Upload 18 files

.gitattributes

@@ -25,7 +25,6 @@
 *.safetensors filter=lfs diff=lfs merge=lfs -text
 saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
 *.tflite filter=lfs diff=lfs merge=lfs -text
 *.tgz filter=lfs diff=lfs merge=lfs -text
 *.wasm filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,165 @@
+training_data/
+experiments/
+wandb/
+
+
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/#use-with-ide
+.pdm.toml
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/

.pre-commit-config.yaml

@@ -0,0 +1,37 @@
+exclude: train_realfill.py
+repos:
+- repo: https://github.com/pre-commit/pre-commit-hooks
+  rev: v4.2.0
+  hooks:
+  - id: check-executables-have-shebangs
+  - id: check-json
+  - id: check-merge-conflict
+  - id: check-shebang-scripts-are-executable
+  - id: check-toml
+  - id: check-yaml
+  - id: double-quote-string-fixer
+  - id: end-of-file-fixer
+  - id: mixed-line-ending
+    args: ['--fix=lf']
+  - id: requirements-txt-fixer
+  - id: trailing-whitespace
+- repo: https://github.com/myint/docformatter
+  rev: v1.4
+  hooks:
+  - id: docformatter
+    args: ['--in-place']
+- repo: https://github.com/pycqa/isort
+  rev: 5.10.1
+  hooks:
+    - id: isort
+- repo: https://github.com/pre-commit/mirrors-mypy
+  rev: v0.991
+  hooks:
+    - id: mypy
+      args: ['--ignore-missing-imports']
+      additional_dependencies: ['types-python-slugify']
+- repo: https://github.com/google/yapf
+  rev: v0.32.0
+  hooks:
+  - id: yapf
+    args: ['--parallel', '--in-place']

.style.yapf

@@ -0,0 +1,5 @@
+[style]
+based_on_style = pep8
+blank_line_before_nested_class_or_def = false
+spaces_before_comment = 2
+split_before_logical_operator = true

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2022 hysts
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -1,10 +1,11 @@
 ---
 title: RealFill Training UI
-emoji: 📈
-colorFrom: purple
-colorTo: blue
+emoji: ⚡
+colorFrom: red
+colorTo: purple
 sdk: gradio
-sdk_version: 3.50.2
+sdk_version: 3.16.2
+python_version: 3.10.9
 app_file: app.py
 pinned: false
 license: mit

app.py

@@ -0,0 +1,75 @@
+#!/usr/bin/env python
+
+from __future__ import annotations
+
+import os
+
+import gradio as gr
+import torch
+
+from app_inference import create_inference_demo
+from app_training import create_training_demo
+from app_upload import create_upload_demo
+from inference import InferencePipeline
+from trainer import Trainer
+
+TITLE = '# RealFill Training UI'
+
+ORIGINAL_SPACE_ID = 'realfill-library/RealFill-Training-UI'
+SPACE_ID = os.getenv('SPACE_ID', ORIGINAL_SPACE_ID)
+SHARED_UI_WARNING = f'''# Attention - This Space doesn't work in this shared UI. You can duplicate and use it with a paid private T4 GPU.
+<center><a class="duplicate-button" style="display:inline-block" target="_blank" href="https://huggingface.co/spaces/{SPACE_ID}?duplicate=true"><img src="https://img.shields.io/badge/-Duplicate%20Space-blue?labelColor=white&style=flat&logo=data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAABAAAAAQCAYAAAAf8/9hAAAAAXNSR0IArs4c6QAAAP5JREFUOE+lk7FqAkEURY+ltunEgFXS2sZGIbXfEPdLlnxJyDdYB62sbbUKpLbVNhyYFzbrrA74YJlh9r079973psed0cvUD4A+4HoCjsA85X0Dfn/RBLBgBDxnQPfAEJgBY+A9gALA4tcbamSzS4xq4FOQAJgCDwV2CPKV8tZAJcAjMMkUe1vX+U+SMhfAJEHasQIWmXNN3abzDwHUrgcRGmYcgKe0bxrblHEB4E/pndMazNpSZGcsZdBlYJcEL9Afo75molJyM2FxmPgmgPqlWNLGfwZGG6UiyEvLzHYDmoPkDDiNm9JR9uboiONcBXrpY1qmgs21x1QwyZcpvxt9NS09PlsPAAAAAElFTkSuQmCC&logoWidth=14" alt="Duplicate Space"></a></center>
+'''
+
+if os.getenv('SYSTEM') == 'spaces' and SPACE_ID != ORIGINAL_SPACE_ID:
+    SETTINGS = f'<a href="https://huggingface.co/spaces/{SPACE_ID}/settings">Settings</a>'
+else:
+    SETTINGS = 'Settings'
+CUDA_NOT_AVAILABLE_WARNING = f'''# Attention - Running on CPU.
+<center>
+You can assign a GPU in the {SETTINGS} tab if you are running this on HF Spaces.
+"T4 small" is sufficient to run this demo.
+</center>
+'''
+
+HF_TOKEN_NOT_SPECIFIED_WARNING = f'''# Attention - The environment variable `HF_TOKEN` is not specified. Please specify your Hugging Face token with write permission as the value of it.
+<center>
+You can check and create your Hugging Face tokens <a href="https://huggingface.co/settings/tokens" target="_blank">here</a>.
+You can specify environment variables in the "Repository secrets" section of the {SETTINGS} tab.
+</center>
+'''
+
+HF_TOKEN = os.getenv('HF_TOKEN')
+
+
+def show_warning(warning_text: str) -> gr.Blocks:
+    with gr.Blocks() as demo:
+        with gr.Box():
+            gr.Markdown(warning_text)
+    return demo
+
+
+pipe = InferencePipeline(HF_TOKEN)
+trainer = Trainer(HF_TOKEN)
+
+with gr.Blocks(css='style.css') as demo:
+    if os.getenv('IS_SHARED_UI'):
+        show_warning(SHARED_UI_WARNING)
+    if not torch.cuda.is_available():
+        show_warning(CUDA_NOT_AVAILABLE_WARNING)
+    if not HF_TOKEN:
+        show_warning(HF_TOKEN_NOT_SPECIFIED_WARNING)
+
+    gr.Markdown(TITLE)
+    with gr.Tabs():
+        with gr.TabItem('Train'):
+            create_training_demo(trainer, pipe)
+        with gr.TabItem('Test'):
+            create_inference_demo(pipe, HF_TOKEN)
+        with gr.TabItem('Upload'):
+            gr.Markdown('''
+            - You can use this tab to upload models later if you choose not to upload models in training time or if upload in training time failed.
+            ''')
+            create_upload_demo(HF_TOKEN)
+
+demo.queue(max_size=1).launch(share=True)

