Fix StableDiffusionPipeline when `text_encoder=None` (#8297)

* update

* update

---------
Co-authored-by: Sayak Paul <spsayakpaul@gmail.com>

src/diffusers/pipelines/animatediff/pipeline_animatediff.py

@@ -316,6 +316,7 @@ class AnimateDiffPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
+        if self.text_encoder is not None:
             if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)

src/diffusers/pipelines/animatediff/pipeline_animatediff_video2video.py

@@ -420,6 +420,7 @@ class AnimateDiffVideoToVideoPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
+        if self.text_encoder is not None:
             if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)

src/diffusers/pipelines/controlnet/pipeline_controlnet.py

@@ -463,6 +463,7 @@ class StableDiffusionControlNetPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
+        if self.text_encoder is not None:
             if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)

src/diffusers/pipelines/controlnet/pipeline_controlnet_img2img.py

@@ -441,6 +441,7 @@ class StableDiffusionControlNetImg2ImgPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
+        if self.text_encoder is not None:
             if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)

src/diffusers/pipelines/controlnet/pipeline_controlnet_inpaint.py

@@ -566,6 +566,7 @@ class StableDiffusionControlNetInpaintPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
+        if self.text_encoder is not None:
             if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)

src/diffusers/pipelines/controlnet_xs/pipeline_controlnet_xs.py

@@ -390,6 +390,7 @@ class StableDiffusionControlNetXSPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
+        if self.text_encoder is not None:
             if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)

src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_cycle_diffusion.py

@@ -456,6 +456,7 @@ class CycleDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lor
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
+        if self.text_encoder is not None:
             if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)

src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_inpaint_legacy.py

@@ -426,6 +426,7 @@ class StableDiffusionInpaintPipelineLegacy(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
+        if self.text_encoder is not None:
             if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)

src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_model_editing.py

@@ -364,6 +364,7 @@ class StableDiffusionModelEditingPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
+        if self.text_encoder is not None:
             if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)

src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_paradigms.py

@@ -355,6 +355,7 @@ class StableDiffusionParadigmsPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
+        if self.text_encoder is not None:
             if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)

src/diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_pix2pix_zero.py

@@ -578,6 +578,7 @@ class StableDiffusionPix2PixZeroPipeline(DiffusionPipeline, StableDiffusionMixin
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
+        if self.text_encoder is not None:
             if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)

src/diffusers/pipelines/latent_consistency_models/pipeline_latent_consistency_img2img.py

@@ -405,6 +405,7 @@ class LatentConsistencyModelImg2ImgPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
+        if self.text_encoder is not None:
             if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)

src/diffusers/pipelines/latent_consistency_models/pipeline_latent_consistency_text2img.py

@@ -389,6 +389,7 @@ class LatentConsistencyModelPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
+        if self.text_encoder is not None:
             if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)

src/diffusers/pipelines/pia/pipeline_pia.py

@@ -375,6 +375,7 @@ class PIAPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
+        if self.text_encoder is not None:
             if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py

@@ -474,6 +474,7 @@ class StableDiffusionPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
+        if self.text_encoder is not None:
             if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_depth2img.py

@@ -357,6 +357,7 @@ class StableDiffusionDepth2ImgPipeline(DiffusionPipeline, TextualInversionLoader
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
+        if self.text_encoder is not None:
             if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py

@@ -517,6 +517,7 @@ class StableDiffusionImg2ImgPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
+        if self.text_encoder is not None:
             if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py

@@ -589,6 +589,7 @@ class StableDiffusionInpaintPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
+        if self.text_encoder is not None:
             if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_upscale.py

@@ -377,6 +377,7 @@ class StableDiffusionUpscalePipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
+        if self.text_encoder is not None:
             if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)

src/diffusers/pipelines/stable_diffusion/pipeline_stable_unclip.py

@@ -458,6 +458,7 @@ class StableUnCLIPPipeline(DiffusionPipeline, StableDiffusionMixin, TextualInver
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
+        if self.text_encoder is not None:
             if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)