Add draft for lora text encoder scale (#3626)

* Add draft for lora text encoder scale

* Improve naming

* fix: training dreambooth lora script.

* Apply suggestions from code review

* Update examples/dreambooth/train_dreambooth_lora.py

* Apply suggestions from code review

* Apply suggestions from code review

* add lora mixin when fit

* add lora mixin when fit

* add lora mixin when fit

* fix more

* fix more

---------

Co-authored-by: Sayak Paul <[email protected]>

Author: patrickvonplaten

CONTRIBUTING.md

@@ -125,14 +125,14 @@ Awesome! Tell us what problem it solved for you.
 
 You can open a feature request [here](https://github.com/huggingface/diffusers/issues/new?assignees=&labels=&template=feature_request.md&title=).
 
-#### 2.3 Feedback. 
+#### 2.3 Feedback.
 
 Feedback about the library design and why it is good or not good helps the core maintainers immensely to build a user-friendly library. To understand the philosophy behind the current design philosophy, please have a look [here](https://huggingface.co/docs/diffusers/conceptual/philosophy). If you feel like a certain design choice does not fit with the current design philosophy, please explain why and how it should be changed. If a certain design choice follows the design philosophy too much, hence restricting use cases, explain why and how it should be changed.
 If a certain design choice is very useful for you, please also leave a note as this is great feedback for future design decisions.
 
 You can open an issue about feedback [here](https://github.com/huggingface/diffusers/issues/new?assignees=&labels=&template=feedback.md&title=).
 
-#### 2.4 Technical questions. 
+#### 2.4 Technical questions.
 
 Technical questions are mainly about why certain code of the library was written in a certain way, or what a certain part of the code does. Please make sure to link to the code in question and please provide detail on
 why this part of the code is difficult to understand.
@@ -394,8 +394,8 @@ passes. You should run the tests impacted by your changes like this:
  ```bash
  $ pytest tests/<TEST_TO_RUN>.py
  ```
- 
-Before you run the tests, please make sure you install the dependencies required for testing. You can do so 
+
+Before you run the tests, please make sure you install the dependencies required for testing. You can do so
 with this command:
 
