diffusers 0.27.1__py3-none-any.whl → 0.32.2__py3-none-any.whl

diffusers/loaders/unet_loader_utils.py

@@ -0,0 +1,163 @@
+# Copyright 2024 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.
+import copy
+from typing import TYPE_CHECKING, Dict, List, Union
+
+from ..utils import logging
+
+
+if TYPE_CHECKING:
+    # import here to avoid circular imports
+    from ..models import UNet2DConditionModel
+
+logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
+
+
+def _translate_into_actual_layer_name(name):
+    """Translate user-friendly name (e.g. 'mid') into actual layer name (e.g. 'mid_block.attentions.0')"""
+    if name == "mid":
+        return "mid_block.attentions.0"
+
+    updown, block, attn = name.split(".")
+
+    updown = updown.replace("down", "down_blocks").replace("up", "up_blocks")
+    block = block.replace("block_", "")
+    attn = "attentions." + attn
+
+    return ".".join((updown, block, attn))
+
+
+def _maybe_expand_lora_scales(
+    unet: "UNet2DConditionModel", weight_scales: List[Union[float, Dict]], default_scale=1.0
+):
+    blocks_with_transformer = {
+        "down": [i for i, block in enumerate(unet.down_blocks) if hasattr(block, "attentions")],
+        "up": [i for i, block in enumerate(unet.up_blocks) if hasattr(block, "attentions")],
+    }
+    transformer_per_block = {"down": unet.config.layers_per_block, "up": unet.config.layers_per_block + 1}
+
+    expanded_weight_scales = [
+        _maybe_expand_lora_scales_for_one_adapter(
+            weight_for_adapter,
+            blocks_with_transformer,
+            transformer_per_block,
+            unet.state_dict(),
+            default_scale=default_scale,
+        )
+        for weight_for_adapter in weight_scales
+    ]
+
+    return expanded_weight_scales
+
+
+def _maybe_expand_lora_scales_for_one_adapter(
+    scales: Union[float, Dict],
+    blocks_with_transformer: Dict[str, int],
+    transformer_per_block: Dict[str, int],
+    state_dict: None,
+    default_scale: float = 1.0,
+):
+    """
+    Expands the inputs into a more granular dictionary. See the example below for more details.
+
+    Parameters:
+        scales (`Union[float, Dict]`):
+            Scales dict to expand.
+        blocks_with_transformer (`Dict[str, int]`):
+            Dict with keys 'up' and 'down', showing which blocks have transformer layers
+        transformer_per_block (`Dict[str, int]`):
+            Dict with keys 'up' and 'down', showing how many transformer layers each block has
+
+    E.g. turns
+    ```python
+    scales = {"down": 2, "mid": 3, "up": {"block_0": 4, "block_1": [5, 6, 7]}}
+    blocks_with_transformer = {"down": [1, 2], "up": [0, 1]}
+    transformer_per_block = {"down": 2, "up": 3}
+    ```
+    into
+    ```python
+    {
+        "down.block_1.0": 2,
+        "down.block_1.1": 2,
+        "down.block_2.0": 2,
+        "down.block_2.1": 2,
+        "mid": 3,
+        "up.block_0.0": 4,
+        "up.block_0.1": 4,