app_inference.py

@@ -0,0 +1,162 @@
+#!/usr/bin/env python
+
+from __future__ import annotations
+
+import enum
+
+import gradio as gr
+from huggingface_hub import HfApi
+
+from inference import InferencePipeline
+from utils import find_exp_dirs
+
+SAMPLE_MODEL_IDS = ['thuanz123/flowerwoman']
+
+
+class ModelSource(enum.Enum):
+    SAMPLE = 'Sample'
+    HUB_LIB = 'Hub (realfill-library)'
+    LOCAL = 'Local'
+
+
+class InferenceUtil:
+    def __init__(self, hf_token: str | None):
+        self.hf_token = hf_token
+
+    @staticmethod
+    def load_sample_model_list():
+        return gr.update(choices=SAMPLE_MODEL_IDS, value=SAMPLE_MODEL_IDS[0])
+    
+    def load_hub_model_list(self) -> dict:
+        api = HfApi(token=self.hf_token)
+        choices = [
+            info.modelId for info in api.list_models(author='realfill-library')
+        ]
+        return gr.update(choices=choices,
+                         value=choices[0] if choices else None)
+
+    @staticmethod
+    def load_local_model_list() -> dict:
+        choices = find_exp_dirs()
+        return gr.update(choices=choices,
+                         value=choices[0] if choices else None)
+
+    def reload_model_list(self, model_source: str) -> dict:
+        if model_source == ModelSource.SAMPLE.value:
+            return self.load_sample_model_list()
+        elif model_source == ModelSource.HUB_LIB.value:
+            return self.load_hub_model_list()
+        elif model_source == ModelSource.LOCAL.value:
+            return self.load_local_model_list()
+        else:
+            raise ValueError
+
+    def load_model_info(self, model_id: str) -> tuple[str, str]:
+        try:
+            card = InferencePipeline.get_model_card(model_id, self.hf_token)
+        except Exception:
+            return '', ''
+        target_image = getattr(card.data, 'target_image', '')
+        target_mask = getattr(card.data, 'target_mask', '')
+        return target_image, target_mask 
+
+    def reload_model_list_and_update_model_info(
+            self, model_source: str
+        ) -> tuple[dict, str, str]:
+        model_list_update = self.reload_model_list(model_source)
+        model_list = model_list_update['choices']
+        model_info = self.load_model_info(model_list[0] if model_list else '')
+        return model_list_update, *model_info
+
+
+def create_inference_demo(pipe: InferencePipeline,
+                          hf_token: str | None = None) -> gr.Blocks:
+    app = InferenceUtil(hf_token)
+
+    with gr.Blocks() as demo:
+        with gr.Row():
+            with gr.Column():
+                with gr.Box():
+                    model_source = gr.Radio(
+                        label='Model Source',
+                        choices=[_.value for _ in ModelSource],
+                        value=ModelSource.SAMPLE.value)
+                    reload_button = gr.Button('Reload Model List')
+                    model_id = gr.Dropdown(label='Model ID',
+                                                choices=SAMPLE_MODEL_IDS,
+                                                value=SAMPLE_MODEL_IDS[0])
+                    with gr.Accordion(
+                            label=
+                            'Model info (Target image and mask used for both training and inference)',
+                            open=False):
+                        with gr.Row():
+                            target_image = gr.Image(
+                                label='Target Image', interactive=False)
+                            target_mask = gr.Image(
+                                label='Target Mask', interactive=False)
+                seed = gr.Slider(label='Seed',
+                                 minimum=0,
+                                 maximum=100000,
+                                 step=1,
+                                 value=0)
+                with gr.Accordion('Other Parameters', open=False):
+                    num_steps = gr.Slider(label='Number of Steps',
+                                          minimum=0,
+                                          maximum=100,
+                                          step=1,
+                                          value=25)
+                    guidance_scale = gr.Slider(label='CFG Scale',
+                                               minimum=0,
+                                               maximum=50,
+                                               step=0.1,
+                                               value=5.0)
+
+                run_button = gr.Button('Generate')
+
+                gr.Markdown('''
+                - After training, you can press "Reload Model List" button to load your trained model names.
+                ''')
+            with gr.Column():
+                result = gr.Image(label='Result')
+
+        model_source.change(
+            fn=app.reload_model_list_and_update_model_info,
+            inputs=model_source,
+            outputs=[
+                model_id, 
+                target_image, 
+                target_mask
+            ])
+        reload_button.click(
+            fn=app.reload_model_list_and_update_model_info,
+            inputs=model_source,
+            outputs=[
+                model_id,
+                target_image, 
+                target_mask
+            ])
+        model_id.change(fn=app.load_model_info,
+                             inputs=model_id,
+                             outputs=[
+                                 target_image, 
+                                 target_mask
+                             ])
+        inputs = [
+            model_id,
+            seed,
+            target_image,
+            target_mask,
+            num_steps,
+            guidance_scale,
+        ]
+        run_button.click(fn=pipe.run, inputs=inputs, outputs=result)
+    return demo
+
+
+if __name__ == '__main__':
+    import os
+
+    hf_token = os.getenv('HF_TOKEN')
+    pipe = InferencePipeline(hf_token)
+    demo = create_inference_demo(pipe, hf_token)
+    demo.queue(max_size=10).launch(share=False)

