optimum-rbln 0.9.3.post1__py3-none-any.whl

optimum/rbln/diffusers/configurations/models/configuration_transformer_sd3.py

@@ -0,0 +1,63 @@
+# Copyright 2025 Rebellions Inc. 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, Optional, Tuple, Union
+
+from ....configuration_utils import RBLNModelConfig
+
+
+class RBLNSD3Transformer2DModelConfig(RBLNModelConfig):
+    """
+    Configuration class for RBLN Stable Diffusion 3 Transformer models.
+
+    This class inherits from RBLNModelConfig and provides specific configuration options
+    for Transformer models used in diffusion models like Stable Diffusion 3.
+    """
+
+    subclass_non_save_attributes = ["_batch_size_is_specified"]
+
+    def __init__(
+        self,
+        batch_size: Optional[int] = None,
+        sample_size: Optional[Union[int, Tuple[int, int]]] = None,
+        prompt_embed_length: Optional[int] = None,
+        **kwargs: Any,
+    ):
+        """
+        Args:
+            batch_size (Optional[int]): The batch size for inference. Defaults to 1.
+            sample_size (Optional[Union[int, Tuple[int, int]]]): The spatial dimensions (height, width)
+                of the generated samples. If an integer is provided, it's used for both height and width.
+            prompt_embed_length (Optional[int]): The length of the embedded prompt vectors that
+                will be used to condition the transformer model.
+            kwargs: Additional arguments passed to the parent RBLNModelConfig.
+
+        Raises:
+            ValueError: If batch_size is not a positive integer.
+        """
+        super().__init__(**kwargs)
+        self._batch_size_is_specified = batch_size is not None
+
+        self.batch_size = batch_size or 1
+        if not isinstance(self.batch_size, int) or self.batch_size < 0:
+            raise ValueError(f"batch_size must be a positive integer, got {self.batch_size}")
+
+        self.prompt_embed_length = prompt_embed_length
+        self.sample_size = sample_size
+        if isinstance(self.sample_size, int):
+            self.sample_size = (self.sample_size, self.sample_size)
+
+    @property
+    def batch_size_is_specified(self):
+        return self._batch_size_is_specified

optimum/rbln/diffusers/configurations/models/configuration_unet_2d_condition.py

@@ -0,0 +1,81 @@
+# Copyright 2025 Rebellions Inc. 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, Optional, Tuple
+
+from ....configuration_utils import RBLNModelConfig
+
+
+class RBLNUNet2DConditionModelConfig(RBLNModelConfig):
+    """
+    Configuration class for RBLN UNet2DCondition models.
+
+    This class inherits from RBLNModelConfig and provides specific configuration options
+    for UNet2DCondition models used in diffusion-based image generation.
+    """
+
+    subclass_non_save_attributes = ["_batch_size_is_specified"]
+
+    def __init__(
+        self,
+        batch_size: Optional[int] = None,
+        sample_size: Optional[Tuple[int, int]] = None,
+        in_channels: Optional[int] = None,
+        cross_attention_dim: Optional[int] = None,
+        use_additional_residuals: Optional[bool] = None,
+        max_seq_len: Optional[int] = None,
+        in_features: Optional[int] = None,
+        text_model_hidden_size: Optional[int] = None,
+        image_model_hidden_size: Optional[int] = None,
+        **kwargs: Any,
+    ):
+        """
+        Args:
+            batch_size (Optional[int]): The batch size for inference. Defaults to 1.
+            sample_size (Optional[Tuple[int, int]]): The spatial dimensions (height, width) of the generated samples.
+                If an integer is provided, it's used for both height and width.
+            in_channels (Optional[int]): Number of input channels for the UNet.
+            cross_attention_dim (Optional[int]): Dimension of the cross-attention features.
+            use_additional_residuals (Optional[bool]): Whether to use additional residual connections in the model.
+            max_seq_len (Optional[int]): Maximum sequence length for text inputs when used with cross-attention.
+            in_features (Optional[int]): Number of input features for the model.
+            text_model_hidden_size (Optional[int]): Hidden size of the text encoder model.
+            image_model_hidden_size (Optional[int]): Hidden size of the image encoder model.
+            kwargs: Additional arguments passed to the parent RBLNModelConfig.
+
+        Raises:
+            ValueError: If batch_size is not a positive integer.
+        """
+        super().__init__(**kwargs)
+        self._batch_size_is_specified = batch_size is not None
+
+        self.batch_size = batch_size or 1
+        if not isinstance(self.batch_size, int) or self.batch_size < 0:
+            raise ValueError(f"batch_size must be a positive integer, got {self.batch_size}")
+
+        self.in_channels = in_channels
+        self.cross_attention_dim = cross_attention_dim
+        self.use_additional_residuals = use_additional_residuals
+        self.max_seq_len = max_seq_len
+        self.in_features = in_features
+        self.text_model_hidden_size = text_model_hidden_size
+        self.image_model_hidden_size = image_model_hidden_size
+
+        self.sample_size = sample_size
+        if isinstance(sample_size, int):
+            self.sample_size = (sample_size, sample_size)
+
+    @property
+    def batch_size_is_specified(self):
+        return self._batch_size_is_specified