  ```bash

PHILOSOPHY.md

@@ -27,18 +27,18 @@ In a nutshell, Diffusers is built to be a natural extension of PyTorch. Therefor
 
 ## Simple over easy
 
-As PyTorch states, **explicit is better than implicit** and **simple is better than complex**. This design philosophy is reflected in multiple parts of the library: 
+As PyTorch states, **explicit is better than implicit** and **simple is better than complex**. This design philosophy is reflected in multiple parts of the library:
 - We follow PyTorch's API with methods like [`DiffusionPipeline.to`](https://huggingface.co/docs/diffusers/main/en/api/diffusion_pipeline#diffusers.DiffusionPipeline.to) to let the user handle device management.
 - Raising concise error messages is preferred to silently correct erroneous input. Diffusers aims at teaching the user, rather than making the library as easy to use as possible.
 - Complex model vs. scheduler logic is exposed instead of magically handled inside. Schedulers/Samplers are separated from diffusion models with minimal dependencies on each other. This forces the user to write the unrolled denoising loop. However, the separation allows for easier debugging and gives the user more control over adapting the denoising process or switching out diffusion models or schedulers.
-- Separately trained components of the diffusion pipeline, *e.g.* the text encoder, the unet, and the variational autoencoder, each have their own model class. This forces the user to handle the interaction between the different model components, and the serialization format separates the model components into different files. However, this allows for easier debugging and customization. Dreambooth or textual inversion training 
+- Separately trained components of the diffusion pipeline, *e.g.* the text encoder, the unet, and the variational autoencoder, each have their own model class. This forces the user to handle the interaction between the different model components, and the serialization format separates the model components into different files. However, this allows for easier debugging and customization. Dreambooth or textual inversion training
 is very simple thanks to diffusers' ability to separate single components of the diffusion pipeline.
 
 ## Tweakable, contributor-friendly over abstraction
 
-For large parts of the library, Diffusers adopts an important design principle of the [Transformers library](https://github.com/huggingface/transformers), which is to prefer copy-pasted code over hasty abstractions. This design principle is very opinionated and stands in stark contrast to popular design principles such as [Don't repeat yourself (DRY)](https://en.wikipedia.org/wiki/Don%27t_repeat_yourself). 
+For large parts of the library, Diffusers adopts an important design principle of the [Transformers library](https://github.com/huggingface/transformers), which is to prefer copy-pasted code over hasty abstractions. This design principle is very opinionated and stands in stark contrast to popular design principles such as [Don't repeat yourself (DRY)](https://en.wikipedia.org/wiki/Don%27t_repeat_yourself).
 In short, just like Transformers does for modeling files, diffusers prefers to keep an extremely low level of abstraction and very self-contained code for pipelines and schedulers.
-Functions, long code blocks, and even classes can be copied across multiple files which at first can look like a bad, sloppy design choice that makes the library unmaintainable. 
+Functions, long code blocks, and even classes can be copied across multiple files which at first can look like a bad, sloppy design choice that makes the library unmaintainable.
 **However**, this design has proven to be extremely successful for Transformers and makes a lot of sense for community-driven, open-source machine learning libraries because:
 - Machine Learning is an extremely fast-moving field in which paradigms, model architectures, and algorithms are changing rapidly, which therefore makes it very difficult to define long-lasting code abstractions.
 - Machine Learning practitioners like to be able to quickly tweak existing code for ideation and research and therefore prefer self-contained code over one that contains many abstractions.
@@ -47,10 +47,10 @@ Functions, long code blocks, and even classes can be copied across multiple file
 At Hugging Face, we call this design the **single-file policy** which means that almost all of the code of a certain class should be written in a single, self-contained file. To read more about the philosophy, you can have a look
 at [this blog post](https://huggingface.co/blog/transformers-design-philosophy).
 
-In diffusers, we follow this philosophy for both pipelines and schedulers, but only partly for diffusion models. The reason we don't follow this design fully for diffusion models is because almost all diffusion pipelines, such 
+In diffusers, we follow this philosophy for both pipelines and schedulers, but only partly for diffusion models. The reason we don't follow this design fully for diffusion models is because almost all diffusion pipelines, such
 as [DDPM](https://huggingface.co/docs/diffusers/v0.12.0/en/api/pipelines/ddpm), [Stable Diffusion](https://huggingface.co/docs/diffusers/v0.12.0/en/api/pipelines/stable_diffusion/overview#stable-diffusion-pipelines), [UnCLIP (Dalle-2)](https://huggingface.co/docs/diffusers/v0.12.0/en/api/pipelines/unclip#overview) and [Imagen](https://imagen.research.google/) all rely on the same diffusion model, the [UNet](https://huggingface.co/docs/diffusers/api/models#diffusers.UNet2DConditionModel).
 
-Great, now you should have generally understood why 🧨 Diffusers is designed the way it is 🤗. 
+Great, now you should have generally understood why 🧨 Diffusers is designed the way it is 🤗.
 We try to apply these design principles consistently across the library. Nevertheless, there are some minor exceptions to the philosophy or some unlucky design choices. If you have feedback regarding the design, we would ❤️  to hear it [directly on GitHub](https://github.com/huggingface/diffusers/issues/new?assignees=&labels=&template=feedback.md&title=).
 
