AI/diffusers
1
0
Fork 0
mirror of https://github.com/huggingface/diffusers.git synced 2026-01-27 17:22:53 +03:00
Code Activity

Continuation of #1035 (#1120)

Browse Source 
* remove batch size from repeat

* repeat empty string if uncond_tokens is none

* fix inpaint pipes

* return back whitespace to pass code quality

* Apply suggestions from code review

* Fix typos.

Co-authored-by: Had <had-95@yandex.ru>
This commit is contained in:
Pedro Cuenca
2022-11-03 15:49:20 +01:00
committed by GitHub
parent d38c804320
commit 269109dbfb
7 changed files with 18 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
examples/community/interpolate_stable_diffusion.py
View File

@@ -278,7 +278,7 @@ class StableDiffusionWalkPipeline(DiffusionPipeline):
if do_classifier_free_guidance:
uncond_tokens: List[str]
if negative_prompt is None:
uncond_tokens = [""]
uncond_tokens = [""] * batch_size
elif type(prompt) is not type(negative_prompt):
raise TypeError(
f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
@@ -307,7 +307,7 @@ class StableDiffusionWalkPipeline(DiffusionPipeline):
# duplicate unconditional embeddings for each generation per prompt, using mps friendly method
seq_len = uncond_embeddings.shape[1]
uncond_embeddings = uncond_embeddings.repeat(batch_size, num_images_per_prompt, 1)
uncond_embeddings = uncond_embeddings.repeat(1, num_images_per_prompt, 1)
uncond_embeddings = uncond_embeddings.view(batch_size * num_images_per_prompt, seq_len, -1)
# For classifier free guidance, we need to do two forward passes.

4
examples/community/speech_to_image_diffusion.py
View File

@@ -148,7 +148,7 @@ class SpeechToImagePipeline(DiffusionPipeline):
if do_classifier_free_guidance:
uncond_tokens: List[str]
if negative_prompt is None:
uncond_tokens = [""]
uncond_tokens = [""] * batch_size
elif type(prompt) is not type(negative_prompt):
raise TypeError(
f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
@@ -177,7 +177,7 @@ class SpeechToImagePipeline(DiffusionPipeline):
# duplicate unconditional embeddings for each generation per prompt, using mps friendly method
seq_len = uncond_embeddings.shape[1]
uncond_embeddings = uncond_embeddings.repeat(batch_size, num_images_per_prompt, 1)
uncond_embeddings = uncond_embeddings.repeat(1, num_images_per_prompt, 1)
uncond_embeddings = uncond_embeddings.view(batch_size * num_images_per_prompt, seq_len, -1)
# For classifier free guidance, we need to do two forward passes.

4
examples/community/wildcard_stable_diffusion.py
View File

@@ -295,7 +295,7 @@ class WildcardStableDiffusionPipeline(DiffusionPipeline):
if do_classifier_free_guidance:
uncond_tokens: List[str]
if negative_prompt is None:
uncond_tokens = [""]
uncond_tokens = [""] * batch_size
elif type(prompt) is not type(negative_prompt):
raise TypeError(
f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
@@ -324,7 +324,7 @@ class WildcardStableDiffusionPipeline(DiffusionPipeline):
# duplicate unconditional embeddings for each generation per prompt, using mps friendly method
seq_len = uncond_embeddings.shape[1]
uncond_embeddings = uncond_embeddings.repeat(batch_size, num_images_per_prompt, 1)
uncond_embeddings = uncond_embeddings.repeat(1, num_images_per_prompt, 1)
uncond_embeddings = uncond_embeddings.view(batch_size * num_images_per_prompt, seq_len, -1)
# For classifier free guidance, we need to do two forward passes.