optimum/rbln/diffusers/configurations/models/configuration_unet_spatio_temporal_condition.py

@@ -0,0 +1,59 @@
+# Copyright 2025 Rebellions Inc. 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, Optional, Tuple
+
+from ....configuration_utils import RBLNModelConfig
+
+
+class RBLNUNetSpatioTemporalConditionModelConfig(RBLNModelConfig):
+    subclass_non_save_attributes = ["_batch_size_is_specified"]
+
+    def __init__(
+        self,
+        batch_size: Optional[int] = None,
+        sample_size: Optional[Tuple[int, int]] = None,
+        in_features: Optional[int] = None,
+        num_frames: Optional[int] = None,
+        **kwargs: Any,
+    ):
+        """
+        Args:
+            batch_size (Optional[int]): The batch size for inference. Defaults to 1.
+            sample_size (Optional[Tuple[int, int]]): The spatial dimensions (height, width) of the generated samples.
+                If an integer is provided, it's used for both height and width.
+            in_features (Optional[int]): Number of input features for the model.
+            num_frames (Optional[int]): The number of frames in the generated video.
+            kwargs: Additional arguments passed to the parent RBLNModelConfig.
+
+        Raises:
+            ValueError: If batch_size is not a positive integer.
+        """
+        super().__init__(**kwargs)
+        self._batch_size_is_specified = batch_size is not None
+
+        self.batch_size = batch_size or 1
+        if not isinstance(self.batch_size, int) or self.batch_size < 0:
+            raise ValueError(f"batch_size must be a positive integer, got {self.batch_size}")
+
+        self.in_features = in_features
+        self.num_frames = num_frames
+
+        self.sample_size = sample_size
+        if isinstance(sample_size, int):
+            self.sample_size = (sample_size, sample_size)
+
+    @property
+    def batch_size_is_specified(self):
+        return self._batch_size_is_specified

optimum/rbln/diffusers/configurations/models/configuration_vq_model.py

@@ -0,0 +1,74 @@
+# Copyright 2025 Rebellions Inc. 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, Optional, Tuple
+
+from ....configuration_utils import RBLNModelConfig
+
+
+class RBLNVQModelConfig(RBLNModelConfig):
+    """
+    Configuration class for RBLN VQModel models, used in Kandinsky.
+
+    This class inherits from RBLNModelConfig and provides specific configuration options
+    for VQModel, which acts similarly to a VAE but uses vector quantization.
+    """
+
+    def __init__(
+        self,
+        batch_size: Optional[int] = None,
+        sample_size: Optional[Tuple[int, int]] = None,
+        uses_encoder: Optional[bool] = None,
+        vqmodel_scale_factor: Optional[float] = None,  # TODO: rename to scaling_factor
+        in_channels: Optional[int] = None,
+        latent_channels: Optional[int] = None,
+        **kwargs: Any,
+    ):
+        """
+        Args:
+            batch_size (Optional[int]): The batch size for inference. Defaults to 1.
+            sample_size (Optional[Tuple[int, int]]): The spatial dimensions (height, width) of the input/output images.
+                If an integer is provided, it's used for both height and width.
+            uses_encoder (Optional[bool]): Whether to include the encoder part of the VAE in the model.
+                When False, only the decoder is used (for latent-to-image conversion).
+            vqmodel_scale_factor (Optional[float]): The scaling factor between pixel space and latent space.
+                Determines the downsampling ratio between original images and latent representations.
+            in_channels (Optional[int]): Number of input channels for the model.
+            latent_channels (Optional[int]): Number of channels in the latent space.
+            kwargs: Additional arguments passed to the parent RBLNModelConfig.
+
+        Raises:
+            ValueError: If batch_size is not a positive integer.
+        """
+        super().__init__(**kwargs)
+        self.batch_size = batch_size or 1
+        if not isinstance(self.batch_size, int) or self.batch_size < 0:
+            raise ValueError(f"batch_size must be a positive integer, got {self.batch_size}")
+
+        self.uses_encoder = uses_encoder
+        self.sample_size = sample_size
+        if isinstance(self.sample_size, int):
+            self.sample_size = (self.sample_size, self.sample_size)
+
+        self.vqmodel_scale_factor = vqmodel_scale_factor
+        self.in_channels = in_channels
+        self.latent_channels = latent_channels
+
+    @property
+    def image_size(self):
+        return self.sample_size
+
+    @property
+    def latent_sample_size(self):
+        return (self.image_size[0] // self.vqmodel_scale_factor, self.image_size[1] // self.vqmodel_scale_factor)