+        "up.block_0.2": 4,
+        "up.block_1.0": 5,
+        "up.block_1.1": 6,
+        "up.block_1.2": 7,
+    }
+    ```
+    """
+    if sorted(blocks_with_transformer.keys()) != ["down", "up"]:
+        raise ValueError("blocks_with_transformer needs to be a dict with keys `'down' and `'up'`")
+
+    if sorted(transformer_per_block.keys()) != ["down", "up"]:
+        raise ValueError("transformer_per_block needs to be a dict with keys `'down' and `'up'`")
+
+    if not isinstance(scales, dict):
+        # don't expand if scales is a single number
+        return scales
+
+    scales = copy.deepcopy(scales)
+
+    if "mid" not in scales:
+        scales["mid"] = default_scale
+    elif isinstance(scales["mid"], list):
+        if len(scales["mid"]) == 1:
+            scales["mid"] = scales["mid"][0]
+        else:
+            raise ValueError(f"Expected 1 scales for mid, got {len(scales['mid'])}.")
+
+    for updown in ["up", "down"]:
+        if updown not in scales:
+            scales[updown] = default_scale
+
+        # eg {"down": 1} to {"down": {"block_1": 1, "block_2": 1}}}
+        if not isinstance(scales[updown], dict):
+            scales[updown] = {f"block_{i}": copy.deepcopy(scales[updown]) for i in blocks_with_transformer[updown]}
+
+        # eg {"down": {"block_1": 1}} to {"down": {"block_1": [1, 1]}}
+        for i in blocks_with_transformer[updown]:
+            block = f"block_{i}"
+            # set not assigned blocks to default scale
+            if block not in scales[updown]:
+                scales[updown][block] = default_scale
+            if not isinstance(scales[updown][block], list):
+                scales[updown][block] = [scales[updown][block] for _ in range(transformer_per_block[updown])]
+            elif len(scales[updown][block]) == 1:
+                # a list specifying scale to each masked IP input
+                scales[updown][block] = scales[updown][block] * transformer_per_block[updown]
+            elif len(scales[updown][block]) != transformer_per_block[updown]:
+                raise ValueError(
+                    f"Expected {transformer_per_block[updown]} scales for {updown}.{block}, got {len(scales[updown][block])}."
+                )
+
+        # eg {"down": "block_1": [1, 1]}}  to {"down.block_1.0": 1, "down.block_1.1": 1}
+        for i in blocks_with_transformer[updown]:
+            block = f"block_{i}"
+            for tf_idx, value in enumerate(scales[updown][block]):
+                scales[f"{updown}.{block}.{tf_idx}"] = value
+
+        del scales[updown]
+
+    for layer in scales.keys():
+        if not any(_translate_into_actual_layer_name(layer) in module for module in state_dict.keys()):
+            raise ValueError(
+                f"Can't set lora scale for layer {layer}. It either doesn't exist in this unet or it has no attentions."
+            )
+
+    return {_translate_into_actual_layer_name(name): weight for name, weight in scales.items()}

