diffusers 0.23.1__py3-none-any.whl → 0.25.0__py3-none-any.whl

Sign up to get free protection for your applications and to get access to all the features.
Files changed (238) hide show
  1. diffusers/__init__.py +26 -2
  2. diffusers/commands/fp16_safetensors.py +10 -11
  3. diffusers/configuration_utils.py +13 -8
  4. diffusers/dependency_versions_check.py +0 -1
  5. diffusers/dependency_versions_table.py +5 -5
  6. diffusers/experimental/rl/value_guided_sampling.py +1 -1
  7. diffusers/image_processor.py +463 -51
  8. diffusers/loaders/__init__.py +82 -0
  9. diffusers/loaders/ip_adapter.py +159 -0
  10. diffusers/loaders/lora.py +1553 -0
  11. diffusers/loaders/lora_conversion_utils.py +284 -0
  12. diffusers/loaders/single_file.py +637 -0
  13. diffusers/loaders/textual_inversion.py +455 -0
  14. diffusers/loaders/unet.py +828 -0
  15. diffusers/loaders/utils.py +59 -0
  16. diffusers/models/__init__.py +26 -9
  17. diffusers/models/activations.py +9 -6
  18. diffusers/models/attention.py +301 -29
  19. diffusers/models/attention_flax.py +9 -1
  20. diffusers/models/attention_processor.py +378 -6
  21. diffusers/models/autoencoders/__init__.py +5 -0
  22. diffusers/models/{autoencoder_asym_kl.py → autoencoders/autoencoder_asym_kl.py} +17 -12
  23. diffusers/models/{autoencoder_kl.py → autoencoders/autoencoder_kl.py} +47 -23
  24. diffusers/models/autoencoders/autoencoder_kl_temporal_decoder.py +402 -0
  25. diffusers/models/{autoencoder_tiny.py → autoencoders/autoencoder_tiny.py} +24 -28
  26. diffusers/models/{consistency_decoder_vae.py → autoencoders/consistency_decoder_vae.py} +51 -44
  27. diffusers/models/{vae.py → autoencoders/vae.py} +71 -17
  28. diffusers/models/controlnet.py +59 -39
  29. diffusers/models/controlnet_flax.py +19 -18
  30. diffusers/models/downsampling.py +338 -0
  31. diffusers/models/embeddings.py +112 -29
  32. diffusers/models/embeddings_flax.py +2 -0
  33. diffusers/models/lora.py +131 -1
  34. diffusers/models/modeling_flax_utils.py +14 -8
  35. diffusers/models/modeling_outputs.py +17 -0
  36. diffusers/models/modeling_utils.py +37 -29
  37. diffusers/models/normalization.py +110 -4
  38. diffusers/models/resnet.py +299 -652
  39. diffusers/models/transformer_2d.py +22 -5
  40. diffusers/models/transformer_temporal.py +183 -1
  41. diffusers/models/unet_2d_blocks_flax.py +5 -0
  42. diffusers/models/unet_2d_condition.py +46 -0
  43. diffusers/models/unet_2d_condition_flax.py +13 -13
  44. diffusers/models/unet_3d_blocks.py +957 -173
  45. diffusers/models/unet_3d_condition.py +16 -8
  46. diffusers/models/unet_kandinsky3.py +535 -0
  47. diffusers/models/unet_motion_model.py +48 -33
  48. diffusers/models/unet_spatio_temporal_condition.py +489 -0
  49. diffusers/models/upsampling.py +454 -0
  50. diffusers/models/uvit_2d.py +471 -0
  51. diffusers/models/vae_flax.py +7 -0
  52. diffusers/models/vq_model.py +12 -3
  53. diffusers/optimization.py +16 -9
  54. diffusers/pipelines/__init__.py +137 -76
  55. diffusers/pipelines/amused/__init__.py +62 -0
  56. diffusers/pipelines/amused/pipeline_amused.py +328 -0
  57. diffusers/pipelines/amused/pipeline_amused_img2img.py +347 -0
  58. diffusers/pipelines/amused/pipeline_amused_inpaint.py +378 -0
  59. diffusers/pipelines/animatediff/pipeline_animatediff.py +66 -8
  60. diffusers/pipelines/audioldm/pipeline_audioldm.py +1 -0
  61. diffusers/pipelines/auto_pipeline.py +23 -13
  62. diffusers/pipelines/consistency_models/pipeline_consistency_models.py +1 -0
  63. diffusers/pipelines/controlnet/pipeline_controlnet.py +238 -35
  64. diffusers/pipelines/controlnet/pipeline_controlnet_img2img.py +148 -37
  65. diffusers/pipelines/controlnet/pipeline_controlnet_inpaint.py +155 -41
  66. diffusers/pipelines/controlnet/pipeline_controlnet_inpaint_sd_xl.py +123 -43
  67. diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl.py +216 -39
  68. diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl_img2img.py +106 -34
  69. diffusers/pipelines/dance_diffusion/pipeline_dance_diffusion.py +1 -0
  70. diffusers/pipelines/ddim/pipeline_ddim.py +1 -0
  71. diffusers/pipelines/ddpm/pipeline_ddpm.py +1 -0
  72. diffusers/pipelines/deepfloyd_if/pipeline_if.py +13 -1
  73. diffusers/pipelines/deepfloyd_if/pipeline_if_img2img.py +13 -1
  74. diffusers/pipelines/deepfloyd_if/pipeline_if_img2img_superresolution.py +13 -1
  75. diffusers/pipelines/deepfloyd_if/pipeline_if_inpainting.py +13 -1
  76. diffusers/pipelines/deepfloyd_if/pipeline_if_inpainting_superresolution.py +13 -1
  77. diffusers/pipelines/deepfloyd_if/pipeline_if_superresolution.py +13 -1
  78. diffusers/pipelines/deprecated/__init__.py +153 -0
  79. diffusers/pipelines/{alt_diffusion → deprecated/alt_diffusion}/__init__.py +3 -3
  80. diffusers/pipelines/{alt_diffusion → deprecated/alt_diffusion}/pipeline_alt_diffusion.py +177 -34
  81. diffusers/pipelines/{alt_diffusion → deprecated/alt_diffusion}/pipeline_alt_diffusion_img2img.py +182 -37