optimum/rbln/diffusers/configurations/pipelines/__init__.py

@@ -0,0 +1,34 @@
+from .configuration_controlnet import (
+    RBLNStableDiffusionControlNetImg2ImgPipelineConfig,
+    RBLNStableDiffusionControlNetPipelineConfig,
+    RBLNStableDiffusionXLControlNetImg2ImgPipelineConfig,
+    RBLNStableDiffusionXLControlNetPipelineConfig,
+)
+from .configuration_cosmos import RBLNCosmosTextToWorldPipelineConfig, RBLNCosmosVideoToWorldPipelineConfig
+from .configuration_kandinsky2_2 import (
+    RBLNKandinskyV22CombinedPipelineConfig,
+    RBLNKandinskyV22Img2ImgCombinedPipelineConfig,
+    RBLNKandinskyV22Img2ImgPipelineConfig,
+    RBLNKandinskyV22InpaintCombinedPipelineConfig,
+    RBLNKandinskyV22InpaintPipelineConfig,
+    RBLNKandinskyV22PipelineConfig,
+    RBLNKandinskyV22PriorPipelineConfig,
+)
+from .configuration_stable_diffusion import (
+    RBLNStableDiffusionImg2ImgPipelineConfig,
+    RBLNStableDiffusionInpaintPipelineConfig,
+    RBLNStableDiffusionPipelineConfig,
+)
+from .configuration_stable_diffusion_3 import (
+    RBLNStableDiffusion3Img2ImgPipelineConfig,
+    RBLNStableDiffusion3InpaintPipelineConfig,
+    RBLNStableDiffusion3PipelineConfig,
+)
+from .configuration_stable_diffusion_xl import (
+    RBLNStableDiffusionXLImg2ImgPipelineConfig,
+    RBLNStableDiffusionXLInpaintPipelineConfig,
+    RBLNStableDiffusionXLPipelineConfig,
+)
+from .configuration_stable_video_diffusion import (
+    RBLNStableVideoDiffusionPipelineConfig,
+)

