels: int = 16,
        num_layers: int = 18,
        attention_head_dim: int = 64,
        num_attention_heads: int = 18,
        joint_attention_dim: int = 4096,
        caption_projection_dim: int = 1152,
        pooled_projection_dim: int = 2048,
        out_channels: int = 16,
        pos_embed_max_size: int = 96,
    ):
        super().__init__()
        default_out_channels = in_channels
        self.out_channels = out_channels if out_channels is not None else default_out_channels
        self.inner_dim = self.config.num_attention_heads * self.config.attention_head_dim

        self.pos_embed = PatchEmbed(
            height=self.config.sample_size,
            width=self.config.sample_size,
            patch_size=self.config.patch_size,
            in_channels=self.config.in_channels,
            embed_dim=self.inner_dim,
            pos_embed_max_size=pos_embed_max_size,  # hard-code for now.
        )
        self.time_text_embed = CombinedTimestepTextProjEmbeddings(
            embedding_dim=self.inner_dim, pooled_projection_dim=self.config.pooled_projection_dim
        )
        self.context_embedder = nn.Dense(self.config.joint_attention_dim, self.config.caption_projection_dim)

        # `attention_head_dim` is doubled to account for the mixing.
        # It needs to crafted when we get the actual checkpoints.
        self.transformer_blocks = nn.CellList(
            [
                JointTransformerBlock(
                    dim=self.inner_dim,
                    num_attention_heads=self.config.num_attention_heads,
                    attention_head_dim=self.config.attention_head_dim,
                    context_pre_only=i == num_layers - 1,
                )
                for i in range(self.config.num_layers)
            ]
        )

        self.norm_out = AdaLayerNormContinuous(self.inner_dim, self.inner_dim, elementwise_affine=False, eps=1e-6)
        self.proj_out = nn.Dense(self.inner_dim, patch_size * patch_size * self.out_channels, has_bias=True)

        self._gradient_checkpointing = False

    @property
    # Copied from diffusers.models.unets.unet_2d_condition.UNet2DConditionModel.attn_processors
    def attn_processors(self) -> Dict[str, AttentionProcessor]:
        r"""
        Returns:
            `dict` of attention processors: A dictionary containing all attention processors used in the model with
            indexed by its weight name.
        """
        # set recursively
        processors = {}

        def fn_recursive_add_processors(name: str, module: nn.Cell, processors: Dict[str, AttentionProcessor]):
            if hasattr(module, "get_processor"):
                processors[f"{name}.processor"] = module.get_processor()

            for sub_name, child in module.name_cells().items():
                fn_recursive_add_processors(f"{name}.{sub_name}", child, processors)

            return processors

        for name, module in self.name_cells().items():
            fn_recursive_add_processors(name, module, processors)

        return processors

    # Copied from diffusers.models.unets.unet_2d_condition.UNet2DConditionModel.set_attn_processor
    def set_attn_processor(self, processor: Union[AttentionProcessor, Dict[str, AttentionProcessor]]):
        r"""
        Sets the attention processor to use to compute attention.

        Parameters:
            processor (`dict` of `AttentionProcessor` or only `AttentionProcessor`):
                The instantiated processor class or a dictionary of processor classes that will be set as the processor
                for **all** `Attention` layers.

                If `processor` is a dict, the key needs to define the path to the corresponding cross attention
                processor. This is strongly recommended when setting trainable attention processors.