app_training.py

@@ -0,0 +1,152 @@
+#!/usr/bin/env python
+
+from __future__ import annotations
+
+import os
+
+import gradio as gr
+
+from constants import UploadTarget
+from inference import InferencePipeline
+from trainer import Trainer
+
+
+def create_training_demo(trainer: Trainer,
+                         pipe: InferencePipeline | None = None) -> gr.Blocks:
+    with gr.Blocks() as demo:
+        with gr.Row():
+            with gr.Column():
+                with gr.Box():
+                    gr.Markdown('Training Data')
+                    reference_images = gr.Files(label='Reference images')
+                    target_image = gr.Files(label='Target image')
+                    target_mask = gr.Files(label='Target mask')
+                    gr.Markdown('''
+                        - Upload reference images of the scene you are planning on training on.
+                        - For the target image, the inpainting region should be white.
+                        - For the target mask, white for inpainting and black for keeping as is.
+                        ''')
+                with gr.Box():
+                    gr.Markdown('Output Model')
+                    output_model_name = gr.Text(label='Name of your model',
+                                                max_lines=1)
+                    delete_existing_model = gr.Checkbox(
+                        label='Delete existing model of the same name',
+                        value=False)
+                with gr.Box():
+                    gr.Markdown('Upload Settings')
+                    with gr.Row():
+                        upload_to_hub = gr.Checkbox(
+                            label='Upload model to Hub', value=True)
+                        use_private_repo = gr.Checkbox(label='Private',
+                                                       value=True)
+                        delete_existing_repo = gr.Checkbox(
+                            label='Delete existing repo of the same name',
+                            value=False)
+                    upload_to = gr.Radio(
+                        label='Upload to',
+                        choices=[_.value for _ in UploadTarget],
+                        value=UploadTarget.REALFILL_LIBRARY.value)
+                    gr.Markdown('''
+                    - By default, trained models will be uploaded to [ReaFill Library](https://huggingface.co/realfill-library).
+                    - You can also choose "Personal Profile", in which case, the model will be uploaded to https://huggingface.co/{your_username}/{model_name}.
+                    ''')
+
+            with gr.Box():
+                gr.Markdown('Training Parameters')
+                with gr.Row():
+                    base_model = gr.Text(
+                        label='Base Model',
+                        value='stabilityai/stable-diffusion-2-inpainting',
+                        max_lines=1)
+                    resolution = gr.Dropdown(choices=['512', '768'],
+                                             value='512',
+                                             label='Resolution')
+                num_training_steps = gr.Number(
+                    label='Number of Training Steps', value=2000, precision=0)
+                unet_learning_rate = gr.Number(label='Unet Learning Rate', value=0.0002)
+                text_encoder_learning_rate = gr.Number(label='Text Encoder Learning Rate', value=0.00004)
+                lora_rank = gr.Number(label='LoRA rank value', value=8, precision=0)
+                lora_dropout = gr.Number(label='LoRA dropout rate', value=0.1)
+                lora_alpha = gr.Number(label='LoRA alpha value', value=16, precision=0)
+                gradient_accumulation = gr.Number(
+                    label='Number of Gradient Accumulation',
+                    value=1,
+                    precision=0)
+                seed = gr.Slider(label='Seed',
+                                 minimum=0,
+                                 maximum=100000,
+                                 step=1,
+                                 value=0)
+                fp16 = gr.Checkbox(label='FP16', value=True)
+                use_8bit_adam = gr.Checkbox(label='Use 8bit Adam', value=True)
+                checkpointing_steps = gr.Number(label='Checkpointing Steps',
+                                                value=100,
+                                                precision=0)
+                use_wandb = gr.Checkbox(label='Use W&B',
+                                        value=False,
+                                        interactive=bool(
+                                            os.getenv('WANDB_API_KEY')))
+                validation_steps = gr.Number(label='Validation Steps',
+                                              value=100,
+                                              precision=0)
+                gr.Markdown('''
+                    - The base model must be a model that is compatible with [diffusers](https://github.com/huggingface/diffusers) library.
+                    - It takes a few minutes to download the base model first.
+                    - It will take about 16 minutes to train for 2000 steps with a T4 GPU.
+                    - You may want to try a small number of steps first, like 1, to see if everything works fine in your environment.
+                    - You can check the training status by pressing the "Open logs" button if you are running this on your Space.
+                    - You need to set the environment variable `WANDB_API_KEY` if you'd like to use [W&B](https://wandb.ai/site). See [W&B documentation](https://docs.wandb.ai/guides/track/advanced/environment-variables).
+                    - **Note:** Due to [this issue](https://github.com/huggingface/accelerate/issues/944), currently, training will not terminate properly if you use W&B.
+                    ''')
+
+        remove_gpu_after_training = gr.Checkbox(
+            label='Remove GPU after training',
+            value=False,
+            interactive=bool(os.getenv('SPACE_ID')),
+            visible=False)
+        run_button = gr.Button('Start Training')
+
+        with gr.Box():
+            gr.Markdown('Output message')
+            output_message = gr.Markdown()
+
+        if pipe is not None:
+            run_button.click(fn=pipe.clear)
+        run_button.click(fn=trainer.run,
+                         inputs=[
+                             reference_images,
+                             target_image,
+                             target_mask,
+                             output_model_name,
+                             delete_existing_model,
+                             base_model,
+                             resolution,
+                             num_training_steps,
+                             unet_learning_rate,
+                             text_encoder_learning_rate,
+                             lora_rank,
+                             lora_dropout,
+                             lora_alpha,
+                             gradient_accumulation,
+                             seed,
+                             fp16,
+                             use_8bit_adam,
+                             checkpointing_steps,
+                             use_wandb,
+                             validation_steps,
+                             upload_to_hub,
+                             use_private_repo,
+                             delete_existing_repo,
+                             upload_to,
+                             remove_gpu_after_training,
+                         ],
+                         outputs=output_message)
+    return demo
+
+
+if __name__ == '__main__':
+    hf_token = os.getenv('HF_TOKEN')
+    trainer = Trainer(hf_token)
+    demo = create_training_demo(trainer)
+    demo.queue(max_size=1).launch(share=False)