  82. diffusers/pipelines/{alt_diffusion → deprecated/alt_diffusion}/pipeline_output.py +1 -1
  83. diffusers/pipelines/{audio_diffusion → deprecated/audio_diffusion}/__init__.py +1 -1
  84. diffusers/pipelines/{audio_diffusion → deprecated/audio_diffusion}/mel.py +2 -2
  85. diffusers/pipelines/{audio_diffusion → deprecated/audio_diffusion}/pipeline_audio_diffusion.py +4 -4
  86. diffusers/pipelines/{latent_diffusion_uncond → deprecated/latent_diffusion_uncond}/__init__.py +1 -1
  87. diffusers/pipelines/{latent_diffusion_uncond → deprecated/latent_diffusion_uncond}/pipeline_latent_diffusion_uncond.py +4 -4
  88. diffusers/pipelines/{pndm → deprecated/pndm}/__init__.py +1 -1
  89. diffusers/pipelines/{pndm → deprecated/pndm}/pipeline_pndm.py +4 -4
  90. diffusers/pipelines/{repaint → deprecated/repaint}/__init__.py +1 -1
  91. diffusers/pipelines/{repaint → deprecated/repaint}/pipeline_repaint.py +5 -5
  92. diffusers/pipelines/{score_sde_ve → deprecated/score_sde_ve}/__init__.py +1 -1
  93. diffusers/pipelines/{score_sde_ve → deprecated/score_sde_ve}/pipeline_score_sde_ve.py +5 -4
  94. diffusers/pipelines/{spectrogram_diffusion → deprecated/spectrogram_diffusion}/__init__.py +6 -6
  95. diffusers/pipelines/{spectrogram_diffusion/continous_encoder.py → deprecated/spectrogram_diffusion/continuous_encoder.py} +2 -2
  96. diffusers/pipelines/{spectrogram_diffusion → deprecated/spectrogram_diffusion}/midi_utils.py +1 -1
  97. diffusers/pipelines/{spectrogram_diffusion → deprecated/spectrogram_diffusion}/notes_encoder.py +2 -2
  98. diffusers/pipelines/{spectrogram_diffusion → deprecated/spectrogram_diffusion}/pipeline_spectrogram_diffusion.py +8 -7
  99. diffusers/pipelines/deprecated/stable_diffusion_variants/__init__.py +55 -0
  100. diffusers/pipelines/{stable_diffusion → deprecated/stable_diffusion_variants}/pipeline_cycle_diffusion.py +34 -13
  101. diffusers/pipelines/{stable_diffusion → deprecated/stable_diffusion_variants}/pipeline_onnx_stable_diffusion_inpaint_legacy.py +7 -6
  102. diffusers/pipelines/{stable_diffusion → deprecated/stable_diffusion_variants}/pipeline_stable_diffusion_inpaint_legacy.py +12 -11
  103. diffusers/pipelines/{stable_diffusion → deprecated/stable_diffusion_variants}/pipeline_stable_diffusion_model_editing.py +17 -11
  104. diffusers/pipelines/{stable_diffusion → deprecated/stable_diffusion_variants}/pipeline_stable_diffusion_paradigms.py +11 -10
  105. diffusers/pipelines/{stable_diffusion → deprecated/stable_diffusion_variants}/pipeline_stable_diffusion_pix2pix_zero.py +14 -13
  106. diffusers/pipelines/{stochastic_karras_ve → deprecated/stochastic_karras_ve}/__init__.py +1 -1
  107. diffusers/pipelines/{stochastic_karras_ve → deprecated/stochastic_karras_ve}/pipeline_stochastic_karras_ve.py +4 -4
  108. diffusers/pipelines/{versatile_diffusion → deprecated/versatile_diffusion}/__init__.py +3 -3
  109. diffusers/pipelines/{versatile_diffusion → deprecated/versatile_diffusion}/modeling_text_unet.py +83 -51
  110. diffusers/pipelines/{versatile_diffusion → deprecated/versatile_diffusion}/pipeline_versatile_diffusion.py +4 -4
  111. diffusers/pipelines/{versatile_diffusion → deprecated/versatile_diffusion}/pipeline_versatile_diffusion_dual_guided.py +7 -6
  112. diffusers/pipelines/{versatile_diffusion → deprecated/versatile_diffusion}/pipeline_versatile_diffusion_image_variation.py +7 -6
  113. diffusers/pipelines/{versatile_diffusion → deprecated/versatile_diffusion}/pipeline_versatile_diffusion_text_to_image.py +7 -6
  114. diffusers/pipelines/{vq_diffusion → deprecated/vq_diffusion}/__init__.py +3 -3
  115. diffusers/pipelines/{vq_diffusion → deprecated/vq_diffusion}/pipeline_vq_diffusion.py +5 -5
  116. diffusers/pipelines/dit/pipeline_dit.py +1 -0
  117. diffusers/pipelines/kandinsky2_2/pipeline_kandinsky2_2.py +1 -1
  118. diffusers/pipelines/kandinsky2_2/pipeline_kandinsky2_2_combined.py +3 -3
  119. diffusers/pipelines/kandinsky2_2/pipeline_kandinsky2_2_img2img.py +1 -1
  120. diffusers/pipelines/kandinsky2_2/pipeline_kandinsky2_2_inpainting.py +1 -1
  121. diffusers/pipelines/kandinsky2_2/pipeline_kandinsky2_2_prior.py +1 -1
  122. diffusers/pipelines/kandinsky3/__init__.py +49 -0
  123. diffusers/pipelines/kandinsky3/convert_kandinsky3_unet.py +98 -0
  124. diffusers/pipelines/kandinsky3/pipeline_kandinsky3.py +589 -0
  125. diffusers/pipelines/kandinsky3/pipeline_kandinsky3_img2img.py +654 -0
  126. diffusers/pipelines/latent_consistency_models/pipeline_latent_consistency_img2img.py +111 -11
  127. diffusers/pipelines/latent_consistency_models/pipeline_latent_consistency_text2img.py +102 -9
  128. diffusers/pipelines/latent_diffusion/pipeline_latent_diffusion.py +1 -0
  129. diffusers/pipelines/musicldm/pipeline_musicldm.py +1 -1
  130. diffusers/pipelines/onnx_utils.py +8 -5
  131. diffusers/pipelines/paint_by_example/pipeline_paint_by_example.py +7 -2
  132. diffusers/pipelines/pipeline_flax_utils.py +11 -8
  133. diffusers/pipelines/pipeline_utils.py +63 -42
