Skip to content

fix for #7365, prevent pipelines from overriding provided prompt embeds #7926

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 15 commits into from
Jan 8, 2025
Merged

fix for #7365, prevent pipelines from overriding provided prompt embeds #7926

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
15 commits
Select commit Hold shift + click to select a range
037a3d6
fix for #7365, prevent pipelines from overriding provided prompt embeds
May 12, 2024
03f8626
Merge branch 'main' into issue/7365b
bghira May 16, 2024
79f08b5
Merge branch 'main' into issue/7365b
bghira May 20, 2024
ffab8f5
Merge branch 'main' into issue/7365b
bghira May 26, 2024
f8e33e9
Merge branch 'main' into issue/7365b
a-r-r-o-w Nov 23, 2024
197d8a7
fix-copies
sayakpaul Nov 23, 2024
0ffece9
Merge branch 'main' into issue/7365b
bghira Nov 23, 2024
c8fb43c
Merge branch 'main' into issue/7365b
a-r-r-o-w Nov 25, 2024
b7d6792
Merge branch 'main' into issue/7365b
bghira Dec 19, 2024
559660b
Merge branch 'main' into issue/7365b
a-r-r-o-w Dec 29, 2024
8eddfc5
fix implementation
a-r-r-o-w Dec 29, 2024
907a87a
update
a-r-r-o-w Dec 29, 2024
686bb98
Merge branch 'main' into issue/7365b
a-r-r-o-w Dec 29, 2024
1c7fd37
Merge branch 'main' into issue/7365b
a-r-r-o-w Jan 4, 2025
fa38877
Merge branch 'main' into issue/7365b
a-r-r-o-w Jan 7, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 5 additions & 2 deletions examples/community/lpw_stable_diffusion_xl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -827,7 +827,9 @@ def encode_prompt(
)

# We are only ALWAYS interested in the pooled output of the final text encoder
pooled_prompt_embeds = prompt_embeds[0]
if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
pooled_prompt_embeds = prompt_embeds[0]

prompt_embeds = prompt_embeds.hidden_states[-2]

prompt_embeds_list.append(prompt_embeds)
Expand Down Expand Up @@ -879,7 +881,8 @@ def encode_prompt(
output_hidden_states=True,
)
# We are only ALWAYS interested in the pooled output of the final text encoder
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]

negative_prompt_embeds_list.append(negative_prompt_embeds)
Expand Down
7 changes: 5 additions & 2 deletions examples/community/pipeline_demofusion_sdxl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -290,7 +290,9 @@ def encode_prompt(
)

# We are only ALWAYS interested in the pooled output of the final text encoder
pooled_prompt_embeds = prompt_embeds[0]
if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
pooled_prompt_embeds = prompt_embeds[0]

prompt_embeds = prompt_embeds.hidden_states[-2]

prompt_embeds_list.append(prompt_embeds)
Expand Down Expand Up @@ -342,7 +344,8 @@ def encode_prompt(
output_hidden_states=True,
)
# We are only ALWAYS interested in the pooled output of the final text encoder
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]

negative_prompt_embeds_list.append(negative_prompt_embeds)
Expand Down
7 changes: 5 additions & 2 deletions examples/community/pipeline_sdxl_style_aligned.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -628,7 +628,9 @@ def encode_prompt(
prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)

# We are only ALWAYS interested in the pooled output of the final text encoder
pooled_prompt_embeds = prompt_embeds[0]
if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
pooled_prompt_embeds = prompt_embeds[0]

if clip_skip is None:
prompt_embeds = prompt_embeds.hidden_states[-2]
else:
Expand Down Expand Up @@ -688,7 +690,8 @@ def encode_prompt(
output_hidden_states=True,
)
# We are only ALWAYS interested in the pooled output of the final text encoder
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]

negative_prompt_embeds_list.append(negative_prompt_embeds)
Expand Down
7 changes: 5 additions & 2 deletions examples/community/pipeline_stable_diffusion_xl_controlnet_adapter.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -359,7 +359,9 @@ def encode_prompt(
prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)

# We are only ALWAYS interested in the pooled output of the final text encoder
pooled_prompt_embeds = prompt_embeds[0]
if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
pooled_prompt_embeds = prompt_embeds[0]

if clip_skip is None:
prompt_embeds = prompt_embeds.hidden_states[-2]
else:
Expand Down Expand Up @@ -419,7 +421,8 @@ def encode_prompt(
output_hidden_states=True,
)
# We are only ALWAYS interested in the pooled output of the final text encoder
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]

negative_prompt_embeds_list.append(negative_prompt_embeds)
Expand Down
7 changes: 5 additions & 2 deletions examples/community/pipeline_stable_diffusion_xl_controlnet_adapter_inpaint.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -507,7 +507,9 @@ def encode_prompt(
prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)