        """
        count = len(self.attn_processors.keys())

        if isinstance(processor, dict) and len(processor) != count:
            raise ValueError(
                f"A dict of processors was passed, but the number of processors {len(processor)} does not match the"
                f" number of attention layers: {count}. Please make sure to pass {count} processor classes."
            )

        def fn_recursive_attn_processor(name: str, module: nn.Cell, processor):
            if hasattr(module, "set_processor"):
                if not isinstance(processor, dict):
                    module.set_processor(processor)
                else:
                    module.set_processor(processor.pop(f"{name}.processor"))

            for sub_name, child in module.name_cells().items():
                fn_recursive_attn_processor(f"{name}.{sub_name}", child, processor)

        for name, module in self.name_cells().items():
            fn_recursive_attn_processor(name, module, processor)

    # Copied from diffusers.models.unets.unet_2d_condition.UNet2DConditionModel.fuse_qkv_projections
    def fuse_qkv_projections(self):
        """
        Enables fused QKV projections. For self-attention modules, all projection matrices (i.e., query, key, value)
        are fused. For cross-attention modules, key and value projection matrices are fused.

        <Tip warning={true}>

        This API is 🧪 experimental.

        </Tip>
        """
        self.original_attn_processors = None

        for _, attn_processor in self.attn_processors.items():
            if "Added" in str(attn_processor.__class__.__name__):
                raise ValueError("`fuse_qkv_projections()` is not supported for models having added KV projections.")

        self.original_attn_processors = self.attn_processors

        for _, module in self.cells_and_names():
            if isinstance(module, Attention):
                module.fuse_projections(fuse=True)

        self.set_attn_processor(FusedJointAttnProcessor2_0())

    # Copied from diffusers.models.unets.unet_2d_condition.UNet2DConditionModel.unfuse_qkv_projections
    def unfuse_qkv_projections(self):
        """Disables the fused QKV projection if enabled.

        <Tip warning={true}>

        This API is 🧪 experimental.

        </Tip>

        """
        if self.original_attn_processors is not None:
            self.set_attn_processor(self.original_attn_processors)

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

    @property
    def gradient_checkpointing(self):
        return self._gradient_checkpointing

    @gradient_checkpointing.setter
    def gradient_checkpointing(self, value):
        self._gradient_checkpointing = value
        for block in self.transformer_blocks:
            block._recompute(value)

    def construct(
        self,
        hidden_states: ms.Tensor,
        encoder_hidden_states: ms.Tensor = None,
        pooled_projections: ms.Tensor = None,
        timestep: ms.Tensor = None,
        block_controlnet_hidden_states: List = None,
        joint_attention_kwargs: Optional[Dict[str, Any]] = None,
        return_dict: bool = False,
    ) -> Union[ms.Tensor, Transformer2DModelOutput]:
        """
        The [`SD3Transformer2DModel`] forward method.

        Args:
            hidden_states (`ms.Tensor` of shape `(batch size, channel, height, width)`):
                Input `hidden_states`.
            encoder_hidden_states (`ms.Tensor` of shape `(batch size, sequence_len, embed_dims)`):
                Conditional embeddings (embeddings computed from the input conditions such as prompts) to use.
            pooled_projections (`ms.Tensor` of shape `(batch_size, projection_dim)`): Embeddings projected
                from the embeddings of input conditions.
            timestep ( `ms.Tensor`):
                Used to indicate denoising step.
            block_controlnet_hidden_states: (`list` of `mindspore.Tensor`):
                A list of tensors that if specified are added to the residuals of transformer blocks.
            joint_attention_kwargs (`dict`, *optional*):
                A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under
                `self.processor` in
                [diffusers.models.attention_processor](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).
            return_dict (`bool`, *optional*, defaults to `False`):
                Whether or not to return a [`~models.transformer_2d.Transformer2DModelOutput`] instead of a plain
                tuple.