app_upload.py

@@ -0,0 +1,100 @@
+#!/usr/bin/env python
+
+from __future__ import annotations
+
+import pathlib
+
+import gradio as gr
+import slugify
+
+from constants import UploadTarget
+from uploader import Uploader
+from utils import find_exp_dirs
+
+
+class ModelUploader(Uploader):
+    def upload_model(
+        self,
+        folder_path: str,
+        repo_name: str,
+        upload_to: str,
+        private: bool,
+        delete_existing_repo: bool,
+    ) -> str:
+        if not folder_path:
+            raise ValueError
+        if not repo_name:
+            repo_name = pathlib.Path(folder_path).name
+        repo_name = slugify.slugify(repo_name)
+
+        if upload_to == UploadTarget.PERSONAL_PROFILE.value:
+            organization = ''
+        elif upload_to == UploadTarget.REALFILL_LIBRARY.value:
+            organization = 'realfill-library'
+        else:
+            raise ValueError
+
+        return self.upload(folder_path,
+                           repo_name,
+                           organization=organization,
+                           private=private,
+                           delete_existing_repo=delete_existing_repo)
+
+
+def load_local_model_list() -> dict:
+    choices = find_exp_dirs(ignore_repo=True)
+    return gr.update(choices=choices, value=choices[0] if choices else None)
+
+
+def create_upload_demo(hf_token: str | None) -> gr.Blocks:
+    uploader = ModelUploader(hf_token)
+    model_dirs = find_exp_dirs(ignore_repo=True)
+
+    with gr.Blocks() as demo:
+        with gr.Box():
+            gr.Markdown('Local Models')
+            reload_button = gr.Button('Reload Model List')
+            model_dir = gr.Dropdown(
+                label='Model names',
+                choices=model_dirs,
+                value=model_dirs[0] if model_dirs else None)
+        with gr.Box():
+            gr.Markdown('Upload Settings')
+            with gr.Row():
+                use_private_repo = gr.Checkbox(label='Private', value=True)
+                delete_existing_repo = gr.Checkbox(
+                    label='Delete existing repo of the same name', value=False)
+            upload_to = gr.Radio(label='Upload to',
+                                 choices=[_.value for _ in UploadTarget],
+                                 value=UploadTarget.REALFILL_LIBRARY.value)
+            model_name = gr.Textbox(label='Model Name')
+        upload_button = gr.Button('Upload')
+        gr.Markdown('''
+            - You can upload your trained model to your personal profile (i.e. https://huggingface.co/{your_username}/{model_name}) or to the public [ReaFill Library](https://huggingface.co/realfill-library) (i.e. https://huggingface.co/realfill-library/{model_name}).
+            ''')
+        with gr.Box():
+            gr.Markdown('Output message')
+            output_message = gr.Markdown()
+
+        reload_button.click(fn=load_local_model_list,
+                            inputs=None,
+                            outputs=model_dir)
+        upload_button.click(fn=uploader.upload_model,
+                            inputs=[
+                                model_dir,
+                                model_name,
+                                upload_to,
+                                use_private_repo,
+                                delete_existing_repo,
+                            ],
+                            outputs=output_message)
+
+    return demo
+
+
+if __name__ == '__main__':
+    import os
+
+    hf_token = os.getenv('HF_TOKEN')
+    demo = create_upload_demo(hf_token)
+    demo.queue(max_size=1).launch(share=False)

constants.py

@@ -0,0 +1,6 @@
+import enum
+
+
+class UploadTarget(enum.Enum):
+    PERSONAL_PROFILE = 'Personal Profile'
+    REALFILL_LIBRARY = 'RealFill Library'

inference.py

@@ -0,0 +1,83 @@
+from __future__ import annotations
+
+import gc
+import pathlib
+
+import gradio as gr
+import PIL.Image
+import torch
+from diffusers import DiffusionPipeline, DPMSolverMultistepScheduler
+from huggingface_hub import ModelCard
+
+
+class InferencePipeline:
+    def __init__(self, hf_token: str | None = None):
+        self.hf_token = hf_token
+        self.pipe = None
+        self.device = torch.device(
+            'cuda:0' if torch.cuda.is_available() else 'cpu')
+        self.model_id = None
+
+    def clear(self) -> None:
+        self.model_id = None
+        del self.pipe
+        self.pipe = None
+        torch.cuda.empty_cache()
+        gc.collect()
+
+    @staticmethod
+    def check_if_model_is_local(model_id: str) -> bool:
+        return pathlib.Path(model_id).exists()
+
+    @staticmethod
+    def get_model_card(model_id: str,
+                       hf_token: str | None = None) -> ModelCard:
+        if InferencePipeline.check_if_model_is_local(model_id):
+            card_path = (pathlib.Path(model_id) / 'README.md').as_posix()
+        else:
+            card_path = model_id
+        return ModelCard.load(card_path, token=hf_token)
+
+    def load_pipe(self, model_id: str) -> None:
+        if model_id == self.model_id:
+            return
+        
+        if self.device.type == 'cpu':
+            pipe = DiffusionPipeline.from_pretrained(
+                model_id, use_auth_token=self.hf_token)
+        else:
+            pipe = DiffusionPipeline.from_pretrained(
+                model_id, torch_dtype=torch.float16,
+                use_auth_token=self.hf_token)
+            pipe = pipe.to(self.device)
+        pipe.scheduler = DPMSolverMultistepScheduler.from_config(
+            pipe.scheduler.config)
+        self.pipe = pipe
+        
+        pipe.safety_checker = lambda images, **kwargs: (images, [False] * len(images))
+        self.model_id = model_id  # type: ignore
+
+    def run(
+        self,
+        model_id: str,
+        target_image: str,
+        target_mask: str,
+        seed: int,
+        n_steps: int,
+        guidance_scale: float,
+    ) -> PIL.Image.Image:
+        if not torch.cuda.is_available():
+            raise gr.Error('CUDA is not available.')
+
+        self.load_pipe(model_id)
+
+        generator = torch.Generator(device=self.device).manual_seed(seed)
+        out = self.pipe(
+            "a photo of sks",
+            image=target_image,
+            mask_image=target_mask,
+            num_inference_steps=n_steps,
+            guidance_scale=guidance_scale,
+            generator=generator,
+        )  # type: ignore
+        return out.images[0]

requirements.txt

@@ -0,0 +1,9 @@
+diffusers==0.20.1
+accelerate==0.23.0
+transformers==4.34.0
+peft==0.5.0
+torch==2.0.1
+torchvision==0.15.2
+ftfy==6.1.1
+tensorboard==2.14.0
+Jinja2==3.1.2

