sayakpaul HF staff commited on
Commit
0abaa8a
1 Parent(s): 16d8d11
Files changed (1) hide show
  1. unet/my_unet_model.py +3 -3
unet/my_unet_model.py CHANGED
@@ -34,7 +34,7 @@ from diffusers.models.embeddings import (
34
  ImageHintTimeEmbedding,
35
  ImageProjection,
36
  ImageTimeEmbedding,
37
- PositionNet,
38
  TextImageProjection,
39
  TextImageTimeEmbedding,
40
  TextTimeEmbedding,
@@ -42,7 +42,7 @@ from diffusers.models.embeddings import (
42
  Timesteps,
43
  )
44
  from diffusers.models.modeling_utils import ModelMixin
45
- from diffusers.models.unet_2d_blocks import (
46
  UNetMidBlock2DCrossAttn,
47
  UNetMidBlock2DSimpleCrossAttn,
48
  get_down_block,
@@ -586,7 +586,7 @@ class MyUNetModel(ModelMixin, ConfigMixin, UNet2DConditionLoadersMixin):
586
  positive_len = cross_attention_dim[0]
587
 
588
  feature_type = "text-only" if attention_type == "gated" else "text-image"
589
- self.position_net = PositionNet(
590
  positive_len=positive_len, out_dim=cross_attention_dim, feature_type=feature_type
591
  )
592
 
 
34
  ImageHintTimeEmbedding,
35
  ImageProjection,
36
  ImageTimeEmbedding,
37
+ GLIGENTextBoundingboxProjection,
38
  TextImageProjection,
39
  TextImageTimeEmbedding,
40
  TextTimeEmbedding,
 
42
  Timesteps,
43
  )
44
  from diffusers.models.modeling_utils import ModelMixin
45
+ from diffusers.models.unets.unet_2d_blocks import (
46
  UNetMidBlock2DCrossAttn,
47
  UNetMidBlock2DSimpleCrossAttn,
48
  get_down_block,
 
586
  positive_len = cross_attention_dim[0]
587
 
588
  feature_type = "text-only" if attention_type == "gated" else "text-image"
589
+ self.position_net = GLIGENTextBoundingboxProjection(
590
  positive_len=positive_len, out_dim=cross_attention_dim, feature_type=feature_type
591
  )
592