Datasets:

ArXiv:
diffusers-benchmarking-bot commited on
Commit
2247fa2
·
verified ·
1 Parent(s): 759ec77

Upload folder using huggingface_hub

Browse files
Files changed (1) hide show
  1. main/mixture_tiling_sdxl.py +22 -22
main/mixture_tiling_sdxl.py CHANGED
@@ -1,4 +1,4 @@
1
- # Copyright 2025 The HuggingFace Team. All rights reserved.
2
  #
3
  # Licensed under the Apache License, Version 2.0 (the "License");
4
  # you may not use this file except in compliance with the License.
@@ -1070,32 +1070,32 @@ class StableDiffusionXLTilingPipeline(
1070
  text_encoder_projection_dim = int(pooled_prompt_embeds.shape[-1])
1071
  else:
1072
  text_encoder_projection_dim = self.text_encoder_2.config.projection_dim
1073
- add_time_ids = self._get_add_time_ids(
1074
- original_size,
1075
- crops_coords_top_left[row][col],
1076
- target_size,
 
 
 
 
 
 
 
 
1077
  dtype=prompt_embeds.dtype,
1078
  text_encoder_projection_dim=text_encoder_projection_dim,
1079
  )
1080
- if negative_original_size is not None and negative_target_size is not None:
1081
- negative_add_time_ids = self._get_add_time_ids(
1082
- negative_original_size,
1083
- negative_crops_coords_top_left[row][col],
1084
- negative_target_size,
1085
- dtype=prompt_embeds.dtype,
1086
- text_encoder_projection_dim=text_encoder_projection_dim,
1087
- )
1088
- else:
1089
- negative_add_time_ids = add_time_ids
1090
 
1091
- if self.do_classifier_free_guidance:
1092
- prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds], dim=0)
1093
- add_text_embeds = torch.cat([negative_pooled_prompt_embeds, add_text_embeds], dim=0)
1094
- add_time_ids = torch.cat([negative_add_time_ids, add_time_ids], dim=0)
1095
 
1096
- prompt_embeds = prompt_embeds.to(device)
1097
- add_text_embeds = add_text_embeds.to(device)
1098
- add_time_ids = add_time_ids.to(device).repeat(batch_size * num_images_per_prompt, 1)
1099
  addition_embed_type_row.append((prompt_embeds, add_text_embeds, add_time_ids))
1100
  embeddings_and_added_time.append(addition_embed_type_row)
1101
 
 
1
+ # Copyright 2025 The DEVAIEXP Team and The HuggingFace Team. All rights reserved.
2
  #
3
  # Licensed under the Apache License, Version 2.0 (the "License");
4
  # you may not use this file except in compliance with the License.
 
1070
  text_encoder_projection_dim = int(pooled_prompt_embeds.shape[-1])
1071
  else:
1072
  text_encoder_projection_dim = self.text_encoder_2.config.projection_dim
1073
+ add_time_ids = self._get_add_time_ids(
1074
+ original_size,
1075
+ crops_coords_top_left[row][col],
1076
+ target_size,
1077
+ dtype=prompt_embeds.dtype,
1078
+ text_encoder_projection_dim=text_encoder_projection_dim,
1079
+ )
1080
+ if negative_original_size is not None and negative_target_size is not None:
1081
+ negative_add_time_ids = self._get_add_time_ids(
1082
+ negative_original_size,
1083
+ negative_crops_coords_top_left[row][col],
1084
+ negative_target_size,
1085
  dtype=prompt_embeds.dtype,
1086
  text_encoder_projection_dim=text_encoder_projection_dim,
1087
  )
1088
+ else:
1089
+ negative_add_time_ids = add_time_ids
 
 
 
 
 
 
 
 
1090
 
1091
+ if self.do_classifier_free_guidance:
1092
+ prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds], dim=0)
1093
+ add_text_embeds = torch.cat([negative_pooled_prompt_embeds, add_text_embeds], dim=0)
1094
+ add_time_ids = torch.cat([negative_add_time_ids, add_time_ids], dim=0)
1095
 
1096
+ prompt_embeds = prompt_embeds.to(device)
1097
+ add_text_embeds = add_text_embeds.to(device)
1098
+ add_time_ids = add_time_ids.to(device).repeat(batch_size * num_images_per_prompt, 1)
1099
  addition_embed_type_row.append((prompt_embeds, add_text_embeds, add_time_ids))
1100
  embeddings_and_added_time.append(addition_embed_type_row)
1101