style.css

@@ -0,0 +1,3 @@
+h1 {
+  text-align: center;
+}

train_realfill.py

@@ -0,0 +1,952 @@
+import random
+import argparse
+import copy
+import itertools
+import logging
+import math
+import os
+import shutil
+from pathlib import Path
+
+import numpy as np
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+import transformers
+from accelerate import Accelerator
+from accelerate.logging import get_logger
+from accelerate.utils import set_seed
+from huggingface_hub import create_repo, upload_folder
+from packaging import version
+from PIL import Image
+from PIL.ImageOps import exif_transpose
+from torch.utils.data import Dataset
+import torchvision.transforms.v2 as transforms_v2
+from tqdm.auto import tqdm
+from transformers import AutoTokenizer, CLIPTextModel
+
+import diffusers
+from diffusers import (
+    AutoencoderKL,
+    DDPMScheduler,
+    StableDiffusionInpaintPipeline,
+    DPMSolverMultistepScheduler,
+    UNet2DConditionModel,
+)
+from diffusers.optimization import get_scheduler
+from diffusers.utils import check_min_version, is_wandb_available
+from diffusers.utils.import_utils import is_xformers_available
+
+from peft import PeftModel, LoraConfig, get_peft_model
+
+# Will error if the minimal version of diffusers is not installed. Remove at your own risks.
+check_min_version("0.20.1")
+
+logger = get_logger(__name__)
+
+def make_mask(images, resolution, times=30):
+    mask, times = torch.ones_like(images[0:1, :, :]), np.random.randint(1, times)
+    min_size, max_size, margin = np.array([0.03, 0.25, 0.01]) * resolution
+    max_size = min(max_size, resolution - margin * 2)
+
+    for _ in range(times):
+        width = np.random.randint(int(min_size), int(max_size))
+        height = np.random.randint(int(min_size), int(max_size))
+
+        x_start = np.random.randint(int(margin), resolution - int(margin) - width + 1)
+        y_start = np.random.randint(int(margin), resolution - int(margin) - height + 1)
+        mask[:, y_start:y_start + height, x_start:x_start + width] = 0
+
+    mask = 1 - mask if random.random() < 0.5 else mask
+    return mask
+
+def save_model_card(
+    repo_id: str,
+    base_model: str,
+    target_image: str,
+    target_mask: str,
+    repo_folder=None,
+):
+    yaml = f"""
+---
+license: creativeml-openrail-m
+base_model: {base_model}
+target_image: {target_image}
+target_mask: {target_mask}
+tags:
+- stable-diffusion-inpainting
+- stable-diffusion-inpainting-diffusers
+- text-to-image
+- diffusers
+- realfill
+inference: true
+---
+    """
+    model_card = f"""
+# RealFill - {repo_id}
+
+This is a realfill model derived from {base_model}. The weights were trained using [RealFill](https://realfill.github.io/).
+"""
+    with open(os.path.join(repo_folder, "README.md"), "w") as f:
+        f.write(yaml + model_card)
+
+def log_validation(
+    text_encoder,
+    tokenizer,
+    unet,
+    args,
+    accelerator,
+    weight_dtype,
+    epoch,
+):
+    logger.info(
+        f"Running validation... \nGenerating {args.num_validation_images} images"
+    )
+
+    # create pipeline (note: unet and vae are loaded again in float32)
+    pipeline = StableDiffusionInpaintPipeline.from_pretrained(
+        args.pretrained_model_name_or_path,
+        tokenizer=tokenizer,
+        revision=args.revision,
+        torch_dtype=weight_dtype,
+    )
+
+    # set `keep_fp32_wrapper` to True because we do not want to remove
+    # mixed precision hooks while we are still training
+    pipeline.unet = accelerator.unwrap_model(unet, keep_fp32_wrapper=True)
+    pipeline.text_encoder = accelerator.unwrap_model(text_encoder, keep_fp32_wrapper=True)
+    pipeline.scheduler = DPMSolverMultistepScheduler.from_config(pipeline.scheduler.config)
+
+    pipeline = pipeline.to(accelerator.device)
+    pipeline.set_progress_bar_config(disable=True)
+
+    # run inference
+    generator = None if args.seed is None else torch.Generator(device=accelerator.device).manual_seed(args.seed)
+
+    target_dir = Path(args.train_data_dir) / "target"
+    target_image, target_mask = target_dir / "target.png", target_dir / "mask.png"
+    image, mask_image = Image.open(target_image), Image.open(target_mask)
+
+    if image.mode != "RGB":
+        image = image.convert("RGB")
+
+    images = []
+    for _ in range(args.num_validation_images):
+        image = pipeline(
+            prompt="a photo of sks", image=image, mask_image=mask_image,
+            num_inference_steps=25, guidance_scale=5, generator=generator
+        ).images[0]
+        images.append(image)
+
+    for tracker in accelerator.trackers:
+        if tracker.name == "tensorboard":
+            np_images = np.stack([np.asarray(img) for img in images])
+            tracker.writer.add_images(f"validation", np_images, epoch, dataformats="NHWC")
+        if tracker.name == "wandb":
+            tracker.log(
+                {
+                    f"validation": [
+                        wandb.Image(image, caption=str(i)) for i, image in enumerate(images)
+                    ]
+                }
+            )
+
+    del pipeline
+    torch.cuda.empty_cache()
+
+    return images
+
+def parse_args(input_args=None):
+    parser = argparse.ArgumentParser(description="Simple example of a training script.")
+    parser.add_argument(
+        "--pretrained_model_name_or_path",
+        type=str,
+        default=None,
+        required=True,
+        help="Path to pretrained model or model identifier from huggingface.co/models.",
+    )
+    parser.add_argument(
+        "--revision",
+        type=str,
+        default=None,
+        required=False,
+        help="Revision of pretrained model identifier from huggingface.co/models.",
+    )
+    parser.add_argument(
+        "--tokenizer_name",
+        type=str,
+        default=None,
+        help="Pretrained tokenizer name or path if not the same as model_name",
+    )
+    parser.add_argument(
+        "--train_data_dir",
+        type=str,
+        default=None,
+        required=True,
+        help="A folder containing the training data of images.",
+    )
+    parser.add_argument(
+        "--num_validation_images",
+        type=int,
+        default=4,
+        help="Number of images that should be generated during validation with `validation_conditioning`.",
+    )
+    parser.add_argument(
+        "--validation_steps",
