do not automatically enable xformers (#1640)

* do not automatically enable xformers

* uP

examples/dreambooth/train_dreambooth.py

@@ -17,6 +17,7 @@ from accelerate.utils import set_seed
 from diffusers import AutoencoderKL, DDPMScheduler, DiffusionPipeline, UNet2DConditionModel
 from diffusers.optimization import get_scheduler
 from diffusers.utils import check_min_version
+from diffusers.utils.import_utils import is_xformers_available
 from huggingface_hub import HfFolder, Repository, whoami
 from PIL import Image
 from torchvision import transforms
@@ -488,6 +489,15 @@ def main(args):
         revision=args.revision,
     )
 
+    if is_xformers_available():
+        try:
+            unet.enable_xformers_memory_efficient_attention(True)
+        except Exception as e:
+            logger.warning(
+                "Could not enable memory efficient attention. Make sure xformers is installed"
+                f" correctly and a GPU is available: {e}"
+            )
+
     vae.requires_grad_(False)
     if not args.train_text_encoder:
         text_encoder.requires_grad_(False)

examples/text_to_image/train_text_to_image.py

@@ -18,6 +18,7 @@ from datasets import load_dataset
 from diffusers import AutoencoderKL, DDPMScheduler, StableDiffusionPipeline, UNet2DConditionModel
 from diffusers.optimization import get_scheduler
 from diffusers.utils import check_min_version
+from diffusers.utils.import_utils import is_xformers_available
 from huggingface_hub import HfFolder, Repository, whoami
 from torchvision import transforms
 from tqdm.auto import tqdm
@@ -364,6 +365,15 @@ def main():
         revision=args.revision,
     )
 
+    if is_xformers_available():
+        try:
+            unet.enable_xformers_memory_efficient_attention(True)
+        except Exception as e:
+            logger.warning(
+                "Could not enable memory efficient attention. Make sure xformers is installed"
+                f" correctly and a GPU is available: {e}"
+            )
+
     # Freeze vae and text_encoder
     vae.requires_grad_(False)
     text_encoder.requires_grad_(False)

examples/textual_inversion/textual_inversion.py

@@ -20,6 +20,7 @@ from diffusers import AutoencoderKL, DDPMScheduler, PNDMScheduler, StableDiffusi
 from diffusers.optimization import get_scheduler
 from diffusers.pipelines.stable_diffusion import StableDiffusionSafetyChecker
 from diffusers.utils import check_min_version
+from diffusers.utils.import_utils import is_xformers_available
 from huggingface_hub import HfFolder, Repository, whoami
 
 # TODO: remove and import from diffusers.utils when the new version of diffusers is released
@@ -439,6 +440,15 @@ def main():
         revision=args.revision,
     )
 
+    if is_xformers_available():
+        try:
+            unet.enable_xformers_memory_efficient_attention(True)
+        except Exception as e:
+            logger.warning(
+                "Could not enable memory efficient attention. Make sure xformers is installed"
+                f" correctly and a GPU is available: {e}"
+            )
+
     # Resize the token embeddings as we are adding new special tokens to the tokenizer
     text_encoder.resize_token_embeddings(len(tokenizer))
 

src/diffusers/models/attention.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import math
-import warnings
 from dataclasses import dataclass
 from typing import Optional
 
@@ -447,16 +446,6 @@ class BasicTransformerBlock(nn.Module):
         # 3. Feed-forward
         self.norm3 = nn.LayerNorm(dim)
 
-        # if xformers is installed try to use memory_efficient_attention by default
-        if is_xformers_available():
-            try:
-                self.set_use_memory_efficient_attention_xformers(True)
-            except Exception as e:
-                warnings.warn(
-                    "Could not enable memory efficient attention. Make sure xformers is installed"
-                    f" correctly and a GPU is available: {e}"
-                )
-
     def set_use_memory_efficient_attention_xformers(self, use_memory_efficient_attention_xformers: bool):
         if not is_xformers_available():
             print("Here is how to install it")