  134. diffusers/pipelines/pixart_alpha/pipeline_pixart_alpha.py +247 -38
  135. diffusers/pipelines/shap_e/pipeline_shap_e_img2img.py +3 -3
  136. diffusers/pipelines/stable_diffusion/__init__.py +37 -65
  137. diffusers/pipelines/stable_diffusion/convert_from_ckpt.py +75 -78
  138. diffusers/pipelines/stable_diffusion/pipeline_flax_stable_diffusion.py +2 -2
  139. diffusers/pipelines/stable_diffusion/pipeline_onnx_stable_diffusion_img2img.py +2 -4
  140. diffusers/pipelines/stable_diffusion/pipeline_onnx_stable_diffusion_inpaint.py +1 -0
  141. diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py +174 -11
  142. diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_depth2img.py +8 -3
  143. diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_image_variation.py +1 -0
  144. diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py +178 -11
  145. diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py +224 -13
  146. diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_instruct_pix2pix.py +74 -20
  147. diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_latent_upscale.py +4 -0
  148. diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_upscale.py +7 -0
  149. diffusers/pipelines/stable_diffusion/pipeline_stable_unclip.py +5 -0
  150. diffusers/pipelines/stable_diffusion/pipeline_stable_unclip_img2img.py +5 -0
  151. diffusers/pipelines/stable_diffusion_attend_and_excite/__init__.py +48 -0
  152. diffusers/pipelines/{stable_diffusion → stable_diffusion_attend_and_excite}/pipeline_stable_diffusion_attend_and_excite.py +6 -2
  153. diffusers/pipelines/stable_diffusion_diffedit/__init__.py +48 -0
  154. diffusers/pipelines/{stable_diffusion → stable_diffusion_diffedit}/pipeline_stable_diffusion_diffedit.py +3 -3
  155. diffusers/pipelines/stable_diffusion_gligen/__init__.py +50 -0
  156. diffusers/pipelines/{stable_diffusion → stable_diffusion_gligen}/pipeline_stable_diffusion_gligen.py +3 -2
  157. diffusers/pipelines/{stable_diffusion → stable_diffusion_gligen}/pipeline_stable_diffusion_gligen_text_image.py +4 -3
  158. diffusers/pipelines/stable_diffusion_k_diffusion/__init__.py +60 -0
  159. diffusers/pipelines/{stable_diffusion → stable_diffusion_k_diffusion}/pipeline_stable_diffusion_k_diffusion.py +7 -1
  160. diffusers/pipelines/stable_diffusion_ldm3d/__init__.py +48 -0
  161. diffusers/pipelines/{stable_diffusion → stable_diffusion_ldm3d}/pipeline_stable_diffusion_ldm3d.py +51 -7
  162. diffusers/pipelines/stable_diffusion_panorama/__init__.py +48 -0
  163. diffusers/pipelines/{stable_diffusion → stable_diffusion_panorama}/pipeline_stable_diffusion_panorama.py +57 -8
  164. diffusers/pipelines/stable_diffusion_safe/pipeline_stable_diffusion_safe.py +58 -6
  165. diffusers/pipelines/stable_diffusion_sag/__init__.py +48 -0
  166. diffusers/pipelines/{stable_diffusion → stable_diffusion_sag}/pipeline_stable_diffusion_sag.py +68 -10
  167. diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl.py +194 -17
  168. diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_img2img.py +205 -16
  169. diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_inpaint.py +206 -17
  170. diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_instruct_pix2pix.py +23 -17
  171. diffusers/pipelines/stable_video_diffusion/__init__.py +58 -0
  172. diffusers/pipelines/stable_video_diffusion/pipeline_stable_video_diffusion.py +652 -0
  173. diffusers/pipelines/t2i_adapter/pipeline_stable_diffusion_adapter.py +108 -12
  174. diffusers/pipelines/t2i_adapter/pipeline_stable_diffusion_xl_adapter.py +115 -14
  175. diffusers/pipelines/text_to_video_synthesis/__init__.py +2 -0
  176. diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_synth.py +6 -0
  177. diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_synth_img2img.py +23 -3
  178. diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_zero.py +334 -10
  179. diffusers/pipelines/text_to_video_synthesis/pipeline_text_to_video_zero_sdxl.py +1331 -0
  180. diffusers/pipelines/unclip/pipeline_unclip.py +2 -1
  181. diffusers/pipelines/unclip/pipeline_unclip_image_variation.py +1 -0
  182. diffusers/pipelines/wuerstchen/modeling_paella_vq_model.py +1 -1
  183. diffusers/pipelines/wuerstchen/modeling_wuerstchen_common.py +14 -4
  184. diffusers/pipelines/wuerstchen/modeling_wuerstchen_prior.py +9 -5
  185. diffusers/pipelines/wuerstchen/pipeline_wuerstchen.py +1 -1
  186. diffusers/pipelines/wuerstchen/pipeline_wuerstchen_combined.py +2 -2
  187. diffusers/pipelines/wuerstchen/pipeline_wuerstchen_prior.py +5 -1
  188. diffusers/schedulers/__init__.py +4 -4
  189. diffusers/schedulers/deprecated/__init__.py +50 -0
  190. diffusers/schedulers/{scheduling_karras_ve.py → deprecated/scheduling_karras_ve.py} +4 -4
  191. diffusers/schedulers/{scheduling_sde_vp.py → deprecated/scheduling_sde_vp.py} +4 -6
  192. diffusers/schedulers/scheduling_amused.py +162 -0
  193. diffusers/schedulers/scheduling_consistency_models.py +2 -0
  194. diffusers/schedulers/scheduling_ddim.py +1 -3