diffusers/models/__init__.py

@@ -27,17 +27,51 @@ _import_structure = {}
 if is_torch_available():
     _import_structure["adapter"] = ["MultiAdapter", "T2IAdapter"]
     _import_structure["autoencoders.autoencoder_asym_kl"] = ["AsymmetricAutoencoderKL"]
+    _import_structure["autoencoders.autoencoder_dc"] = ["AutoencoderDC"]
     _import_structure["autoencoders.autoencoder_kl"] = ["AutoencoderKL"]
+    _import_structure["autoencoders.autoencoder_kl_allegro"] = ["AutoencoderKLAllegro"]
+    _import_structure["autoencoders.autoencoder_kl_cogvideox"] = ["AutoencoderKLCogVideoX"]
+    _import_structure["autoencoders.autoencoder_kl_hunyuan_video"] = ["AutoencoderKLHunyuanVideo"]
+    _import_structure["autoencoders.autoencoder_kl_ltx"] = ["AutoencoderKLLTXVideo"]
+    _import_structure["autoencoders.autoencoder_kl_mochi"] = ["AutoencoderKLMochi"]
     _import_structure["autoencoders.autoencoder_kl_temporal_decoder"] = ["AutoencoderKLTemporalDecoder"]
+    _import_structure["autoencoders.autoencoder_oobleck"] = ["AutoencoderOobleck"]
     _import_structure["autoencoders.autoencoder_tiny"] = ["AutoencoderTiny"]
     _import_structure["autoencoders.consistency_decoder_vae"] = ["ConsistencyDecoderVAE"]
-    _import_structure["controlnet"] = ["ControlNetModel"]
-    _import_structure["dual_transformer_2d"] = ["DualTransformer2DModel"]
+    _import_structure["autoencoders.vq_model"] = ["VQModel"]
+    _import_structure["controlnets.controlnet"] = ["ControlNetModel"]
+    _import_structure["controlnets.controlnet_flux"] = ["FluxControlNetModel", "FluxMultiControlNetModel"]
+    _import_structure["controlnets.controlnet_hunyuan"] = [
+        "HunyuanDiT2DControlNetModel",
+        "HunyuanDiT2DMultiControlNetModel",
+    ]
+    _import_structure["controlnets.controlnet_sd3"] = ["SD3ControlNetModel", "SD3MultiControlNetModel"]
+    _import_structure["controlnets.controlnet_sparsectrl"] = ["SparseControlNetModel"]
+    _import_structure["controlnets.controlnet_union"] = ["ControlNetUnionModel"]
+    _import_structure["controlnets.controlnet_xs"] = ["ControlNetXSAdapter", "UNetControlNetXSModel"]
+    _import_structure["controlnets.multicontrolnet"] = ["MultiControlNetModel"]
     _import_structure["embeddings"] = ["ImageProjection"]
     _import_structure["modeling_utils"] = ["ModelMixin"]
+    _import_structure["transformers.auraflow_transformer_2d"] = ["AuraFlowTransformer2DModel"]
+    _import_structure["transformers.cogvideox_transformer_3d"] = ["CogVideoXTransformer3DModel"]
+    _import_structure["transformers.dit_transformer_2d"] = ["DiTTransformer2DModel"]
+    _import_structure["transformers.dual_transformer_2d"] = ["DualTransformer2DModel"]
+    _import_structure["transformers.hunyuan_transformer_2d"] = ["HunyuanDiT2DModel"]
+    _import_structure["transformers.latte_transformer_3d"] = ["LatteTransformer3DModel"]
+    _import_structure["transformers.lumina_nextdit2d"] = ["LuminaNextDiT2DModel"]
+    _import_structure["transformers.pixart_transformer_2d"] = ["PixArtTransformer2DModel"]
     _import_structure["transformers.prior_transformer"] = ["PriorTransformer"]
+    _import_structure["transformers.sana_transformer"] = ["SanaTransformer2DModel"]
+    _import_structure["transformers.stable_audio_transformer"] = ["StableAudioDiTModel"]
     _import_structure["transformers.t5_film_transformer"] = ["T5FilmDecoder"]
     _import_structure["transformers.transformer_2d"] = ["Transformer2DModel"]
+    _import_structure["transformers.transformer_allegro"] = ["AllegroTransformer3DModel"]
+    _import_structure["transformers.transformer_cogview3plus"] = ["CogView3PlusTransformer2DModel"]
+    _import_structure["transformers.transformer_flux"] = ["FluxTransformer2DModel"]
+    _import_structure["transformers.transformer_hunyuan_video"] = ["HunyuanVideoTransformer3DModel"]
+    _import_structure["transformers.transformer_ltx"] = ["LTXVideoTransformer3DModel"]
+    _import_structure["transformers.transformer_mochi"] = ["MochiTransformer3DModel"]
+    _import_structure["transformers.transformer_sd3"] = ["SD3Transformer2DModel"]
     _import_structure["transformers.transformer_temporal"] = ["TransformerTemporalModel"]
     _import_structure["unets.unet_1d"] = ["UNet1DModel"]
     _import_structure["unets.unet_2d"] = ["UNet2DModel"]
@@ -49,10 +83,9 @@ if is_torch_available():
     _import_structure["unets.unet_spatio_temporal_condition"] = ["UNetSpatioTemporalConditionModel"]
     _import_structure["unets.unet_stable_cascade"] = ["StableCascadeUNet"]
     _import_structure["unets.uvit_2d"] = ["UVit2DModel"]
-    _import_structure["vq_model"] = ["VQModel"]
 
 if is_flax_available():
-    _import_structure["controlnet_flax"] = ["FlaxControlNetModel"]
+    _import_structure["controlnets.controlnet_flax"] = ["FlaxControlNetModel"]
     _import_structure["unets.unet_2d_condition_flax"] = ["FlaxUNet2DConditionModel"]
     _import_structure["vae_flax"] = ["FlaxAutoencoderKL"]
 