4
src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py
View File

@@ -297,7 +297,7 @@ class StableDiffusionPipeline(DiffusionPipeline):
if do_classifier_free_guidance:
uncond_tokens: List[str]
if negative_prompt is None:
uncond_tokens = [""]
uncond_tokens = [""] * batch_size
elif type(prompt) is not type(negative_prompt):
raise TypeError(
f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
@@ -326,7 +326,7 @@ class StableDiffusionPipeline(DiffusionPipeline):
# duplicate unconditional embeddings for each generation per prompt, using mps friendly method
seq_len = uncond_embeddings.shape[1]
uncond_embeddings = uncond_embeddings.repeat(batch_size, num_images_per_prompt, 1)
uncond_embeddings = uncond_embeddings.repeat(1, num_images_per_prompt, 1)
uncond_embeddings = uncond_embeddings.view(batch_size * num_images_per_prompt, seq_len, -1)
# For classifier free guidance, we need to do two forward passes.

6
src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py
View File

@@ -295,7 +295,7 @@ class StableDiffusionImg2ImgPipeline(DiffusionPipeline):
if do_classifier_free_guidance:
uncond_tokens: List[str]
if negative_prompt is None:
uncond_tokens = [""]
uncond_tokens = [""] * batch_size
elif type(prompt) is not type(negative_prompt):
raise TypeError(
f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
@@ -319,7 +319,9 @@ class StableDiffusionImg2ImgPipeline(DiffusionPipeline):
uncond_embeddings = self.text_encoder(uncond_input.input_ids.to(self.device))[0]
# duplicate unconditional embeddings for each generation per prompt
uncond_embeddings = uncond_embeddings.repeat_interleave(batch_size * num_images_per_prompt, dim=0)
seq_len = uncond_embeddings.shape[1]
uncond_embeddings = uncond_embeddings.repeat(1, num_images_per_prompt, 1)
uncond_embeddings = uncond_embeddings.view(batch_size * num_images_per_prompt, seq_len, -1)
# For classifier free guidance, we need to do two forward passes.
# Here we concatenate the unconditional and text embeddings into a single batch

4
src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py
View File

@@ -302,7 +302,7 @@ class StableDiffusionInpaintPipeline(DiffusionPipeline):
if do_classifier_free_guidance:
uncond_tokens: List[str]
if negative_prompt is None:
uncond_tokens = [""]
uncond_tokens = [""] * batch_size
elif type(prompt) is not type(negative_prompt):
raise TypeError(
f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
@@ -331,7 +331,7 @@ class StableDiffusionInpaintPipeline(DiffusionPipeline):
# duplicate unconditional embeddings for each generation per prompt, using mps friendly method
seq_len = uncond_embeddings.shape[1]
uncond_embeddings = uncond_embeddings.repeat(batch_size, num_images_per_prompt, 1)
uncond_embeddings = uncond_embeddings.repeat(1, num_images_per_prompt, 1)
uncond_embeddings = uncond_embeddings.view(batch_size * num_images_per_prompt, seq_len, -1)
# For classifier free guidance, we need to do two forward passes.

6
src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint_legacy.py
View File

@@ -284,7 +284,7 @@ class StableDiffusionInpaintPipelineLegacy(DiffusionPipeline):
if do_classifier_free_guidance:
uncond_tokens: List[str]
if negative_prompt is None:
uncond_tokens = [""]
uncond_tokens = [""] * batch_size
elif type(prompt) is not type(negative_prompt):
raise TypeError(
f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
@@ -312,7 +312,9 @@ class StableDiffusionInpaintPipelineLegacy(DiffusionPipeline):
uncond_embeddings = self.text_encoder(uncond_input.input_ids.to(self.device))[0]
# duplicate unconditional embeddings for each generation per prompt
uncond_embeddings = uncond_embeddings.repeat_interleave(batch_size * num_images_per_prompt, dim=0)
seq_len = uncond_embeddings.shape[1]
uncond_embeddings = uncond_embeddings.repeat(1, num_images_per_prompt, 1)
uncond_embeddings = uncond_embeddings.view(batch_size * num_images_per_prompt, seq_len, -1)
# For classifier free guidance, we need to do two forward passes.
# Here we concatenate the unconditional and text embeddings into a single batch