# We are only ALWAYS interested in the pooled output of the final text encoder
pooled_prompt_embeds = prompt_embeds[0]
if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
pooled_prompt_embeds = prompt_embeds[0]

if clip_skip is None:
prompt_embeds = prompt_embeds.hidden_states[-2]
else:
Expand Down Expand Up @@ -567,7 +569,8 @@ def encode_prompt(
output_hidden_states=True,
)
# We are only ALWAYS interested in the pooled output of the final text encoder
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]

negative_prompt_embeds_list.append(negative_prompt_embeds)
Expand Down
7 changes: 5 additions & 2 deletions examples/community/pipeline_stable_diffusion_xl_differential_img2img.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -394,7 +394,9 @@ def encode_prompt(
prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)

# We are only ALWAYS interested in the pooled output of the final text encoder
pooled_prompt_embeds = prompt_embeds[0]
if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
pooled_prompt_embeds = prompt_embeds[0]

if clip_skip is None:
prompt_embeds = prompt_embeds.hidden_states[-2]
else:
Expand Down Expand Up @@ -454,7 +456,8 @@ def encode_prompt(
output_hidden_states=True,
)
# We are only ALWAYS interested in the pooled output of the final text encoder
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]

negative_prompt_embeds_list.append(negative_prompt_embeds)
Expand Down
7 changes: 5 additions & 2 deletions examples/community/pipeline_stable_diffusion_xl_ipex.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -390,7 +390,9 @@ def encode_prompt(
prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)

# We are only ALWAYS interested in the pooled output of the final text encoder
pooled_prompt_embeds = prompt_embeds[0]
if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
pooled_prompt_embeds = prompt_embeds[0]

if clip_skip is None:
prompt_embeds = prompt_embeds.hidden_states[-2]
else:
Expand Down Expand Up @@ -450,7 +452,8 @@ def encode_prompt(
output_hidden_states=True,
)
# We are only ALWAYS interested in the pooled output of the final text encoder
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]

negative_prompt_embeds_list.append(negative_prompt_embeds)
Expand Down
8 changes: 6 additions & 2 deletions src/diffusers/pipelines/animatediff/pipeline_animatediff_sdxl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -438,7 +438,9 @@ def encode_prompt(
prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)

# We are only ALWAYS interested in the pooled output of the final text encoder
pooled_prompt_embeds = prompt_embeds[0]
if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
pooled_prompt_embeds = prompt_embeds[0]

if clip_skip is None:
prompt_embeds = prompt_embeds.hidden_states[-2]
else:
Expand Down Expand Up @@ -497,8 +499,10 @@ def encode_prompt(
uncond_input.input_ids.to(device),
output_hidden_states=True,
)

# We are only ALWAYS interested in the pooled output of the final text encoder
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]

negative_prompt_embeds_list.append(negative_prompt_embeds)
Expand Down
8 changes: 6 additions & 2 deletions src/diffusers/pipelines/controlnet/pipeline_controlnet_inpaint_sd_xl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -406,7 +406,9 @@ def encode_prompt(
prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)

# We are only ALWAYS interested in the pooled output of the final text encoder
pooled_prompt_embeds = prompt_embeds[0]
if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
pooled_prompt_embeds = prompt_embeds[0]

if clip_skip is None:
prompt_embeds = prompt_embeds.hidden_states[-2]
else:
Expand Down Expand Up @@ -465,8 +467,10 @@ def encode_prompt(
uncond_input.input_ids.to(device),
output_hidden_states=True,
)

# We are only ALWAYS interested in the pooled output of the final text encoder
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]

negative_prompt_embeds_list.append(negative_prompt_embeds)
Expand Down
8 changes: 6 additions & 2 deletions src/diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -415,7 +415,9 @@ def encode_prompt(
prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)

# We are only ALWAYS interested in the pooled output of the final text encoder
pooled_prompt_embeds = prompt_embeds[0]
if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
pooled_prompt_embeds = prompt_embeds[0]

if clip_skip is None:
prompt_embeds = prompt_embeds.hidden_states[-2]
else:
Expand Down Expand Up @@ -474,8 +476,10 @@ def encode_prompt(
uncond_input.input_ids.to(device),
output_hidden_states=True,
)

# We are only ALWAYS interested in the pooled output of the final text encoder
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]

negative_prompt_embeds_list.append(negative_prompt_embeds)
Expand Down
8 changes: 6 additions & 2 deletions src/diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl_img2img.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -408,7 +408,9 @@ def encode_prompt(
prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)