@@ -62,17 +95,54 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
         from .adapter import MultiAdapter, T2IAdapter
         from .autoencoders import (
             AsymmetricAutoencoderKL,
+            AutoencoderDC,
             AutoencoderKL,
+            AutoencoderKLAllegro,
+            AutoencoderKLCogVideoX,
+            AutoencoderKLHunyuanVideo,
+            AutoencoderKLLTXVideo,
+            AutoencoderKLMochi,
             AutoencoderKLTemporalDecoder,
+            AutoencoderOobleck,
             AutoencoderTiny,
             ConsistencyDecoderVAE,
+            VQModel,
+        )
+        from .controlnets import (
+            ControlNetModel,
+            ControlNetUnionModel,
+            ControlNetXSAdapter,
+            FluxControlNetModel,
+            FluxMultiControlNetModel,
+            HunyuanDiT2DControlNetModel,
+            HunyuanDiT2DMultiControlNetModel,
+            MultiControlNetModel,
+            SD3ControlNetModel,
+            SD3MultiControlNetModel,
+            SparseControlNetModel,
+            UNetControlNetXSModel,
         )
-        from .controlnet import ControlNetModel
         from .embeddings import ImageProjection
         from .modeling_utils import ModelMixin
         from .transformers import (
+            AllegroTransformer3DModel,
+            AuraFlowTransformer2DModel,
+            CogVideoXTransformer3DModel,
+            CogView3PlusTransformer2DModel,
+            DiTTransformer2DModel,
             DualTransformer2DModel,
+            FluxTransformer2DModel,
+            HunyuanDiT2DModel,
+            HunyuanVideoTransformer3DModel,
+            LatteTransformer3DModel,
+            LTXVideoTransformer3DModel,
+            LuminaNextDiT2DModel,
+            MochiTransformer3DModel,
+            PixArtTransformer2DModel,
             PriorTransformer,
+            SanaTransformer2DModel,
+            SD3Transformer2DModel,
+            StableAudioDiTModel,
             T5FilmDecoder,
             Transformer2DModel,
             TransformerTemporalModel,
@@ -90,10 +160,9 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             UNetSpatioTemporalConditionModel,
             UVit2DModel,
         )
-        from .vq_model import VQModel
 
     if is_flax_available():
-        from .controlnet_flax import FlaxControlNetModel
+        from .controlnets import FlaxControlNetModel
         from .unets import FlaxUNet2DConditionModel
         from .vae_flax import FlaxAutoencoderKL
 

diffusers/models/activations.py

@@ -18,8 +18,12 @@ import torch.nn.functional as F
 from torch import nn
 
 from ..utils import deprecate
+from ..utils.import_utils import is_torch_npu_available, is_torch_version
 
 
+if is_torch_npu_available():
+    import torch_npu
+
 ACTIVATION_FUNCTIONS = {
     "swish": nn.SiLU(),
     "silu": nn.SiLU(),
@@ -46,6 +50,18 @@ def get_activation(act_fn: str) -> nn.Module:
         raise ValueError(f"Unsupported activation function: {act_fn}")
 