+        type=int,
+        default=100,
+        help=(
+            "Run realfill validation every X steps. RealFill validation consists of running the conditioning"
+            " `args.validation_conditioning` multiple times: `args.num_validation_images`."
+        ),
+    )
+    parser.add_argument(
+        "--output_dir",
+        type=str,
+        default="realfill-model",
+        help="The output directory where the model predictions and checkpoints will be written.",
+    )
+    parser.add_argument("--seed", type=int, default=None, help="A seed for reproducible training.")
+    parser.add_argument(
+        "--resolution",
+        type=int,
+        default=512,
+        help=(
+            "The resolution for input images, all the images in the train/validation dataset will be resized to this"
+            " resolution"
+        ),
+    )
+    parser.add_argument(
+        "--train_batch_size", type=int, default=4, help="Batch size (per device) for the training dataloader."
+    )
+    parser.add_argument("--num_train_epochs", type=int, default=1)
+    parser.add_argument(
+        "--max_train_steps",
+        type=int,
+        default=None,
+        help="Total number of training steps to perform.  If provided, overrides num_train_epochs.",
+    )
+    parser.add_argument(
+        "--checkpointing_steps",
+        type=int,
+        default=500,
+        help=(
+            "Save a checkpoint of the training state every X updates. These checkpoints can be used both as final"
+            " checkpoints in case they are better than the last checkpoint, and are also suitable for resuming"
+            " training using `--resume_from_checkpoint`."
+        ),
+    )
+    parser.add_argument(
+        "--checkpoints_total_limit",
+        type=int,
+        default=None,
+        help=("Max number of checkpoints to store."),
+    )
+    parser.add_argument(
+        "--resume_from_checkpoint",
+        type=str,
+        default=None,
+        help=(
+            "Whether training should be resumed from a previous checkpoint. Use a path saved by"
+            ' `--checkpointing_steps`, or `"latest"` to automatically select the last available checkpoint.'
+        ),
+    )
+    parser.add_argument(
+        "--gradient_accumulation_steps",
+        type=int,
+        default=1,
+        help="Number of updates steps to accumulate before performing a backward/update pass.",
+    )
+    parser.add_argument(
+        "--gradient_checkpointing",
+        action="store_true",
+        help="Whether or not to use gradient checkpointing to save memory at the expense of slower backward pass.",
+    )
+    parser.add_argument(
+        "--unet_learning_rate",
+        type=float,
+        default=2e-4,
+        help="Learning rate to use for unet.",
+    )
+    parser.add_argument(
+        "--text_encoder_learning_rate",
+        type=float,
+        default=4e-5,
+        help="Learning rate to use for text encoder.",
+    )
+    parser.add_argument(
+        "--scale_lr",
+        action="store_true",
+        default=False,
+        help="Scale the learning rate by the number of GPUs, gradient accumulation steps, and batch size.",
+    )
+    parser.add_argument(
+        "--lr_scheduler",
+        type=str,
+        default="constant",
+        help=(
+            'The scheduler type to use. Choose between ["linear", "cosine", "cosine_with_restarts", "polynomial",'
+            ' "constant", "constant_with_warmup"]'
+        ),
+    )
+    parser.add_argument(
+        "--lr_warmup_steps", type=int, default=500, help="Number of steps for the warmup in the lr scheduler."
+    )
+    parser.add_argument(
+        "--lr_num_cycles",
+        type=int,
+        default=1,
+        help="Number of hard resets of the lr in cosine_with_restarts scheduler.",
+    )
+    parser.add_argument("--lr_power", type=float, default=1.0, help="Power factor of the polynomial scheduler.")
+    parser.add_argument(
+        "--use_8bit_adam", action="store_true", help="Whether or not to use 8-bit Adam from bitsandbytes."
+    )
+    parser.add_argument("--adam_beta1", type=float, default=0.9, help="The beta1 parameter for the Adam optimizer.")
+    parser.add_argument("--adam_beta2", type=float, default=0.999, help="The beta2 parameter for the Adam optimizer.")
+    parser.add_argument("--adam_weight_decay", type=float, default=1e-2, help="Weight decay to use.")
+    parser.add_argument("--adam_epsilon", type=float, default=1e-08, help="Epsilon value for the Adam optimizer")
+    parser.add_argument("--max_grad_norm", default=1.0, type=float, help="Max gradient norm.")
+    parser.add_argument("--push_to_hub", action="store_true", help="Whether or not to push the model to the Hub.")
+    parser.add_argument("--hub_token", type=str, default=None, help="The token to use to push to the Model Hub.")
+    parser.add_argument(
+        "--hub_model_id",
+        type=str,
+        default=None,
+        help="The name of the repository to keep in sync with the local `output_dir`.",
+    )
+    parser.add_argument(
+        "--logging_dir",
+        type=str,
+        default="logs",
+        help=(
+            "[TensorBoard](https://www.tensorflow.org/tensorboard) log directory. Will default to"
+            " *output_dir/runs/**CURRENT_DATETIME_HOSTNAME***."
+        ),
+    )
+    parser.add_argument(
+        "--allow_tf32",
+        action="store_true",
+        help=(
+            "Whether or not to allow TF32 on Ampere GPUs. Can be used to speed up training. For more information, see"
+            " https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices"
+        ),
+    )
+    parser.add_argument(
+        "--report_to",
+        type=str,
+        default="tensorboard",
+        help=(
+            'The integration to report the results and logs to. Supported platforms are `"tensorboard"`'
+            ' (default), `"wandb"` and `"comet_ml"`. Use `"all"` to report to all integrations.'
+        ),
+    )
+    parser.add_argument(
+        "--wandb_key",
+        type=str,
+        default=None,
+        help=("If report to option is set to wandb, api-key for wandb used for login to wandb "),
+    )
+    parser.add_argument(
+        "--wandb_project_name",
+        type=str,
+        default=None,
+        help=("If report to option is set to wandb, project name in wandb for log tracking  "),