# We are only ALWAYS interested in the pooled output of the final text encoder
pooled_prompt_embeds = prompt_embeds[0]
if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
pooled_prompt_embeds = prompt_embeds[0]

if clip_skip is None:
prompt_embeds = prompt_embeds.hidden_states[-2]
else:
Expand Down Expand Up @@ -467,8 +469,10 @@ def encode_prompt(
uncond_input.input_ids.to(device),
output_hidden_states=True,
)

# We are only ALWAYS interested in the pooled output of the final text encoder
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]

negative_prompt_embeds_list.append(negative_prompt_embeds)
Expand Down
8 changes: 6 additions & 2 deletions src/diffusers/pipelines/controlnet/pipeline_controlnet_union_inpaint_sd_xl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -388,7 +388,9 @@ def encode_prompt(
prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)

# We are only ALWAYS interested in the pooled output of the final text encoder
pooled_prompt_embeds = prompt_embeds[0]
if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
pooled_prompt_embeds = prompt_embeds[0]

if clip_skip is None:
prompt_embeds = prompt_embeds.hidden_states[-2]
else:
Expand Down Expand Up @@ -447,8 +449,10 @@ def encode_prompt(
uncond_input.input_ids.to(device),
output_hidden_states=True,
)

# We are only ALWAYS interested in the pooled output of the final text encoder
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]

negative_prompt_embeds_list.append(negative_prompt_embeds)
Expand Down
8 changes: 6 additions & 2 deletions src/diffusers/pipelines/controlnet/pipeline_controlnet_union_sd_xl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -397,7 +397,9 @@ def encode_prompt(
prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)

# We are only ALWAYS interested in the pooled output of the final text encoder
pooled_prompt_embeds = prompt_embeds[0]
if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
pooled_prompt_embeds = prompt_embeds[0]

if clip_skip is None:
prompt_embeds = prompt_embeds.hidden_states[-2]
else:
Expand Down Expand Up @@ -456,8 +458,10 @@ def encode_prompt(
uncond_input.input_ids.to(device),
output_hidden_states=True,
)

# We are only ALWAYS interested in the pooled output of the final text encoder
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]

negative_prompt_embeds_list.append(negative_prompt_embeds)
Expand Down
8 changes: 6 additions & 2 deletions src/diffusers/pipelines/controlnet/pipeline_controlnet_union_sd_xl_img2img.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -422,7 +422,9 @@ def encode_prompt(
prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)

# We are only ALWAYS interested in the pooled output of the final text encoder
pooled_prompt_embeds = prompt_embeds[0]
if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
pooled_prompt_embeds = prompt_embeds[0]

if clip_skip is None:
prompt_embeds = prompt_embeds.hidden_states[-2]
else:
Expand Down Expand Up @@ -481,8 +483,10 @@ def encode_prompt(
uncond_input.input_ids.to(device),
output_hidden_states=True,
)

# We are only ALWAYS interested in the pooled output of the final text encoder
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]

negative_prompt_embeds_list.append(negative_prompt_embeds)
Expand Down
8 changes: 6 additions & 2 deletions src/diffusers/pipelines/controlnet_xs/pipeline_controlnet_xs_sd_xl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -336,7 +336,9 @@ def encode_prompt(
prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)

# We are only ALWAYS interested in the pooled output of the final text encoder
pooled_prompt_embeds = prompt_embeds[0]
if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
pooled_prompt_embeds = prompt_embeds[0]

if clip_skip is None:
prompt_embeds = prompt_embeds.hidden_states[-2]
else:
Expand Down Expand Up @@ -395,8 +397,10 @@ def encode_prompt(
uncond_input.input_ids.to(device),
output_hidden_states=True,
)

# We are only ALWAYS interested in the pooled output of the final text encoder
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]

negative_prompt_embeds_list.append(negative_prompt_embeds)
Expand Down
8 changes: 6 additions & 2 deletions src/diffusers/pipelines/pag/pipeline_pag_controlnet_sd_xl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -421,7 +421,9 @@ def encode_prompt(
prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)

# We are only ALWAYS interested in the pooled output of the final text encoder
pooled_prompt_embeds = prompt_embeds[0]
if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
pooled_prompt_embeds = prompt_embeds[0]

if clip_skip is None:
prompt_embeds = prompt_embeds.hidden_states[-2]
else:
Expand Down Expand Up @@ -480,8 +482,10 @@ def encode_prompt(
uncond_input.input_ids.to(device),
output_hidden_states=True,
)

# We are only ALWAYS interested in the pooled output of the final text encoder
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]

negative_prompt_embeds_list.append(negative_prompt_embeds)
Expand Down
Loading
Loading