Rename attention (huggingface#2691)

* rename file

* rename attention

* fix more

* rename more

* up

* more deprecation imports

* fixes

Author: patrickvonplaten

docs/source/en/optimization/torch2.0.mdx

@@ -50,7 +50,7 @@ pip install --pre torch torchvision --index-url https://download.pytorch.org/whl
     ```Python
     import torch
     from diffusers import StableDiffusionPipeline
-    from diffusers.models.cross_attention import AttnProcessor2_0
+    from diffusers.models.attention_processor import AttnProcessor2_0
 
     pipe = StableDiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5", torch_dtype=torch.float16).to("cuda")
     pipe.unet.set_attn_processor(AttnProcessor2_0())

examples/community/stable_diffusion_controlnet_img2img.py

@@ -713,7 +713,7 @@ def __call__(
                 The frequency at which the `callback` function will be called. If not specified, the callback will be
                 called at every step.
             cross_attention_kwargs (`dict`, *optional*):
-                A kwargs dictionary that if specified is passed along to the `AttnProcessor` as defined under
+                A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under
                 `self.processor` in
                 [diffusers.cross_attention](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/cross_attention.py).
             controlnet_conditioning_scale (`float`, *optional*, defaults to 1.0):

examples/community/stable_diffusion_controlnet_inpaint.py

@@ -868,7 +868,7 @@ def __call__(
                 The frequency at which the `callback` function will be called. If not specified, the callback will be
                 called at every step.
             cross_attention_kwargs (`dict`, *optional*):
-                A kwargs dictionary that if specified is passed along to the `AttnProcessor` as defined under
+                A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under
                 `self.processor` in
                 [diffusers.cross_attention](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/cross_attention.py).
             controlnet_conditioning_scale (`float`, *optional*, defaults to 1.0):

examples/community/stable_diffusion_controlnet_inpaint_img2img.py

@@ -911,7 +911,7 @@ def __call__(
                 The frequency at which the `callback` function will be called. If not specified, the callback will be
                 called at every step.
             cross_attention_kwargs (`dict`, *optional*):
-                A kwargs dictionary that if specified is passed along to the `AttnProcessor` as defined under
+                A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under
                 `self.processor` in
                 [diffusers.cross_attention](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/cross_attention.py).
             controlnet_conditioning_scale (`float`, *optional*, defaults to 1.0):

examples/dreambooth/train_dreambooth_lora.py

@@ -47,7 +47,7 @@
     UNet2DConditionModel,
 )
 from diffusers.loaders import AttnProcsLayers
-from diffusers.models.cross_attention import LoRACrossAttnProcessor
+from diffusers.models.attention_processor import LoRAAttnProcessor
 from diffusers.optimization import get_scheduler
 from diffusers.utils import check_min_version, is_wandb_available
 from diffusers.utils.import_utils import is_xformers_available
@@ -723,9 +723,7 @@ def main(args):
             block_id = int(name[len("down_blocks.")])
             hidden_size = unet.config.block_out_channels[block_id]
 
-        lora_attn_procs[name] = LoRACrossAttnProcessor(
-            hidden_size=hidden_size, cross_attention_dim=cross_attention_dim
-        )
+        lora_attn_procs[name] = LoRAAttnProcessor(hidden_size=hidden_size, cross_attention_dim=cross_attention_dim)
 
     unet.set_attn_processor(lora_attn_procs)
     lora_layers = AttnProcsLayers(unet.attn_processors)

examples/research_projects/dreambooth_inpaint/train_dreambooth_inpaint_lora.py

@@ -22,7 +22,7 @@
 
 from diffusers import AutoencoderKL, DDPMScheduler, StableDiffusionInpaintPipeline, UNet2DConditionModel
 from diffusers.loaders import AttnProcsLayers
-from diffusers.models.cross_attention import LoRACrossAttnProcessor
+from diffusers.models.attention_processor import LoRAAttnProcessor
 from diffusers.optimization import get_scheduler
 from diffusers.utils import check_min_version
 from diffusers.utils.import_utils import is_xformers_available
@@ -561,9 +561,7 @@ def main():
             block_id = int(name[len("down_blocks.")])
             hidden_size = unet.config.block_out_channels[block_id]
 
-        lora_attn_procs[name] = LoRACrossAttnProcessor(
-            hidden_size=hidden_size, cross_attention_dim=cross_attention_dim
-        )
+        lora_attn_procs[name] = LoRAAttnProcessor(hidden_size=hidden_size, cross_attention_dim=cross_attention_dim)
 
     unet.set_attn_processor(lora_attn_procs)
     lora_layers = AttnProcsLayers(unet.attn_processors)

examples/research_projects/lora/train_text_to_image_lora.py

@@ -43,7 +43,7 @@
 import diffusers
 from diffusers import AutoencoderKL, DDPMScheduler, DiffusionPipeline, UNet2DConditionModel
 from diffusers.loaders import AttnProcsLayers
-from diffusers.models.cross_attention import LoRACrossAttnProcessor
+from diffusers.models.attention_processor import LoRAAttnProcessor
 from diffusers.optimization import get_scheduler
 from diffusers.utils import check_min_version, is_wandb_available
 from diffusers.utils.import_utils import is_xformers_available
@@ -536,9 +536,7 @@ def main():
                 block_id = int(name[len("down_blocks.")])
                 hidden_size = unet.config.block_out_channels[block_id]
 