optimum/rbln/diffusers/configurations/pipelines/configuration_controlnet.py

@@ -0,0 +1,316 @@
+# Copyright 2025 Rebellions Inc. 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, Optional, Tuple
+
+from ....configuration_utils import RBLNModelConfig
+from ....transformers import RBLNCLIPTextModelConfig, RBLNCLIPTextModelWithProjectionConfig
+from ..models import RBLNAutoencoderKLConfig, RBLNControlNetModelConfig, RBLNUNet2DConditionModelConfig
+
+
+class RBLNStableDiffusionControlNetPipelineBaseConfig(RBLNModelConfig):
+    submodules = ["text_encoder", "unet", "vae", "controlnet"]
+    _vae_uses_encoder = False
+
+    def __init__(
+        self,
+        text_encoder: Optional[RBLNCLIPTextModelConfig] = None,
+        unet: Optional[RBLNUNet2DConditionModelConfig] = None,
+        vae: Optional[RBLNAutoencoderKLConfig] = None,
+        controlnet: Optional[RBLNControlNetModelConfig] = None,
+        *,
+        batch_size: Optional[int] = None,
+        img_height: Optional[int] = None,
+        img_width: Optional[int] = None,
+        height: Optional[int] = None,
+        width: Optional[int] = None,
+        sample_size: Optional[Tuple[int, int]] = None,
+        image_size: Optional[Tuple[int, int]] = None,
+        guidance_scale: Optional[float] = None,
+        **kwargs: Any,
+    ):
+        """
+        Args:
+            text_encoder (Optional[RBLNCLIPTextModelConfig]): Configuration for the text encoder component.
+                Initialized as RBLNCLIPTextModelConfig if not provided.
+            unet (Optional[RBLNUNet2DConditionModelConfig]): Configuration for the UNet model component.
+                Initialized as RBLNUNet2DConditionModelConfig if not provided.
+            vae (Optional[RBLNAutoencoderKLConfig]): Configuration for the VAE model component.
+                Initialized as RBLNAutoencoderKLConfig if not provided.
+            controlnet (Optional[RBLNControlNetModelConfig]): Configuration for the ControlNet model component.
+                Initialized as RBLNControlNetModelConfig if not provided.
+            batch_size (Optional[int]): Batch size for inference, applied to all submodules.
+            img_height (Optional[int]): Height of the generated images.
+            img_width (Optional[int]): Width of the generated images.
+            height (Optional[int]): Height of the generated images.
+            width (Optional[int]): Width of the generated images.
+            sample_size (Optional[Tuple[int, int]]): Spatial dimensions for the UNet model.
+            image_size (Optional[Tuple[int, int]]): Alternative way to specify image dimensions.
+                Cannot be used together with img_height/img_width.
+            guidance_scale (Optional[float]): Scale for classifier-free guidance.
+            kwargs: Additional arguments passed to the parent RBLNModelConfig.
+
+        Raises:
+            ValueError: If both image_size and img_height/img_width are provided.
+
+        Note:
+            When guidance_scale > 1.0, the UNet batch size is automatically doubled to
+            accommodate classifier-free guidance.
+        """
+        super().__init__(**kwargs)
+
+        # Initial check for image_size conflict remains as is
+        if image_size is not None and (
+            img_height is not None or img_width is not None or height is not None or width is not None
+        ):
+            raise ValueError("image_size cannot be provided alongside img_height/img_width or height/width")
+
+        # Prioritize height/width (HF-aligned)
+        if height is not None and width is not None:
+            if img_height is not None or img_width is not None:
+                # Raise error if both sets of arguments are provided
+                raise ValueError(
+                    "Cannot provide both 'height'/'width' and 'img_height'/'img_width' simultaneously. "
+                    "Please use one set of arguments for image dimensions, preferring 'height'/'width'."
+                )
+            image_size = (height, width)
+        elif (height is not None and width is None) or (height is None and width is not None):
+            raise ValueError("Both height and width must be provided together if used")
+        # Fallback to img_height/img_width for backward compatibility
+        elif img_height is not None and img_width is not None:
+            image_size = (img_height, img_width)
+        elif (img_height is not None and img_width is None) or (img_height is None and img_width is not None):
+            raise ValueError("Both img_height and img_width must be provided together if used")
+
+        self.text_encoder = self.initialize_submodule_config(
+            text_encoder,
+            cls_name="RBLNCLIPTextModelConfig",
+            batch_size=batch_size,