  195. diffusers/schedulers/scheduling_ddim_inverse.py +2 -7
  196. diffusers/schedulers/scheduling_ddim_parallel.py +1 -3
  197. diffusers/schedulers/scheduling_ddpm.py +47 -3
  198. diffusers/schedulers/scheduling_ddpm_parallel.py +47 -3
  199. diffusers/schedulers/scheduling_deis_multistep.py +28 -6
  200. diffusers/schedulers/scheduling_dpmsolver_multistep.py +28 -6
  201. diffusers/schedulers/scheduling_dpmsolver_multistep_inverse.py +28 -6
  202. diffusers/schedulers/scheduling_dpmsolver_sde.py +3 -3
  203. diffusers/schedulers/scheduling_dpmsolver_singlestep.py +28 -6
  204. diffusers/schedulers/scheduling_euler_ancestral_discrete.py +59 -3
  205. diffusers/schedulers/scheduling_euler_discrete.py +102 -16
  206. diffusers/schedulers/scheduling_heun_discrete.py +17 -5
  207. diffusers/schedulers/scheduling_k_dpm_2_ancestral_discrete.py +17 -5
  208. diffusers/schedulers/scheduling_k_dpm_2_discrete.py +17 -5
  209. diffusers/schedulers/scheduling_lcm.py +123 -29
  210. diffusers/schedulers/scheduling_lms_discrete.py +3 -3
  211. diffusers/schedulers/scheduling_pndm.py +1 -3
  212. diffusers/schedulers/scheduling_repaint.py +1 -3
  213. diffusers/schedulers/scheduling_unipc_multistep.py +28 -6
  214. diffusers/schedulers/scheduling_utils.py +3 -1
  215. diffusers/schedulers/scheduling_utils_flax.py +3 -1
  216. diffusers/training_utils.py +1 -1
  217. diffusers/utils/__init__.py +1 -2
  218. diffusers/utils/constants.py +10 -12
  219. diffusers/utils/dummy_pt_objects.py +75 -0
  220. diffusers/utils/dummy_torch_and_transformers_objects.py +105 -0
  221. diffusers/utils/dynamic_modules_utils.py +18 -22
  222. diffusers/utils/export_utils.py +8 -3
  223. diffusers/utils/hub_utils.py +24 -36
  224. diffusers/utils/logging.py +11 -11
  225. diffusers/utils/outputs.py +5 -5
  226. diffusers/utils/peft_utils.py +88 -44
  227. diffusers/utils/state_dict_utils.py +8 -0
  228. diffusers/utils/testing_utils.py +199 -1
  229. diffusers/utils/torch_utils.py +4 -4
  230. {diffusers-0.23.1.dist-info → diffusers-0.25.0.dist-info}/METADATA +86 -69
  231. diffusers-0.25.0.dist-info/RECORD +360 -0
  232. {diffusers-0.23.1.dist-info → diffusers-0.25.0.dist-info}/WHEEL +1 -1
  233. {diffusers-0.23.1.dist-info → diffusers-0.25.0.dist-info}/entry_points.txt +0 -1
  234. diffusers/loaders.py +0 -3336
  235. diffusers-0.23.1.dist-info/RECORD +0 -323
  236. /diffusers/pipelines/{alt_diffusion → deprecated/alt_diffusion}/modeling_roberta_series.py +0 -0
  237. {diffusers-0.23.1.dist-info → diffusers-0.25.0.dist-info}/LICENSE +0 -0
  238. {diffusers-0.23.1.dist-info → diffusers-0.25.0.dist-info}/top_level.txt +0 -0
@@ -0,0 +1,55 @@
1
+ from typing import TYPE_CHECKING
2
+
3
+ from ....utils import (
4
+ DIFFUSERS_SLOW_IMPORT,
5
+ OptionalDependencyNotAvailable,
6
+ _LazyModule,
7
+ get_objects_from_module,
8
+ is_torch_available,
9
+ is_transformers_available,
10
+ )
11
+
12
+
13
+ _dummy_objects = {}
14
+ _import_structure = {}
15
+
16
+ try:
17
+ if not (is_transformers_available() and is_torch_available()):
18
+ raise OptionalDependencyNotAvailable()
19
+ except OptionalDependencyNotAvailable:
20
+ from ....utils import dummy_torch_and_transformers_objects
21
+
22
+ _dummy_objects.update(get_objects_from_module(dummy_torch_and_transformers_objects))
23
+ else:
24
+ _import_structure["pipeline_cycle_diffusion"] = ["CycleDiffusionPipeline"]
25
+ _import_structure["pipeline_stable_diffusion_inpaint_legacy"] = ["StableDiffusionInpaintPipelineLegacy"]
26
+ _import_structure["pipeline_stable_diffusion_model_editing"] = ["StableDiffusionModelEditingPipeline"]
27
+
28
+ _import_structure["pipeline_stable_diffusion_paradigms"] = ["StableDiffusionParadigmsPipeline"]
29
+ _import_structure["pipeline_stable_diffusion_pix2pix_zero"] = ["StableDiffusionPix2PixZeroPipeline"]
30
+
31
+ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
32
+ try:
33
+ if not (is_transformers_available() and is_torch_available()):
34
+ raise OptionalDependencyNotAvailable()
35
+ except OptionalDependencyNotAvailable:
36
+ from ....utils.dummy_torch_and_transformers_objects import *
37
+
38
+ else:
39
+ from .pipeline_cycle_diffusion import CycleDiffusionPipeline
40
+ from .pipeline_stable_diffusion_inpaint_legacy import StableDiffusionInpaintPipelineLegacy
41
+ from .pipeline_stable_diffusion_model_editing import StableDiffusionModelEditingPipeline
42
+ from .pipeline_stable_diffusion_paradigms import StableDiffusionParadigmsPipeline
43
+ from .pipeline_stable_diffusion_pix2pix_zero import StableDiffusionPix2PixZeroPipeline
44
+
45
+ else:
46
+ import sys
47
+
48
+ sys.modules[__name__] = _LazyModule(
49
+ __name__,
50
+ globals()["__file__"],
51
+ _import_structure,
52
+ module_spec=__spec__,
53
+ )
54
+ for name, value in _dummy_objects.items():
55
+ setattr(sys.modules[__name__], name, value)
@@ -21,17 +21,17 @@ import torch
21
21
  from packaging import version
22
22
  from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer
23
23
 