        Returns:
            If `return_dict` is True, an [`~models.transformer_2d.Transformer2DModelOutput`] is returned, otherwise a
            `tuple` where the first element is the sample tensor.
        """
        if joint_attention_kwargs is not None and "scale" in joint_attention_kwargs:
            # weight the lora layers by setting `lora_scale` for each PEFT layer here
            # and remove `lora_scale` from each PEFT layer at the end.
            # scale_lora_layers & unscale_lora_layers maybe contains some operation forbidden in graph mode
            raise RuntimeError(
                f"You are trying to set scaling of lora layer by passing {joint_attention_kwargs['scale']=}. "
                f"However it's not allowed in on-the-fly model forwarding. "
                f"Please manually call `scale_lora_layers(model, lora_scale)` before model forwarding and "
                f"`unscale_lora_layers(model, lora_scale)` after model forwarding. "
                f"For example, it can be done in a pipeline call like `StableDiffusionPipeline.__call__`."
            )

        height, width = hidden_states.shape[-2:]

        hidden_states = self.pos_embed(hidden_states)  # takes care of adding positional embeddings too.
        temb = self.time_text_embed(timestep, pooled_projections)
        encoder_hidden_states = self.context_embedder(encoder_hidden_states)

        for index_block, block in enumerate(self.transformer_blocks):
            encoder_hidden_states, hidden_states = block(
                hidden_states=hidden_states, encoder_hidden_states=encoder_hidden_states, temb=temb
            )

            # controlnet residual
            if block_controlnet_hidden_states is not None and block.context_pre_only is False:
                interval_control = len(self.transformer_blocks) // len(block_controlnet_hidden_states)
                hidden_states = hidden_states + block_controlnet_hidden_states[index_block // interval_control]

        hidden_states = self.norm_out(hidden_states, temb)
        hidden_states = self.proj_out(hidden_states)

        # unpatchify
        patch_size = self.config["patch_size"]
        height = height // patch_size
        width = width // patch_size

        hidden_states = hidden_states.reshape(
            hidden_states.shape[0],
            height,
            width,
            patch_size,
            patch_size,
            self.out_channels,
        )
        # hidden_states = torch.einsum("nhwpqc->nchpwq", hidden_states)
        hidden_states = hidden_states.transpose(0, 5, 1, 3, 2, 4)
        output = hidden_states.reshape(
            hidden_states.shape[0], self.out_channels, height * patch_size, width * patch_size
        )

        if not return_dict:
            return (output,)

        return Transformer2DModelOutput(sample=output)

mindone.diffusers.SD3Transformer2DModel.attn_processors: Dict[str, AttentionProcessor] property

RETURNS	DESCRIPTION
Dict[str, AttentionProcessor]	dict of attention processors: A dictionary containing all attention processors used in the model with
Dict[str, AttentionProcessor]	indexed by its weight name.

mindone.diffusers.SD3Transformer2DModel.construct(hidden_states, encoder_hidden_states=None, pooled_projections=None, timestep=None, block_controlnet_hidden_states=None, joint_attention_kwargs=None, return_dict=False)

The [SD3Transformer2DModel] forward method.

PARAMETER	DESCRIPTION
hidden_states	Input hidden_states. TYPE: `ms.Tensor` of shape `(batch size, channel, height, width)`
encoder_hidden_states	Conditional embeddings (embeddings computed from the input conditions such as prompts) to use. TYPE: `ms.Tensor` of shape `(batch size, sequence_len, embed_dims)` DEFAULT: None
pooled_projections	Embeddings projected from the embeddings of input conditions. TYPE: `ms.Tensor` of shape `(batch_size, projection_dim)` DEFAULT: None
timestep	Used to indicate denoising step. TYPE: `ms.Tensor` DEFAULT: None
block_controlnet_hidden_states	(list of mindspore.Tensor): A list of tensors that if specified are added to the residuals of transformer blocks. TYPE: List DEFAULT: None
joint_attention_kwargs	A kwargs dictionary that if specified is passed along to the AttentionProcessor as defined under self.processor in diffusers.models.attention_processor. TYPE: `dict`, *optional* DEFAULT: None
return_dict	Whether or not to return a [~models.transformer_2d.Transformer2DModelOutput] instead of a plain tuple. TYPE: `bool`, *optional*, defaults to `False` DEFAULT: False

RETURNS	DESCRIPTION
Union[Tensor, Transformer2DModelOutput]	If return_dict is True, an [~models.transformer_2d.Transformer2DModelOutput] is returned, otherwise a
Union[Tensor, Transformer2DModelOutput]	tuple where the first element is the sample tensor.