 ## Design Philosophy in Details
@@ -89,17 +89,17 @@ The following design principles are followed:
 - Models should by default have the highest precision and lowest performance setting.
 - To integrate new model checkpoints whose general architecture can be classified as an architecture that already exists in Diffusers, the existing model architecture shall be adapted to make it work with the new checkpoint. One should only create a new file if the model architecture is fundamentally different.
 - Models should be designed to be easily extendable to future changes. This can be achieved by limiting public function arguments, configuration arguments, and "foreseeing" future changes, *e.g.* it is usually better to add `string` "...type" arguments that can easily be extended to new future types instead of boolean `is_..._type` arguments. Only the minimum amount of changes shall be made to existing architectures to make a new model checkpoint work.
-- The model design is a difficult trade-off between keeping code readable and concise and supporting many model checkpoints. For most parts of the modeling code, classes shall be adapted for new model checkpoints, while there are some exceptions where it is preferred to add new classes to make sure the code is kept concise and 
+- The model design is a difficult trade-off between keeping code readable and concise and supporting many model checkpoints. For most parts of the modeling code, classes shall be adapted for new model checkpoints, while there are some exceptions where it is preferred to add new classes to make sure the code is kept concise and
 readable longterm, such as [UNet blocks](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/unet_2d_blocks.py) and [Attention processors](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/cross_attention.py).
 
 ### Schedulers
 
 Schedulers are responsible to guide the denoising process for inference as well as to define a noise schedule for training. They are designed as individual classes with loadable configuration files and strongly follow the **single-file policy**.
 
 The following design principles are followed:
-- All schedulers are found in [`src/diffusers/schedulers`](https://github.com/huggingface/diffusers/tree/main/src/diffusers/schedulers). 
-- Schedulers are **not** allowed to import from large utils files and shall be kept very self-contained. 
-- One scheduler python file corresponds to one scheduler algorithm (as might be defined in a paper). 
+- All schedulers are found in [`src/diffusers/schedulers`](https://github.com/huggingface/diffusers/tree/main/src/diffusers/schedulers).
+- Schedulers are **not** allowed to import from large utils files and shall be kept very self-contained.
+- One scheduler python file corresponds to one scheduler algorithm (as might be defined in a paper).
 - If schedulers share similar functionalities, we can make use of the `#Copied from` mechanism.
 - Schedulers all inherit from `SchedulerMixin` and `ConfigMixin`.
 - Schedulers can be easily swapped out with the [`ConfigMixin.from_config`](https://huggingface.co/docs/diffusers/main/en/api/configuration#diffusers.ConfigMixin.from_config) method as explained in detail [here](./using-diffusers/schedulers.mdx).

README.md

@@ -30,7 +30,7 @@ We recommend installing 🤗 Diffusers in a virtual environment from PyPi or Con
 ### PyTorch
 
 With `pip` (official package):
-    
+
 ```bash
 pip install --upgrade diffusers[torch]
 ```
@@ -107,7 +107,7 @@ Check out the [Quickstart](https://huggingface.co/docs/diffusers/quicktour) to l
 | [Training](https://huggingface.co/docs/diffusers/training/overview) | Guides for how to train a diffusion model for different tasks with different training techniques.                                                                                               |
 ## Contribution
 
-We ❤️  contributions from the open-source community! 
+We ❤️  contributions from the open-source community!
 If you want to contribute to this library, please check out our [Contribution guide](https://github.com/huggingface/diffusers/blob/main/CONTRIBUTING.md).
 You can look out for [issues](https://github.com/huggingface/diffusers/issues) you'd like to tackle to contribute to the library.
 - See [Good first issues](https://github.com/huggingface/diffusers/issues?q=is%3Aopen+is%3Aissue+label%3A%22good+first+issue%22) for general opportunities to contribute
@@ -128,7 +128,7 @@ just hang out ☕.
   </tr>
   <tr style="border-top: 2px solid black">
     <td>Unconditional Image Generation</td>
-    <td><a href="https://huggingface.co/docs/diffusers/api/pipelines/ddpm"> DDPM </a></td> 
+    <td><a href="https://huggingface.co/docs/diffusers/api/pipelines/ddpm"> DDPM </a></td>
     <td><a href="https://huggingface.co/google/ddpm-ema-church-256"> google/ddpm-ema-church-256 </a></td>
   </tr>
   <tr style="border-top: 2px solid black">
@@ -185,13 +185,13 @@ just hang out ☕.
 
 ## Popular libraries using 🧨 Diffusers
 
-- https://github.com/microsoft/TaskMatrix 
-- https://github.com/invoke-ai/InvokeAI 
-- https://github.com/apple/ml-stable-diffusion   
-- https://github.com/Sanster/lama-cleaner 
+- https://github.com/microsoft/TaskMatrix
+- https://github.com/invoke-ai/InvokeAI
+- https://github.com/apple/ml-stable-diffusion
+- https://github.com/Sanster/lama-cleaner
 - https://github.com/IDEA-Research/Grounded-Segment-Anything
-- https://github.com/ashawkey/stable-dreamfusion 
-- https://github.com/deep-floyd/IF  
+- https://github.com/ashawkey/stable-dreamfusion
+- https://github.com/deep-floyd/IF
 - https://github.com/bentoml/BentoML
 - https://github.com/bmaltais/kohya_ss
 - +3000 other amazing GitHub repositories 💪

docs/source/_config.py

@@ -6,4 +6,4 @@
 # ! pip install git+https://github.com/huggingface/diffusers.git
 """
 
-notebook_first_cells = [{"type": "code", "content": INSTALL_CONTENT}]
+notebook_first_cells = [{"type": "code", "content": INSTALL_CONTENT}]

docs/source/en/training/lora.mdx

@@ -260,6 +260,14 @@ pipe.load_lora_weights(lora_model_id)
 image = pipe("A picture of a sks dog in a bucket", num_inference_steps=25).images[0]
 ```
 