 
+class FP32SiLU(nn.Module):
+    r"""
+    SiLU activation function with input upcasted to torch.float32.
+    """
+
+    def __init__(self):
+        super().__init__()
+
+    def forward(self, inputs: torch.Tensor) -> torch.Tensor:
+        return F.silu(inputs.float(), inplace=False).to(inputs.dtype)
+
+
 class GELU(nn.Module):
     r"""
     GELU activation function with tanh approximation support with `approximate="tanh"`.
@@ -63,10 +79,10 @@ class GELU(nn.Module):
         self.approximate = approximate
 
     def gelu(self, gate: torch.Tensor) -> torch.Tensor:
-        if gate.device.type != "mps":
-            return F.gelu(gate, approximate=self.approximate)
-        # mps: gelu is not implemented for float16
-        return F.gelu(gate.to(dtype=torch.float32), approximate=self.approximate).to(dtype=gate.dtype)
+        if gate.device.type == "mps" and is_torch_version("<", "2.0.0"):
+            # fp16 gelu not supported on mps before torch 2.0
+            return F.gelu(gate.to(dtype=torch.float32), approximate=self.approximate).to(dtype=gate.dtype)
+        return F.gelu(gate, approximate=self.approximate)
 
     def forward(self, hidden_states):
         hidden_states = self.proj(hidden_states)
@@ -89,18 +105,45 @@ class GEGLU(nn.Module):
         self.proj = nn.Linear(dim_in, dim_out * 2, bias=bias)
 
     def gelu(self, gate: torch.Tensor) -> torch.Tensor:
-        if gate.device.type != "mps":
-            return F.gelu(gate)
-        # mps: gelu is not implemented for float16
-        return F.gelu(gate.to(dtype=torch.float32)).to(dtype=gate.dtype)
+        if gate.device.type == "mps" and is_torch_version("<", "2.0.0"):
+            # fp16 gelu not supported on mps before torch 2.0
+            return F.gelu(gate.to(dtype=torch.float32)).to(dtype=gate.dtype)
+        return F.gelu(gate)
 
     def forward(self, hidden_states, *args, **kwargs):
         if len(args) > 0 or kwargs.get("scale", None) is not None:
             deprecation_message = "The `scale` argument is deprecated and will be ignored. Please remove it, as passing it will raise an error in the future. `scale` should directly be passed while calling the underlying pipeline component i.e., via `cross_attention_kwargs`."
             deprecate("scale", "1.0.0", deprecation_message)
+        hidden_states = self.proj(hidden_states)
+        if is_torch_npu_available():
+            # using torch_npu.npu_geglu can run faster and save memory on NPU.
+            return torch_npu.npu_geglu(hidden_states, dim=-1, approximate=1)[0]
+        else:
+            hidden_states, gate = hidden_states.chunk(2, dim=-1)
+            return hidden_states * self.gelu(gate)
 
-        hidden_states, gate = self.proj(hidden_states).chunk(2, dim=-1)
-        return hidden_states * self.gelu(gate)
+
+class SwiGLU(nn.Module):
+    r"""
+    A [variant](https://arxiv.org/abs/2002.05202) of the gated linear unit activation function. It's similar to `GEGLU`
+    but uses SiLU / Swish instead of GeLU.
+
+    Parameters:
+        dim_in (`int`): The number of channels in the input.
+        dim_out (`int`): The number of channels in the output.
+        bias (`bool`, defaults to True): Whether to use a bias in the linear layer.
+    """
+
+    def __init__(self, dim_in: int, dim_out: int, bias: bool = True):
+        super().__init__()
+
+        self.proj = nn.Linear(dim_in, dim_out * 2, bias=bias)
+        self.activation = nn.SiLU()
+
+    def forward(self, hidden_states):
+        hidden_states = self.proj(hidden_states)
+        hidden_states, gate = hidden_states.chunk(2, dim=-1)
+        return hidden_states * self.activation(gate)
 
 
 class ApproximateGELU(nn.Module):
@@ -121,3 +164,15 @@ class ApproximateGELU(nn.Module):
     def forward(self, x: torch.Tensor) -> torch.Tensor:
         x = self.proj(x)
         return x * torch.sigmoid(1.702 * x)
+
+
+class LinearActivation(nn.Module):
+    def __init__(self, dim_in: int, dim_out: int, bias: bool = True, activation: str = "silu"):
+        super().__init__()
+
+        self.proj = nn.Linear(dim_in, dim_out, bias=bias)
+        self.activation = get_activation(activation)
+
+    def forward(self, hidden_states):
+        hidden_states = self.proj(hidden_states)
+        return self.activation(hidden_states)

diffusers/models/adapter.py

@@ -30,10 +30,10 @@ class MultiAdapter(ModelMixin):
     MultiAdapter is a wrapper model that contains multiple adapter models and merges their outputs according to
     user-assigned weighting.
 