Source code in mindone/diffusers/models/transformers/transformer_sd3.py

def construct(
    self,
    hidden_states: ms.Tensor,
    encoder_hidden_states: ms.Tensor = None,
    pooled_projections: ms.Tensor = None,
    timestep: ms.Tensor = None,
    block_controlnet_hidden_states: List = None,
    joint_attention_kwargs: Optional[Dict[str, Any]] = None,
    return_dict: bool = False,
) -> Union[ms.Tensor, Transformer2DModelOutput]:
    """
    The [`SD3Transformer2DModel`] forward method.

    Args:
        hidden_states (`ms.Tensor` of shape `(batch size, channel, height, width)`):
            Input `hidden_states`.
        encoder_hidden_states (`ms.Tensor` of shape `(batch size, sequence_len, embed_dims)`):
            Conditional embeddings (embeddings computed from the input conditions such as prompts) to use.
        pooled_projections (`ms.Tensor` of shape `(batch_size, projection_dim)`): Embeddings projected
            from the embeddings of input conditions.
        timestep ( `ms.Tensor`):
            Used to indicate denoising step.
        block_controlnet_hidden_states: (`list` of `mindspore.Tensor`):
            A list of tensors that if specified are added to the residuals of transformer blocks.
        joint_attention_kwargs (`dict`, *optional*):
            A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under
            `self.processor` in
            [diffusers.models.attention_processor](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).
        return_dict (`bool`, *optional*, defaults to `False`):
            Whether or not to return a [`~models.transformer_2d.Transformer2DModelOutput`] instead of a plain
            tuple.

    Returns:
        If `return_dict` is True, an [`~models.transformer_2d.Transformer2DModelOutput`] is returned, otherwise a
        `tuple` where the first element is the sample tensor.
    """
    if joint_attention_kwargs is not None and "scale" in joint_attention_kwargs:
        # weight the lora layers by setting `lora_scale` for each PEFT layer here
        # and remove `lora_scale` from each PEFT layer at the end.
        # scale_lora_layers & unscale_lora_layers maybe contains some operation forbidden in graph mode
        raise RuntimeError(
            f"You are trying to set scaling of lora layer by passing {joint_attention_kwargs['scale']=}. "
            f"However it's not allowed in on-the-fly model forwarding. "
            f"Please manually call `scale_lora_layers(model, lora_scale)` before model forwarding and "
            f"`unscale_lora_layers(model, lora_scale)` after model forwarding. "
            f"For example, it can be done in a pipeline call like `StableDiffusionPipeline.__call__`."
        )

    height, width = hidden_states.shape[-2:]

    hidden_states = self.pos_embed(hidden_states)  # takes care of adding positional embeddings too.
    temb = self.time_text_embed(timestep, pooled_projections)
    encoder_hidden_states = self.context_embedder(encoder_hidden_states)

    for index_block, block in enumerate(self.transformer_blocks):
        encoder_hidden_states, hidden_states = block(
            hidden_states=hidden_states, encoder_hidden_states=encoder_hidden_states, temb=temb
        )

        # controlnet residual
        if block_controlnet_hidden_states is not None and block.context_pre_only is False:
            interval_control = len(self.transformer_blocks) // len(block_controlnet_hidden_states)
            hidden_states = hidden_states + block_controlnet_hidden_states[index_block // interval_control]

    hidden_states = self.norm_out(hidden_states, temb)
    hidden_states = self.proj_out(hidden_states)