24
- from ...configuration_utils import FrozenDict
25
- from ...image_processor import PipelineImageInput, VaeImageProcessor
26
- from ...loaders import LoraLoaderMixin, TextualInversionLoaderMixin
27
- from ...models import AutoencoderKL, UNet2DConditionModel
28
- from ...models.lora import adjust_lora_scale_text_encoder
29
- from ...schedulers import DDIMScheduler
30
- from ...utils import PIL_INTERPOLATION, USE_PEFT_BACKEND, deprecate, logging, scale_lora_layers, unscale_lora_layers
31
- from ...utils.torch_utils import randn_tensor
32
- from ..pipeline_utils import DiffusionPipeline
33
- from .pipeline_output import StableDiffusionPipelineOutput
34
- from .safety_checker import StableDiffusionSafetyChecker
24
+ from ....configuration_utils import FrozenDict
25
+ from ....image_processor import PipelineImageInput, VaeImageProcessor
26
+ from ....loaders import LoraLoaderMixin, TextualInversionLoaderMixin
27
+ from ....models import AutoencoderKL, UNet2DConditionModel
28
+ from ....models.lora import adjust_lora_scale_text_encoder
29
+ from ....schedulers import DDIMScheduler
30
+ from ....utils import PIL_INTERPOLATION, USE_PEFT_BACKEND, deprecate, logging, scale_lora_layers, unscale_lora_layers
31
+ from ....utils.torch_utils import randn_tensor
32
+ from ...pipeline_utils import DiffusionPipeline
33
+ from ...stable_diffusion.pipeline_output import StableDiffusionPipelineOutput
34
+ from ...stable_diffusion.safety_checker import StableDiffusionSafetyChecker
35
35
 