-    This model inherits from [`ModelMixin`]. Check the superclass documentation for the generic methods the library
-    implements for all the model (such as downloading or saving, etc.)
+    This model inherits from [`ModelMixin`]. Check the superclass documentation for common methods such as downloading
+    or saving.
 
-    Parameters:
+    Args:
         adapters (`List[T2IAdapter]`, *optional*, defaults to None):
             A list of `T2IAdapter` model instances.
     """
@@ -77,11 +77,13 @@ class MultiAdapter(ModelMixin):
         r"""
         Args:
             xs (`torch.Tensor`):
-                (batch, channel, height, width) input images for multiple adapter models concated along dimension 1,
-                `channel` should equal to `num_adapter` * "number of channel of image".
+                A tensor of shape (batch, channel, height, width) representing input images for multiple adapter
+                models, concatenated along dimension 1(channel dimension). The `channel` dimension should be equal to
+                `num_adapter` * number of channel per image.
+
             adapter_weights (`List[float]`, *optional*, defaults to None):
-                List of floats representing the weight which will be multiply to each adapter's output before adding
-                them together.
+                A list of floats representing the weights which will be multiplied by each adapter's output before
+                summing them together. If `None`, equal weights will be used for all adapters.
         """
         if adapter_weights is None:
             adapter_weights = torch.tensor([1 / self.num_adapter] * self.num_adapter)
@@ -109,24 +111,24 @@ class MultiAdapter(ModelMixin):
         variant: Optional[str] = None,
     ):
         """
-        Save a model and its configuration file to a directory, so that it can be re-loaded using the
+        Save a model and its configuration file to a specified directory, allowing it to be re-loaded with the
         `[`~models.adapter.MultiAdapter.from_pretrained`]` class method.
 
-        Arguments:
+        Args:
             save_directory (`str` or `os.PathLike`):
-                Directory to which to save. Will be created if it doesn't exist.
-            is_main_process (`bool`, *optional*, defaults to `True`):
-                Whether the process calling this is the main process or not. Useful when in distributed training like
-                TPUs and need to call this function on all processes. In this case, set `is_main_process=True` only on
-                the main process to avoid race conditions.
+                The directory where the model will be saved. If the directory does not exist, it will be created.
+            is_main_process (`bool`, optional, defaults=True):
+                Indicates whether current process is the main process or not. Useful for distributed training (e.g.,
+                TPUs) and need to call this function on all processes. In this case, set `is_main_process=True` only
+                for the main process to avoid race conditions.
             save_function (`Callable`):
-                The function to use to save the state dictionary. Useful on distributed training like TPUs when one
-                need to replace `torch.save` by another method. Can be configured with the environment variable
-                `DIFFUSERS_SAVE_MODE`.
-            safe_serialization (`bool`, *optional*, defaults to `True`):
-                Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+                Function used to save the state dictionary. Useful for distributed training (e.g., TPUs) to replace
+                `torch.save` with another method. Can also be configured using`DIFFUSERS_SAVE_MODE` environment
+                variable.
+            safe_serialization (`bool`, optional, defaults=True):
+                If `True`, save the model using `safetensors`. If `False`, save the model with `pickle`.
             variant (`str`, *optional*):
-                If specified, weights are saved in the format pytorch_model.<variant>.bin.
+                If specified, weights are saved in the format `pytorch_model.<variant>.bin`.
         """
         idx = 0
         model_path_to_save = save_directory
@@ -145,19 +147,17 @@ class MultiAdapter(ModelMixin):
     @classmethod
     def from_pretrained(cls, pretrained_model_path: Optional[Union[str, os.PathLike]], **kwargs):
         r"""
-        Instantiate a pretrained MultiAdapter model from multiple pre-trained adapter models.
+        Instantiate a pretrained `MultiAdapter` model from multiple pre-trained adapter models.
 