    # unpatchify
    patch_size = self.config["patch_size"]
    height = height // patch_size
    width = width // patch_size

    hidden_states = hidden_states.reshape(
        hidden_states.shape[0],
        height,
        width,
        patch_size,
        patch_size,
        self.out_channels,
    )
    # hidden_states = torch.einsum("nhwpqc->nchpwq", hidden_states)
    hidden_states = hidden_states.transpose(0, 5, 1, 3, 2, 4)
    output = hidden_states.reshape(
        hidden_states.shape[0], self.out_channels, height * patch_size, width * patch_size
    )

    if not return_dict:
        return (output,)

    return Transformer2DModelOutput(sample=output)

mindone.diffusers.SD3Transformer2DModel.fuse_qkv_projections()

Enables fused QKV projections. For self-attention modules, all projection matrices (i.e., query, key, value) are fused. For cross-attention modules, key and value projection matrices are fused.

This API is 🧪 experimental.

Source code in mindone/diffusers/models/transformers/transformer_sd3.py

def fuse_qkv_projections(self):
    """
    Enables fused QKV projections. For self-attention modules, all projection matrices (i.e., query, key, value)
    are fused. For cross-attention modules, key and value projection matrices are fused.

    <Tip warning={true}>

    This API is 🧪 experimental.

    </Tip>
    """
    self.original_attn_processors = None

    for _, attn_processor in self.attn_processors.items():
        if "Added" in str(attn_processor.__class__.__name__):
            raise ValueError("`fuse_qkv_projections()` is not supported for models having added KV projections.")

    self.original_attn_processors = self.attn_processors

    for _, module in self.cells_and_names():
        if isinstance(module, Attention):
            module.fuse_projections(fuse=True)

    self.set_attn_processor(FusedJointAttnProcessor2_0())

mindone.diffusers.SD3Transformer2DModel.set_attn_processor(processor)

Sets the attention processor to use to compute attention.

PARAMETER	DESCRIPTION
processor	The instantiated processor class or a dictionary of processor classes that will be set as the processor for all Attention layers. If processor is a dict, the key needs to define the path to the corresponding cross attention processor. This is strongly recommended when setting trainable attention processors. TYPE: `dict` of `AttentionProcessor` or only `AttentionProcessor`

Source code in mindone/diffusers/models/transformers/transformer_sd3.py

def set_attn_processor(self, processor: Union[AttentionProcessor, Dict[str, AttentionProcessor]]):
    r"""
    Sets the attention processor to use to compute attention.

    Parameters:
        processor (`dict` of `AttentionProcessor` or only `AttentionProcessor`):
            The instantiated processor class or a dictionary of processor classes that will be set as the processor
            for **all** `Attention` layers.

            If `processor` is a dict, the key needs to define the path to the corresponding cross attention
            processor. This is strongly recommended when setting trainable attention processors.

    """
    count = len(self.attn_processors.keys())

    if isinstance(processor, dict) and len(processor) != count:
        raise ValueError(
            f"A dict of processors was passed, but the number of processors {len(processor)} does not match the"
            f" number of attention layers: {count}. Please make sure to pass {count} processor classes."
        )

    def fn_recursive_attn_processor(name: str, module: nn.Cell, processor):
        if hasattr(module, "set_processor"):
            if not isinstance(processor, dict):
                module.set_processor(processor)
            else:
                module.set_processor(processor.pop(f"{name}.processor"))

        for sub_name, child in module.name_cells().items():
            fn_recursive_attn_processor(f"{name}.{sub_name}", child, processor)

    for name, module in self.name_cells().items():
        fn_recursive_attn_processor(name, module, processor)

mindone.diffusers.SD3Transformer2DModel.unfuse_qkv_projections()

Disables the fused QKV projection if enabled.

This API is 🧪 experimental.

Source code in mindone/diffusers/models/transformers/transformer_sd3.py

def unfuse_qkv_projections(self):
    """Disables the fused QKV projection if enabled.

    <Tip warning={true}>

    This API is 🧪 experimental.

    </Tip>

    """
    if self.original_attn_processors is not None:
        self.set_attn_processor(self.original_attn_processors)