36
36
 
37
37
  logger = logging.get_logger(__name__) # pylint: disable=invalid-name
@@ -61,6 +61,20 @@ def preprocess(image):
61
61
  return image
62
62
 
63
63
 
64
+ # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
65
+ def retrieve_latents(
66
+ encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
67
+ ):
68
+ if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
69
+ return encoder_output.latent_dist.sample(generator)
70
+ elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
71
+ return encoder_output.latent_dist.mode()
72
+ elif hasattr(encoder_output, "latents"):
73
+ return encoder_output.latents
74
+ else:
75
+ raise AttributeError("Could not access latents of provided encoder_output")
76
+
77
+
64
78
  def posterior_sample(scheduler, latents, timestep, clean_latents, generator, eta):
65
79
  # 1. get previous step value (=t-1)
66
80
  prev_timestep = timestep - scheduler.config.num_train_timesteps // scheduler.num_inference_steps
@@ -129,6 +143,11 @@ class CycleDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lor
129
143
  This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods
130
144
  implemented for all pipelines (downloading, saving, running on a particular device, etc.).
131
145
 
146
+ The pipeline also inherits the following loading methods:
147
+ - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
148
+ - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
149
+ - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
150
+
132
151
  Args:
133
152
  vae ([`AutoencoderKL`]):
134
153
  Variational Auto-Encoder (VAE) model to encode and decode images to and from latent representations.
@@ -148,6 +167,7 @@ class CycleDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lor
148
167
  feature_extractor ([`~transformers.CLIPImageProcessor`]):
149
168
  A `CLIPImageProcessor` to extract features from generated images; used as inputs to the `safety_checker`.
150
169
  """
170
+
151
171
  model_cpu_offload_seq = "text_encoder->unet->vae"
152
172
  _optional_components = ["safety_checker", "feature_extractor"]
153
173
 
@@ -566,11 +586,12 @@ class CycleDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lor
566
586
 
567
587
  if isinstance(generator, list):
568
588
  init_latents = [
569
- self.vae.encode(image[i : i + 1]).latent_dist.sample(generator[i]) for i in range(batch_size)
589
+ retrieve_latents(self.vae.encode(image[i : i + 1]), generator=generator[i])
590
+ for i in range(image.shape[0])
570
591
  ]
571
592
  init_latents = torch.cat(init_latents, dim=0)
572
593
  else:
573
- init_latents = self.vae.encode(image).latent_dist.sample(generator)
594
+ init_latents = retrieve_latents(self.vae.encode(image), generator=generator)
574
595
 
575
596
  init_latents = self.vae.config.scaling_factor * init_latents
576
597
 
@@ -6,12 +6,12 @@ import PIL.Image
6
6
  import torch
7
7
  from transformers import CLIPImageProcessor, CLIPTokenizer
8
8
 
9
- from ...configuration_utils import FrozenDict
10
- from ...schedulers import DDIMScheduler, LMSDiscreteScheduler, PNDMScheduler
11
- from ...utils import deprecate, logging
12
- from ..onnx_utils import ORT_TO_NP_TYPE, OnnxRuntimeModel
13
- from ..pipeline_utils import DiffusionPipeline
14
- from . import StableDiffusionPipelineOutput
9
+ from ....configuration_utils import FrozenDict
10
+ from ....schedulers import DDIMScheduler, LMSDiscreteScheduler, PNDMScheduler
11
+ from ....utils import deprecate, logging
12
+ from ...onnx_utils import ORT_TO_NP_TYPE, OnnxRuntimeModel
13
+ from ...pipeline_utils import DiffusionPipeline
14
+ from ...stable_diffusion.pipeline_output import StableDiffusionPipelineOutput
15
15
 
16
16
 
17
17
  logger = logging.get_logger(__name__) # pylint: disable=invalid-name
@@ -66,6 +66,7 @@ class OnnxStableDiffusionInpaintPipelineLegacy(DiffusionPipeline):
66
66
  feature_extractor ([`CLIPImageProcessor`]):
67
67
  Model that extracts features from generated images to be used as inputs for the `safety_checker`.
68
68
  """
69
+
69
70
  _optional_components = ["safety_checker", "feature_extractor"]
70
71
  _is_onnx = True
71
72
 
@@ -21,17 +21,17 @@ import torch
21
21
  from packaging import version
22
22
  from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer
23
23
 
24
- from ...configuration_utils import FrozenDict
25
- from ...image_processor import VaeImageProcessor
26
- from ...loaders import FromSingleFileMixin, LoraLoaderMixin, TextualInversionLoaderMixin
27
- from ...models import AutoencoderKL, UNet2DConditionModel
28
- from ...models.lora import adjust_lora_scale_text_encoder
29
- from ...schedulers import KarrasDiffusionSchedulers
30
- from ...utils import PIL_INTERPOLATION, USE_PEFT_BACKEND, deprecate, logging, scale_lora_layers, unscale_lora_layers
31
- from ...utils.torch_utils import randn_tensor
32
- from ..pipeline_utils import DiffusionPipeline
33
- from . import StableDiffusionPipelineOutput
34
- from .safety_checker import StableDiffusionSafetyChecker
24
+ from ....configuration_utils import FrozenDict
25
+ from ....image_processor import VaeImageProcessor
26
+ from ....loaders import FromSingleFileMixin, LoraLoaderMixin, TextualInversionLoaderMixin
27
+ from ....models import AutoencoderKL, UNet2DConditionModel
28
+ from ....models.lora import adjust_lora_scale_text_encoder
29
+ from ....schedulers import KarrasDiffusionSchedulers
30
+ from ....utils import PIL_INTERPOLATION, USE_PEFT_BACKEND, deprecate, logging, scale_lora_layers, unscale_lora_layers
31
+ from ....utils.torch_utils import randn_tensor
32
+ from ...pipeline_utils import DiffusionPipeline
33
+ from ...stable_diffusion import StableDiffusionPipelineOutput
34
+ from ...stable_diffusion.safety_checker import StableDiffusionSafetyChecker
35
35
 
36
36
 
37
37
  logger = logging.get_logger(__name__)
@@ -115,6 +115,7 @@ class StableDiffusionInpaintPipelineLegacy(
115
115
  feature_extractor ([`CLIPImageProcessor`]):
116
116
  Model that extracts features from generated images to be used as inputs for the `safety_checker`.
117
117
  """
118
+
118
119
  model_cpu_offload_seq = "text_encoder->unet->vae"
119
120
  _optional_components = ["feature_extractor"]
120
121
  _exclude_from_cpu_offload = ["safety_checker"]
@@ -18,17 +18,17 @@ from typing import Any, Callable, Dict, List, Optional, Union
18
18
  import torch
19
19
  from transformers import CLIPFeatureExtractor, CLIPTextModel, CLIPTokenizer
20
20
 