         The model is set in evaluation mode by default using `model.eval()` (Dropout modules are deactivated). To train
-        the model, you should first set it back in training mode with `model.train()`.
+        the model, set it back to training mode using `model.train()`.
 
-        The warning *Weights from XXX not initialized from pretrained model* means that the weights of XXX do not come
-        pretrained with the rest of the model. It is up to you to train those weights with a downstream fine-tuning
-        task.
+        Warnings:
+            *Weights from XXX not initialized from pretrained model* means that the weights of XXX are not pretrained
+            with the rest of the model. It is up to you to train those weights with a downstream fine-tuning. *Weights
+            from XXX not used in YYY* means that the layer XXX is not used by YYY, so those weights are discarded.
 
-        The warning *Weights from XXX not used in YYY* means that the layer XXX is not used by YYY, therefore those
-        weights are discarded.
-
-        Parameters:
+        Args:
             pretrained_model_path (`os.PathLike`):
                 A path to a *directory* containing model weights saved using
                 [`~diffusers.models.adapter.MultiAdapter.save_pretrained`], e.g., `./my_model_directory/adapter`.
@@ -175,20 +175,20 @@ class MultiAdapter(ModelMixin):
                 more information about each option see [designing a device
                 map](https://hf.co/docs/accelerate/main/en/usage_guides/big_modeling#designing-a-device-map).
             max_memory (`Dict`, *optional*):
-                A dictionary device identifier to maximum memory. Will default to the maximum memory available for each
-                GPU and the available CPU RAM if unset.
+                A dictionary mapping device identifiers to their maximum memory. Default to the maximum memory
+                available for each GPU and the available CPU RAM if unset.
             low_cpu_mem_usage (`bool`, *optional*, defaults to `True` if torch version >= 1.9.0 else `False`):
                 Speed up model loading by not initializing the weights and only loading the pre-trained weights. This
                 also tries to not use more than 1x model size in CPU memory (including peak memory) while loading the
                 model. This is only supported when torch version >= 1.9.0. If you are using an older version of torch,
                 setting this argument to `True` will raise an error.
             variant (`str`, *optional*):
-                If specified load weights from `variant` filename, *e.g.* pytorch_model.<variant>.bin. `variant` is
-                ignored when using `from_flax`.
+                If specified, load weights from a `variant` file (*e.g.* pytorch_model.<variant>.bin). `variant` will
+                be ignored when using `from_flax`.
             use_safetensors (`bool`, *optional*, defaults to `None`):
-                If set to `None`, the `safetensors` weights will be downloaded if they're available **and** if the
-                `safetensors` library is installed. If set to `True`, the model will be forcibly loaded from
-                `safetensors` weights. If set to `False`, loading will *not* use `safetensors`.
+                If `None`, the `safetensors` weights will be downloaded if available **and** if`safetensors` library is
+                installed. If `True`, the model will be forcibly loaded from`safetensors` weights. If `False`,
+                `safetensors` is not used.
         """
         idx = 0
         adapters = []
@@ -223,22 +223,22 @@ class T2IAdapter(ModelMixin, ConfigMixin):
      and
      [AdapterLight](https://github.com/TencentARC/T2I-Adapter/blob/686de4681515662c0ac2ffa07bf5dda83af1038a/ldm/modules/encoders/adapter.py#L235).
 
-    This model inherits from [`ModelMixin`]. Check the superclass documentation for the generic methods the library
-    implements for all the model (such as downloading or saving, etc.)
+    This model inherits from [`ModelMixin`]. Check the superclass documentation for the common methods, such as
+    downloading or saving.
 