+    )
+    parser.add_argument(
+        "--mixed_precision",
+        type=str,
+        default=None,
+        choices=["no", "fp16", "bf16"],
+        help=(
+            "Whether to use mixed precision. Choose between fp16 and bf16 (bfloat16). Bf16 requires PyTorch >="
+            " 1.10.and an Nvidia Ampere GPU.  Default to the value of accelerate config of the current system or the"
+            " flag passed with the `accelerate.launch` command. Use this argument to override the accelerate config."
+        ),
+    )
+    parser.add_argument("--local_rank", type=int, default=-1, help="For distributed training: local_rank")
+    parser.add_argument(
+        "--enable_xformers_memory_efficient_attention", action="store_true", help="Whether or not to use xformers."
+    )
+    parser.add_argument(
+        "--set_grads_to_none",
+        action="store_true",
+        help=(
+            "Save more memory by using setting grads to None instead of zero. Be aware, that this changes certain"
+            " behaviors, so disable this argument if it causes any problems. More info:"
+            " https://pytorch.org/docs/stable/generated/torch.optim.Optimizer.zero_grad.html"
+        ),
+    )
+    parser.add_argument(
+        "--lora_rank",
+        type=int,
+        default=16,
+        help=("The dimension of the LoRA update matrices."),
+    )
+    parser.add_argument(
+        "--lora_alpha",
+        type=int,
+        default=27,
+        help=("The alpha constant of the LoRA update matrices."),
+    )
+    parser.add_argument(
+        "--lora_dropout",
+        type=float,
+        default=0.0,
+        help="The dropout rate of the LoRA update matrices.",
+    )
+    parser.add_argument(
+        "--lora_bias",
+        type=str,
+        default="none",
+        help="The bias type of the Lora update matrices. Must be 'none', 'all' or 'lora_only'.",
+    )
+
+    if input_args is not None:
+        args = parser.parse_args(input_args)
+    else:
+        args = parser.parse_args()
+
+    env_local_rank = int(os.environ.get("LOCAL_RANK", -1))
+    if env_local_rank != -1 and env_local_rank != args.local_rank:
+        args.local_rank = env_local_rank
+
+    return args
+
+class RealFillDataset(Dataset):
+    """
+    A dataset to prepare the training and conditioning images and
+    the masks with the dummy prompt for fine-tuning the model.
+    It pre-processes the images, masks and tokenizes the prompts.
+    """
+
+    def __init__(
+        self,
+        train_data_root,
+        tokenizer,
+        size=512,
+    ):
+        self.size = size
+        self.tokenizer = tokenizer
+
+        self.ref_data_root = Path(train_data_root) / "ref"
+        self.target_image = Path(train_data_root) / "target" / "target.jpg"
+        self.target_mask = Path(train_data_root) / "target" / "mask.jpg"
+        if not (self.ref_data_root.exists() and self.target_image.exists() and self.target_mask.exists()):
+            raise ValueError("Train images root doesn't exists.")
+
+        self.train_images_path = list(self.ref_data_root.iterdir()) + [self.target_image]
+        self.num_train_images = len(self.train_images_path)
+        self.train_prompt = "a photo of sks"
+
+        self.transform = transforms_v2.Compose(
+            [
+                transforms_v2.RandomResize(size, int(1.125 * size)),
+                transforms_v2.RandomCrop(size),
+                transforms_v2.ToImageTensor(),
+                transforms_v2.ConvertImageDtype(),
+                transforms_v2.Normalize([0.5], [0.5]),
+            ]
+        )
+
+    def __len__(self):
+        return self.num_train_images
+
+    def __getitem__(self, index):
+        example = {}
+
+        image = Image.open(self.train_images_path[index])
+        image = exif_transpose(image)
+
+        if not image.mode == "RGB":
+            image = image.convert("RGB")
+
+        if index < len(self) - 1:
+            weighting = Image.new("L", image.size)
+        else:
+            weighting = Image.open(self.target_mask)
+            weighting = exif_transpose(weighting)
+
+        image, weighting = self.transform(image, weighting)
+        example["images"], example["weightings"] = image, weighting < 0
+
+        if random.random() < 0.1:
+            example["masks"] = torch.ones_like(example["images"][0:1, :, :])
+        else:
+            example["masks"] = make_mask(example["images"], self.size)
+
+        example["conditioning_images"] = example["images"] * (example["masks"] < 0.5)
+
+        train_prompt = "" if random.random() < 0.1 else self.train_prompt
+        example["prompt_ids"] = self.tokenizer(
+            train_prompt,
+            truncation=True,
+            padding="max_length",
+            max_length=self.tokenizer.model_max_length,
+            return_tensors="pt",
+        ).input_ids
+
+        return example
+
+def collate_fn(examples):
+    input_ids = [example["prompt_ids"] for example in examples]
+    images = [example["images"] for example in examples]
+
+    masks = [example["masks"] for example in examples]
+    weightings = [example["weightings"] for example in examples]
+    conditioning_images = [example["conditioning_images"] for example in examples]
+
+    images = torch.stack(images)
+    images = images.to(memory_format=torch.contiguous_format).float()
+
+    masks = torch.stack(masks)
+    masks = masks.to(memory_format=torch.contiguous_format).float()
+
+    weightings = torch.stack(weightings)
+    weightings = weightings.to(memory_format=torch.contiguous_format).float()
+
+    conditioning_images = torch.stack(conditioning_images)
+    conditioning_images = conditioning_images.to(memory_format=torch.contiguous_format).float()
+
+    input_ids = torch.cat(input_ids, dim=0)
+
+    batch = {
+        "input_ids": input_ids,
+        "images": images,
+        "masks": masks,
+        "weightings": weightings,
+        "conditioning_images": conditioning_images,
+    }
+    return batch
+
+def main(args):
+    logging_dir = Path(args.output_dir, args.logging_dir)
+
+    accelerator = Accelerator(
+        gradient_accumulation_steps=args.gradient_accumulation_steps,
+        mixed_precision=args.mixed_precision,
+        log_with=args.report_to,
+        project_dir=logging_dir,
+    )
+
+    if args.report_to == "wandb":