21
- from ...image_processor import VaeImageProcessor
22
- from ...loaders import LoraLoaderMixin, TextualInversionLoaderMixin
23
- from ...models import AutoencoderKL, UNet2DConditionModel
24
- from ...models.lora import adjust_lora_scale_text_encoder
25
- from ...schedulers import PNDMScheduler
26
- from ...schedulers.scheduling_utils import SchedulerMixin
27
- from ...utils import USE_PEFT_BACKEND, deprecate, logging, scale_lora_layers, unscale_lora_layers
28
- from ...utils.torch_utils import randn_tensor
29
- from ..pipeline_utils import DiffusionPipeline
30
- from . import StableDiffusionPipelineOutput
31
- from .safety_checker import StableDiffusionSafetyChecker
21
+ from ....image_processor import VaeImageProcessor
22
+ from ....loaders import LoraLoaderMixin, TextualInversionLoaderMixin
23
+ from ....models import AutoencoderKL, UNet2DConditionModel
24
+ from ....models.lora import adjust_lora_scale_text_encoder
25
+ from ....schedulers import PNDMScheduler
26
+ from ....schedulers.scheduling_utils import SchedulerMixin
27
+ from ....utils import USE_PEFT_BACKEND, deprecate, logging, scale_lora_layers, unscale_lora_layers
28
+ from ....utils.torch_utils import randn_tensor
29
+ from ...pipeline_utils import DiffusionPipeline
30
+ from ...stable_diffusion.pipeline_output import StableDiffusionPipelineOutput
31
+ from ...stable_diffusion.safety_checker import StableDiffusionSafetyChecker
32
32
 
33
33
 
34
34
  logger = logging.get_logger(__name__) # pylint: disable=invalid-name
@@ -43,6 +43,11 @@ class StableDiffusionModelEditingPipeline(DiffusionPipeline, TextualInversionLoa
43
43
  This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods
44
44
  implemented for all pipelines (downloading, saving, running on a particular device, etc.).
45
45
 
46
+ The pipeline also inherits the following loading methods:
47
+ - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
48
+ - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
49
+ - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
50
+
46
51
  Args:
47
52
  vae ([`AutoencoderKL`]):
48
53
  Variational Auto-Encoder (VAE) model to encode and decode images to and from latent representations.
@@ -66,6 +71,7 @@ class StableDiffusionModelEditingPipeline(DiffusionPipeline, TextualInversionLoa
66
71
  with_augs ([`list`]):
67
72
  Textual augmentations to apply while editing the text-to-image model. Set to `[]` for no augmentations.
68
73
  """
74
+
69
75
  model_cpu_offload_seq = "text_encoder->unet->vae"
70
76
  _optional_components = ["safety_checker", "feature_extractor"]
71
77
  _exclude_from_cpu_offload = ["safety_checker"]
@@ -18,12 +18,12 @@ from typing import Any, Callable, Dict, List, Optional, Union
18
18
  import torch
19
19
  from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer
20
20
 
21
- from ...image_processor import VaeImageProcessor
22
- from ...loaders import FromSingleFileMixin, LoraLoaderMixin, TextualInversionLoaderMixin
23
- from ...models import AutoencoderKL, UNet2DConditionModel
24
- from ...models.lora import adjust_lora_scale_text_encoder
25
- from ...schedulers import KarrasDiffusionSchedulers
26
- from ...utils import (
21
+ from ....image_processor import VaeImageProcessor
22
+ from ....loaders import FromSingleFileMixin, LoraLoaderMixin, TextualInversionLoaderMixin
23
+ from ....models import AutoencoderKL, UNet2DConditionModel
24
+ from ....models.lora import adjust_lora_scale_text_encoder
25
+ from ....schedulers import KarrasDiffusionSchedulers
26
+ from ....utils import (
27
27
  USE_PEFT_BACKEND,
28
28
  deprecate,
29
29
  logging,
@@ -31,10 +31,10 @@ from ...utils import (
31
31
  scale_lora_layers,
32
32
  unscale_lora_layers,
33
33
  )
34
- from ...utils.torch_utils import randn_tensor
35
- from ..pipeline_utils import DiffusionPipeline
36
- from . import StableDiffusionPipelineOutput
37
- from .safety_checker import StableDiffusionSafetyChecker
34
+ from ....utils.torch_utils import randn_tensor
35
+ from ...pipeline_utils import DiffusionPipeline
36
+ from ...stable_diffusion.pipeline_output import StableDiffusionPipelineOutput
37
+ from ...stable_diffusion.safety_checker import StableDiffusionSafetyChecker
38
38
 
39
39
 
40
40
  logger = logging.get_logger(__name__) # pylint: disable=invalid-name
@@ -96,6 +96,7 @@ class StableDiffusionParadigmsPipeline(
96
96
  feature_extractor ([`~transformers.CLIPImageProcessor`]):
97
97
  A `CLIPImageProcessor` to extract features from generated images; used as inputs to the `safety_checker`.
98
98
  """
99
+
99
100
  model_cpu_offload_seq = "text_encoder->unet->vae"
100
101
  _optional_components = ["safety_checker", "feature_extractor"]
101
102
  _exclude_from_cpu_offload = ["safety_checker"]
@@ -28,14 +28,14 @@ from transformers import (
28
28
  CLIPTokenizer,
29
29
  )
30
30
 