+        )
+        self.unet = self.initialize_submodule_config(
+            unet,
+            cls_name="RBLNUNet2DConditionModelConfig",
+            sample_size=sample_size,
+        )
+        self.vae = self.initialize_submodule_config(
+            vae,
+            cls_name="RBLNAutoencoderKLConfig",
+            batch_size=batch_size,
+            uses_encoder=self.__class__._vae_uses_encoder,
+            sample_size=image_size,  # image size is equal to sample size in vae
+        )
+        self.controlnet = self.initialize_submodule_config(
+            controlnet,
+            cls_name="RBLNControlNetModelConfig",
+        )
+
+        # Get default guidance scale from original class to set UNet and ControlNet batch size
+        if guidance_scale is None:
+            guidance_scale = self.get_default_values_for_original_cls("__call__", ["guidance_scale"])["guidance_scale"]
+
+        if guidance_scale is not None:
+            do_classifier_free_guidance = guidance_scale > 1.0
+            if do_classifier_free_guidance:
+                if not self.unet.batch_size_is_specified:
+                    self.unet.batch_size = self.text_encoder.batch_size * 2
+                if not self.controlnet.batch_size_is_specified:
+                    self.controlnet.batch_size = self.text_encoder.batch_size * 2
+            else:
+                if not self.unet.batch_size_is_specified:
+                    self.unet.batch_size = self.text_encoder.batch_size
+                if not self.controlnet.batch_size_is_specified:
+                    self.controlnet.batch_size = self.text_encoder.batch_size
+
+    @property
+    def batch_size(self):
+        return self.vae.batch_size
+
+    @property
+    def sample_size(self):
+        return self.unet.sample_size
+
+    @property
+    def image_size(self):
+        return self.vae.sample_size
+
+
+class RBLNStableDiffusionControlNetPipelineConfig(RBLNStableDiffusionControlNetPipelineBaseConfig):
+    """
+    Configuration for Stable Diffusion ControlNet pipeline.
+    """
+
+    _vae_uses_encoder = False
+
+
+class RBLNStableDiffusionControlNetImg2ImgPipelineConfig(RBLNStableDiffusionControlNetPipelineBaseConfig):
+    """
+    Configuration for Stable Diffusion ControlNet image-to-image pipeline.
+    """
+
+    _vae_uses_encoder = True
+
+
+class RBLNStableDiffusionXLControlNetPipelineBaseConfig(RBLNModelConfig):
+    """
+    Base configuration for Stable Diffusion XL ControlNet pipelines.
+    """
+
+    submodules = ["text_encoder", "text_encoder_2", "unet", "vae", "controlnet"]
+    _vae_uses_encoder = False
+
+    def __init__(
+        self,
+        text_encoder: Optional[RBLNCLIPTextModelConfig] = None,
+        text_encoder_2: Optional[RBLNCLIPTextModelWithProjectionConfig] = None,
+        unet: Optional[RBLNUNet2DConditionModelConfig] = None,
+        vae: Optional[RBLNAutoencoderKLConfig] = None,
+        controlnet: Optional[RBLNControlNetModelConfig] = None,
+        *,
+        batch_size: Optional[int] = None,
+        img_height: Optional[int] = None,
+        img_width: Optional[int] = None,
+        height: Optional[int] = None,
+        width: Optional[int] = None,
+        sample_size: Optional[Tuple[int, int]] = None,
+        image_size: Optional[Tuple[int, int]] = None,
+        guidance_scale: Optional[float] = None,
+        **kwargs: Any,
+    ):
+        """
+        Args:
+            text_encoder (Optional[RBLNCLIPTextModelConfig]): Configuration for the primary text encoder.
+                Initialized as RBLNCLIPTextModelConfig if not provided.
+            text_encoder_2 (Optional[RBLNCLIPTextModelWithProjectionConfig]): Configuration for the secondary text encoder.
+                Initialized as RBLNCLIPTextModelWithProjectionConfig if not provided.
+            unet (Optional[RBLNUNet2DConditionModelConfig]): Configuration for the UNet model component.
+                Initialized as RBLNUNet2DConditionModelConfig if not provided.
+            vae (Optional[RBLNAutoencoderKLConfig]): Configuration for the VAE model component.
+                Initialized as RBLNAutoencoderKLConfig if not provided.
+            controlnet (Optional[RBLNControlNetModelConfig]): Configuration for the ControlNet model component.
+                Initialized as RBLNControlNetModelConfig if not provided.
+            batch_size (Optional[int]): Batch size for inference, applied to all submodules.
+            img_height (Optional[int]): Height of the generated images.
+            img_width (Optional[int]): Width of the generated images.
+            height (Optional[int]): Height of the generated images.
+            width (Optional[int]): Width of the generated images.
+            sample_size (Optional[Tuple[int, int]]): Spatial dimensions for the UNet model.