-            lora_attn_procs[name] = LoRACrossAttnProcessor(
-                hidden_size=hidden_size, cross_attention_dim=cross_attention_dim
-            )
+            lora_attn_procs[name] = LoRAAttnProcessor(hidden_size=hidden_size, cross_attention_dim=cross_attention_dim)
 
         unet.set_attn_processor(lora_attn_procs)
         lora_layers = AttnProcsLayers(unet.attn_processors)

examples/text_to_image/train_text_to_image_lora.py

@@ -41,7 +41,7 @@
 import diffusers
 from diffusers import AutoencoderKL, DDPMScheduler, DiffusionPipeline, UNet2DConditionModel
 from diffusers.loaders import AttnProcsLayers
-from diffusers.models.cross_attention import LoRACrossAttnProcessor
+from diffusers.models.attention_processor import LoRAAttnProcessor
 from diffusers.optimization import get_scheduler
 from diffusers.utils import check_min_version, is_wandb_available
 from diffusers.utils.import_utils import is_xformers_available
@@ -474,9 +474,7 @@ def main():
             block_id = int(name[len("down_blocks.")])
             hidden_size = unet.config.block_out_channels[block_id]
 
-        lora_attn_procs[name] = LoRACrossAttnProcessor(
-            hidden_size=hidden_size, cross_attention_dim=cross_attention_dim
-        )
+        lora_attn_procs[name] = LoRAAttnProcessor(hidden_size=hidden_size, cross_attention_dim=cross_attention_dim)
 
     unet.set_attn_processor(lora_attn_procs)
 

src/diffusers/loaders.py

@@ -17,7 +17,7 @@
 
 import torch
 
-from .models.cross_attention import LoRACrossAttnProcessor
+from .models.attention_processor import LoRAAttnProcessor
 from .models.modeling_utils import _get_model_file
 from .utils import DIFFUSERS_CACHE, HF_HUB_OFFLINE, deprecate, is_safetensors_available, logging
 
@@ -207,7 +207,7 @@ def load_attn_procs(self, pretrained_model_name_or_path_or_dict: Union[str, Dict
                 cross_attention_dim = value_dict["to_k_lora.down.weight"].shape[1]
                 hidden_size = value_dict["to_k_lora.up.weight"].shape[0]
 
-                attn_processors[key] = LoRACrossAttnProcessor(
+                attn_processors[key] = LoRAAttnProcessor(
                     hidden_size=hidden_size, cross_attention_dim=cross_attention_dim, rank=rank
                 )
                 attn_processors[key].load_state_dict(value_dict)

src/diffusers/models/attention.py

@@ -19,7 +19,7 @@
 from torch import nn
 
 from ..utils.import_utils import is_xformers_available
-from .cross_attention import CrossAttention
+from .attention_processor import Attention
 from .embeddings import CombinedTimestepLabelEmbeddings
 
 
@@ -220,7 +220,7 @@ def __init__(
             )
 
         # 1. Self-Attn
-        self.attn1 = CrossAttention(
+        self.attn1 = Attention(
             query_dim=dim,
             heads=num_attention_heads,
             dim_head=attention_head_dim,
@@ -234,7 +234,7 @@ def __init__(
 
         # 2. Cross-Attn
         if cross_attention_dim is not None:
-            self.attn2 = CrossAttention(
+            self.attn2 = Attention(
                 query_dim=dim,
                 cross_attention_dim=cross_attention_dim,
                 heads=num_attention_heads,

src/diffusers/models/attention_flax.py

@@ -16,7 +16,7 @@
 import jax.numpy as jnp
 
 
-class FlaxCrossAttention(nn.Module):
+class FlaxAttention(nn.Module):
     r"""
     A Flax multi-head attention module as described in: https://arxiv.org/abs/1706.03762
 
@@ -118,9 +118,9 @@ class FlaxBasicTransformerBlock(nn.Module):
 
     def setup(self):
         # self attention (or cross_attention if only_cross_attention is True)
-        self.attn1 = FlaxCrossAttention(self.dim, self.n_heads, self.d_head, self.dropout, dtype=self.dtype)
+        self.attn1 = FlaxAttention(self.dim, self.n_heads, self.d_head, self.dropout, dtype=self.dtype)
         # cross attention
-        self.attn2 = FlaxCrossAttention(self.dim, self.n_heads, self.d_head, self.dropout, dtype=self.dtype)
+        self.attn2 = FlaxAttention(self.dim, self.n_heads, self.d_head, self.dropout, dtype=self.dtype)
         self.ff = FlaxFeedForward(dim=self.dim, dropout=self.dropout, dtype=self.dtype)
         self.norm1 = nn.LayerNorm(epsilon=1e-5, dtype=self.dtype)
         self.norm2 = nn.LayerNorm(epsilon=1e-5, dtype=self.dtype)