+<Tip>
+
+If your LoRA parameters involve the UNet as well as the Text Encoder, then passing
+`cross_attention_kwargs={"scale": 0.5}` will apply the `scale` value to both the UNet 
+and the Text Encoder. 
+
+</Tip>
+
 Note that the use of [`~diffusers.loaders.LoraLoaderMixin.load_lora_weights`] is preferred to [`~diffusers.loaders.UNet2DConditionLoadersMixin.load_attn_procs`] for loading LoRA parameters. This is because
 [`~diffusers.loaders.LoraLoaderMixin.load_lora_weights`] can handle the following situations:
 

src/diffusers/loaders.py

@@ -852,6 +852,9 @@ def load_lora_weights(self, pretrained_model_name_or_path_or_dict: Union[str, Di
         weight_name = kwargs.pop("weight_name", None)
         use_safetensors = kwargs.pop("use_safetensors", None)
 
+        # set lora scale to a reasonable default
+        self._lora_scale = 1.0
+
         if use_safetensors and not is_safetensors_available():
             raise ValueError(
                 "`use_safetensors`=True but safetensors is not installed. Please install safetensors with `pip install safetenstors"
@@ -953,6 +956,12 @@ def load_lora_weights(self, pretrained_model_name_or_path_or_dict: Union[str, Di
             warn_message = "You have saved the LoRA weights using the old format. To convert the old LoRA weights to the new format, you can first load them in a dictionary and then create a new dictionary like the following: `new_state_dict = {f'unet'.{module_name}: params for module_name, params in old_state_dict.items()}`."
             warnings.warn(warn_message)
 
+    @property
+    def lora_scale(self) -> float:
+        # property function that returns the lora scale which can be set at run time by the pipeline.
+        # if _lora_scale has not been set, return 1
+        return self._lora_scale if hasattr(self, "_lora_scale") else 1.0
+
     @property
     def text_encoder_lora_attn_procs(self):
         if hasattr(self, "_text_encoder_lora_attn_procs"):
@@ -1000,7 +1009,8 @@ def _modify_text_encoder(self, attn_processors: Dict[str, LoRAAttnProcessor]):
                     # for more detail, see https://github.com/huggingface/diffusers/pull/3490#issuecomment-1555059060
                     def make_new_forward(old_forward, lora_layer):
                         def new_forward(x):
-                            return old_forward(x) + lora_layer(x)
+                            result = old_forward(x) + self.lora_scale * lora_layer(x)
+                            return result
 
                         return new_forward
 

src/diffusers/pipelines/alt_diffusion/pipeline_alt_diffusion.py

@@ -24,7 +24,7 @@
 
 from ...configuration_utils import FrozenDict
 from ...image_processor import VaeImageProcessor
-from ...loaders import TextualInversionLoaderMixin
+from ...loaders import LoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, UNet2DConditionModel
 from ...schedulers import KarrasDiffusionSchedulers
 from ...utils import deprecate, logging, randn_tensor, replace_example_docstring
@@ -52,7 +52,7 @@
 
 
 # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline with Stable->Alt, CLIPTextModel->RobertaSeriesModelWithTransformation, CLIPTokenizer->XLMRobertaTokenizer, AltDiffusionSafetyChecker->StableDiffusionSafetyChecker
-class AltDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin):
+class AltDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin):
     r"""
     Pipeline for text-to-image generation using Alt Diffusion.
 
@@ -291,6 +291,7 @@ def _encode_prompt(
         negative_prompt=None,
         prompt_embeds: Optional[torch.FloatTensor] = None,
         negative_prompt_embeds: Optional[torch.FloatTensor] = None,
+        lora_scale: Optional[float] = None,
     ):
         r"""
         Encodes the prompt into text encoder hidden states.
@@ -315,7 +316,14 @@ def _encode_prompt(
                 Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
                 weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
                 argument.
+            lora_scale (`float`, *optional*):
+                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
         """
+        # set lora scale so that monkey patched LoRA
+        # function of text encoder can correctly access it
+        if lora_scale is not None and isinstance(self, LoraLoaderMixin):
+            self._lora_scale = lora_scale
+
         if prompt is not None and isinstance(prompt, str):
             batch_size = 1
         elif prompt is not None and isinstance(prompt, list):
@@ -653,6 +661,9 @@ def __call__(
         do_classifier_free_guidance = guidance_scale > 1.0
 
         # 3. Encode input prompt
+        text_encoder_lora_scale = (
+            cross_attention_kwargs.get("scale", None) if cross_attention_kwargs is not None else None
+        )
         prompt_embeds = self._encode_prompt(
             prompt,
             device,
@@ -661,6 +672,7 @@ def __call__(
             negative_prompt,
             prompt_embeds=prompt_embeds,
             negative_prompt_embeds=negative_prompt_embeds,
+            lora_scale=text_encoder_lora_scale,
         )
 
         # 4. Prepare timesteps