31
- from ...image_processor import PipelineImageInput, VaeImageProcessor
32
- from ...loaders import LoraLoaderMixin, TextualInversionLoaderMixin
33
- from ...models import AutoencoderKL, UNet2DConditionModel
34
- from ...models.attention_processor import Attention
35
- from ...models.lora import adjust_lora_scale_text_encoder
36
- from ...schedulers import DDIMScheduler, DDPMScheduler, EulerAncestralDiscreteScheduler, LMSDiscreteScheduler
37
- from ...schedulers.scheduling_ddim_inverse import DDIMInverseScheduler
38
- from ...utils import (
31
+ from ....image_processor import PipelineImageInput, VaeImageProcessor
32
+ from ....loaders import LoraLoaderMixin, TextualInversionLoaderMixin
33
+ from ....models import AutoencoderKL, UNet2DConditionModel
34
+ from ....models.attention_processor import Attention
35
+ from ....models.lora import adjust_lora_scale_text_encoder
36
+ from ....schedulers import DDIMScheduler, DDPMScheduler, EulerAncestralDiscreteScheduler, LMSDiscreteScheduler
37
+ from ....schedulers.scheduling_ddim_inverse import DDIMInverseScheduler
38
+ from ....utils import (
39
39
  PIL_INTERPOLATION,
40
40
  USE_PEFT_BACKEND,
41
41
  BaseOutput,
@@ -45,10 +45,10 @@ from ...utils import (
45
45
  scale_lora_layers,
46
46
  unscale_lora_layers,
47
47
  )
48
- from ...utils.torch_utils import randn_tensor
49
- from ..pipeline_utils import DiffusionPipeline
50
- from . import StableDiffusionPipelineOutput
51
- from .safety_checker import StableDiffusionSafetyChecker
48
+ from ....utils.torch_utils import randn_tensor
49
+ from ...pipeline_utils import DiffusionPipeline
50
+ from ...stable_diffusion.pipeline_output import StableDiffusionPipelineOutput
51
+ from ...stable_diffusion.safety_checker import StableDiffusionSafetyChecker
52
52
 
53
53
 
54
54
  logger = logging.get_logger(__name__) # pylint: disable=invalid-name
@@ -282,7 +282,7 @@ class Pix2PixZeroAttnProcessor:
282
282
 
283
283
  class StableDiffusionPix2PixZeroPipeline(DiffusionPipeline):
284
284
  r"""
285
- Pipeline for pixel-levl image editing using Pix2Pix Zero. Based on Stable Diffusion.
285
+ Pipeline for pixel-level image editing using Pix2Pix Zero. Based on Stable Diffusion.
286
286
 
287
287
  This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods the
288
288
  library implements for all the pipelines (such as downloading or saving, running on a particular device, etc.)
@@ -310,6 +310,7 @@ class StableDiffusionPix2PixZeroPipeline(DiffusionPipeline):
310
310
  Whether the pipeline requires a safety checker. We recommend setting it to True if you're using the
311
311
  pipeline publicly.
312
312
  """
313
+
313
314
  model_cpu_offload_seq = "text_encoder->unet->vae"
314
315
  _optional_components = [
315
316
  "safety_checker",
@@ -1,6 +1,6 @@
1
1
  from typing import TYPE_CHECKING
2
2
 
3
- from ...utils import DIFFUSERS_SLOW_IMPORT, _LazyModule
3
+ from ....utils import DIFFUSERS_SLOW_IMPORT, _LazyModule
4
4
 
5
5
 
6
6
  _import_structure = {"pipeline_stochastic_karras_ve": ["KarrasVePipeline"]}
@@ -16,10 +16,10 @@ from typing import List, Optional, Tuple, Union
16
16
 
17
17
  import torch
18
18
 
19
- from ...models import UNet2DModel
20
- from ...schedulers import KarrasVeScheduler
21
- from ...utils.torch_utils import randn_tensor
22
- from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput
19
+ from ....models import UNet2DModel
20
+ from ....schedulers import KarrasVeScheduler
21
+ from ....utils.torch_utils import randn_tensor
22
+ from ...pipeline_utils import DiffusionPipeline, ImagePipelineOutput
23
23
 
24
24
 
25
25
  class KarrasVePipeline(DiffusionPipeline):
@@ -1,6 +1,6 @@
1
1
  from typing import TYPE_CHECKING
2
2
 
3
- from ...utils import (
3
+ from ....utils import (
4
4
  DIFFUSERS_SLOW_IMPORT,
5
5
  OptionalDependencyNotAvailable,
6
6
  _LazyModule,
@@ -17,7 +17,7 @@ try:
17
17
  if not (is_transformers_available() and is_torch_available() and is_transformers_version(">=", "4.25.0")):
18
18
  raise OptionalDependencyNotAvailable()
19
19
  except OptionalDependencyNotAvailable:
20
- from ...utils.dummy_torch_and_transformers_objects import (
20
+ from ....utils.dummy_torch_and_transformers_objects import (
21
21
  VersatileDiffusionDualGuidedPipeline,
22
22
  VersatileDiffusionImageVariationPipeline,
23
23
  VersatileDiffusionPipeline,
@@ -45,7 +45,7 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
45
45
  if not (is_transformers_available() and is_torch_available() and is_transformers_version(">=", "4.25.0")):
46
46
  raise OptionalDependencyNotAvailable()
47
47
  except OptionalDependencyNotAvailable:
48
- from ...utils.dummy_torch_and_transformers_objects import (
48
+ from ....utils.dummy_torch_and_transformers_objects import (
49
49
  VersatileDiffusionDualGuidedPipeline,
50
50
  VersatileDiffusionImageVariationPipeline,
51
51
  VersatileDiffusionPipeline,