+            image_size (Optional[Tuple[int, int]]): Alternative way to specify image dimensions.
+                Cannot be used together with img_height/img_width.
+            guidance_scale (Optional[float]): Scale for classifier-free guidance.
+            kwargs: Additional arguments passed to the parent RBLNModelConfig.
+
+        Raises:
+            ValueError: If both image_size and img_height/img_width are provided.
+
+        Note:
+            When guidance_scale > 1.0, the UNet batch size is automatically doubled to
+            accommodate classifier-free guidance.
+        """
+        super().__init__(**kwargs)
+
+        # Initial check for image_size conflict remains as is
+        if image_size is not None and (
+            img_height is not None or img_width is not None or height is not None or width is not None
+        ):
+            raise ValueError("image_size cannot be provided alongside img_height/img_width or height/width")
+
+        # Prioritize height/width (HF-aligned)
+        if height is not None and width is not None:
+            if img_height is not None or img_width is not None:
+                # Raise error if both sets of arguments are provided
+                raise ValueError(
+                    "Cannot provide both 'height'/'width' and 'img_height'/'img_width' simultaneously. "
+                    "Please use one set of arguments for image dimensions, preferring 'height'/'width'."
+                )
+            image_size = (height, width)
+        elif (height is not None and width is None) or (height is None and width is not None):
+            raise ValueError("Both height and width must be provided together if used")
+        # Fallback to img_height/img_width for backward compatibility
+        elif img_height is not None and img_width is not None:
+            image_size = (img_height, img_width)
+        elif (img_height is not None and img_width is None) or (img_height is None and img_width is not None):
+            raise ValueError("Both img_height and img_width must be provided together if used")
+
+        self.text_encoder = self.initialize_submodule_config(
+            text_encoder,
+            cls_name="RBLNCLIPTextModelConfig",
+            batch_size=batch_size,
+        )
+        self.text_encoder_2 = self.initialize_submodule_config(
+            text_encoder_2,
+            cls_name="RBLNCLIPTextModelWithProjectionConfig",
+            batch_size=batch_size,
+        )
+        self.unet = self.initialize_submodule_config(
+            unet,
+            cls_name="RBLNUNet2DConditionModelConfig",
+            sample_size=sample_size,
+        )
+        self.vae = self.initialize_submodule_config(
+            vae,
+            cls_name="RBLNAutoencoderKLConfig",
+            batch_size=batch_size,
+            uses_encoder=self.__class__._vae_uses_encoder,
+            sample_size=image_size,  # image size is equal to sample size in vae
+        )
+        self.controlnet = self.initialize_submodule_config(
+            controlnet,
+            cls_name="RBLNControlNetModelConfig",
+        )
+
+        # Get default guidance scale from original class to set UNet and ControlNet batch size
+        guidance_scale = (
+            guidance_scale
+            or self.get_default_values_for_original_cls("__call__", ["guidance_scale"])["guidance_scale"]
+        )
+
+        do_classifier_free_guidance = guidance_scale > 1.0
+        if do_classifier_free_guidance:
+            if not self.unet.batch_size_is_specified:
+                self.unet.batch_size = self.text_encoder.batch_size * 2
+            if not self.controlnet.batch_size_is_specified:
+                self.controlnet.batch_size = self.text_encoder.batch_size * 2
+        else:
+            if not self.unet.batch_size_is_specified:
+                self.unet.batch_size = self.text_encoder.batch_size
+            if not self.controlnet.batch_size_is_specified:
+                self.controlnet.batch_size = self.text_encoder.batch_size
+
+    @property
+    def batch_size(self):
+        return self.vae.batch_size
+
+    @property
+    def sample_size(self):
+        return self.unet.sample_size
+
+    @property
+    def image_size(self):
+        return self.vae.sample_size
+
+
+class RBLNStableDiffusionXLControlNetPipelineConfig(RBLNStableDiffusionXLControlNetPipelineBaseConfig):
+    """
+    Configuration for Stable Diffusion XL ControlNet pipeline.
+    """
+
+    _vae_uses_encoder = False
+
+
+class RBLNStableDiffusionXLControlNetImg2ImgPipelineConfig(RBLNStableDiffusionXLControlNetPipelineBaseConfig):
+    """
+    Configuration for Stable Diffusion XL ControlNet image-to-image pipeline.
+    """
+
+    _vae_uses_encoder = True