-    Parameters:
-        in_channels (`int`, *optional*, defaults to 3):
-            Number of channels of Aapter's input(*control image*). Set this parameter to 1 if you're using gray scale
-            image as *control image*.
+    Args:
+        in_channels (`int`, *optional*, defaults to `3`):
+            The number of channels in the adapter's input (*control image*). Set it to 1 if you're using a gray scale
+            image.
         channels (`List[int]`, *optional*, defaults to `(320, 640, 1280, 1280)`):
-            The number of channel of each downsample block's output hidden state. The `len(block_out_channels)` will
-            also determine the number of downsample blocks in the Adapter.
-        num_res_blocks (`int`, *optional*, defaults to 2):
+            The number of channels in each downsample block's output hidden state. The `len(block_out_channels)`
+            determines the number of downsample blocks in the adapter.
+        num_res_blocks (`int`, *optional*, defaults to `2`):
             Number of ResNet blocks in each downsample block.
-        downscale_factor (`int`, *optional*, defaults to 8):
+        downscale_factor (`int`, *optional*, defaults to `8`):
             A factor that determines the total downscale factor of the Adapter.
         adapter_type (`str`, *optional*, defaults to `full_adapter`):
-            The type of Adapter to use. Choose either `full_adapter` or `full_adapter_xl` or `light_adapter`.
+            Adapter type (`full_adapter` or `full_adapter_xl` or `light_adapter`) to use.
     """
 
     @register_to_config
@@ -393,7 +393,7 @@ class AdapterBlock(nn.Module):
     An AdapterBlock is a helper model that contains multiple ResNet-like blocks. It is used in the `FullAdapter` and
     `FullAdapterXL` models.
 
-    Parameters:
+    Args:
         in_channels (`int`):
             Number of channels of AdapterBlock's input.
         out_channels (`int`):
@@ -401,7 +401,7 @@ class AdapterBlock(nn.Module):
         num_res_blocks (`int`):
             Number of ResNet blocks in the AdapterBlock.
         down (`bool`, *optional*, defaults to `False`):
-            Whether to perform downsampling on AdapterBlock's input.
+            If `True`, perform downsampling on AdapterBlock's input.
     """
 
     def __init__(self, in_channels: int, out_channels: int, num_res_blocks: int, down: bool = False):
@@ -440,7 +440,7 @@ class AdapterResnetBlock(nn.Module):
     r"""
     An `AdapterResnetBlock` is a helper model that implements a ResNet-like block.
 
-    Parameters:
+    Args:
         channels (`int`):
             Number of channels of AdapterResnetBlock's input and output.
     """
@@ -518,7 +518,7 @@ class LightAdapterBlock(nn.Module):
     A `LightAdapterBlock` is a helper model that contains multiple `LightAdapterResnetBlocks`. It is used in the
     `LightAdapter` model.
 
-    Parameters:
+    Args:
         in_channels (`int`):
             Number of channels of LightAdapterBlock's input.
         out_channels (`int`):
@@ -526,7 +526,7 @@ class LightAdapterBlock(nn.Module):
         num_res_blocks (`int`):
             Number of LightAdapterResnetBlocks in the LightAdapterBlock.
         down (`bool`, *optional*, defaults to `False`):
-            Whether to perform downsampling on LightAdapterBlock's input.
+            If `True`, perform downsampling on LightAdapterBlock's input.
     """
 
     def __init__(self, in_channels: int, out_channels: int, num_res_blocks: int, down: bool = False):
@@ -561,7 +561,7 @@ class LightAdapterResnetBlock(nn.Module):
     A `LightAdapterResnetBlock` is a helper model that implements a ResNet-like block with a slightly different
     architecture than `AdapterResnetBlock`.
 
-    Parameters:
+    Args:
         channels (`int`):
             Number of channels of LightAdapterResnetBlock's input and output.
     """