+        if not is_wandb_available():
+            raise ImportError("Make sure to install wandb if you want to use it for logging during training.")
+        import wandb
+
+        wandb.login(key=args.wandb_key)
+        wandb.init(project=args.wandb_project_name)
+
+    # Make one log on every process with the configuration for debugging.
+    logging.basicConfig(
+        format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
+        datefmt="%m/%d/%Y %H:%M:%S",
+        level=logging.INFO,
+    )
+    logger.info(accelerator.state, main_process_only=False)
+    if accelerator.is_local_main_process:
+        transformers.utils.logging.set_verbosity_warning()
+        diffusers.utils.logging.set_verbosity_info()
+    else:
+        transformers.utils.logging.set_verbosity_error()
+        diffusers.utils.logging.set_verbosity_error()
+
+    # If passed along, set the training seed now.
+    if args.seed is not None:
+        set_seed(args.seed)
+
+    # Handle the repository creation
+    if accelerator.is_main_process:
+        if args.output_dir is not None:
+            os.makedirs(args.output_dir, exist_ok=True)
+
+        if args.push_to_hub:
+            repo_id = create_repo(
+                repo_id=args.hub_model_id or Path(args.output_dir).name, exist_ok=True, token=args.hub_token
+            ).repo_id
+
+    # Load the tokenizer
+    if args.tokenizer_name:
+        tokenizer = AutoTokenizer.from_pretrained(args.tokenizer_name, revision=args.revision, use_fast=False)
+    elif args.pretrained_model_name_or_path:
+        tokenizer = AutoTokenizer.from_pretrained(
+            args.pretrained_model_name_or_path,
+            subfolder="tokenizer",
+            revision=args.revision,
+            use_fast=False,
+        )
+
+    # Load scheduler and models
+    noise_scheduler = DDPMScheduler.from_pretrained(args.pretrained_model_name_or_path, subfolder="scheduler")
+    text_encoder = CLIPTextModel.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="text_encoder", revision=args.revision
+    )
+    vae = AutoencoderKL.from_pretrained(args.pretrained_model_name_or_path, subfolder="vae", revision=args.revision)
+    unet = UNet2DConditionModel.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="unet", revision=args.revision
+    )
+
+    config = LoraConfig(
+        r=args.lora_rank,
+        lora_alpha=args.lora_alpha,
+        target_modules=["to_k", "to_q", "to_v", "key", "query", "value"],
+        lora_dropout=args.lora_dropout,
+        bias=args.lora_bias,
+    )
+    unet = get_peft_model(unet, config)
+
+    config = LoraConfig(
+        r=args.lora_rank,
+        lora_alpha=args.lora_alpha,
+        target_modules=["k_proj", "q_proj", "v_proj"],
+        lora_dropout=args.lora_dropout,
+        bias=args.lora_bias,
+    )
+    text_encoder = get_peft_model(text_encoder, config)
+
+    vae.requires_grad_(False)
+
+    if args.enable_xformers_memory_efficient_attention:
+        if is_xformers_available():
+            import xformers
+
+            xformers_version = version.parse(xformers.__version__)
+            if xformers_version == version.parse("0.0.16"):
+                logger.warn(
+                    "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
+                )
+            unet.enable_xformers_memory_efficient_attention()
+        else:
+            raise ValueError("xformers is not available. Make sure it is installed correctly")
+
+    if args.gradient_checkpointing:
+        unet.enable_gradient_checkpointing()
+        text_encoder.gradient_checkpointing_enable()
+
+    # create custom saving & loading hooks so that `accelerator.save_state(...)` serializes in a nice format
+    def save_model_hook(models, weights, output_dir):
+        if accelerator.is_main_process:
+            for model in models:
+                sub_dir = "unet" if isinstance(model.base_model.model, type(accelerator.unwrap_model(unet.base_model.model))) else "text_encoder"
+                model.save_pretrained(os.path.join(output_dir, sub_dir))
+
+                # make sure to pop weight so that corresponding model is not saved again
+                weights.pop()
+
+    def load_model_hook(models, input_dir):
+        while len(models) > 0:
+            # pop models so that they are not loaded again
+            model = models.pop()
+
+            sub_dir = "unet" if isinstance(model.base_model.model, type(accelerator.unwrap_model(unet.base_model.model))) else "text_encoder"
+            model_cls = UNet2DConditionModel if isinstance(model.base_model.model, type(accelerator.unwrap_model(unet.base_model.model))) else CLIPTextModel
+
+            load_model = model_cls.from_pretrained(args.pretrained_model_name_or_path, subfolder=sub_dir)
+            load_model = PeftModel.from_pretrained(load_model, input_dir, subfolder=sub_dir)
+
+            model.load_state_dict(load_model.state_dict())
+            del load_model
+
+    accelerator.register_save_state_pre_hook(save_model_hook)
+    accelerator.register_load_state_pre_hook(load_model_hook)
+
+    # Enable TF32 for faster training on Ampere GPUs,
+    # cf https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices
+    if args.allow_tf32:
+        torch.backends.cuda.matmul.allow_tf32 = True
+
+    if args.scale_lr:
+        args.unet_learning_rate = (
+            args.unet_learning_rate * args.gradient_accumulation_steps * args.train_batch_size * accelerator.num_processes
+        )
+
+        args.text_encoder_learning_rate = (
+            args.text_encoder_learning_rate * args.gradient_accumulation_steps * args.train_batch_size * accelerator.num_processes
+        )
+
+    # Use 8-bit Adam for lower memory usage or to fine-tune the model in 16GB GPUs
+    if args.use_8bit_adam:
+        try:
+            import bitsandbytes as bnb
+        except ImportError:
+            raise ImportError(
+                "To use 8-bit Adam, please install the bitsandbytes library: `pip install bitsandbytes`."
+            )
+
+        optimizer_class = bnb.optim.AdamW8bit
+    else:
+        optimizer_class = torch.optim.AdamW
+
+    # Optimizer creation
+    optimizer = optimizer_class(
+        [
+            {"params": unet.parameters(), "lr": args.unet_learning_rate},
+            {"params": text_encoder.parameters(), "lr": args.text_encoder_learning_rate}
+        ],
+        betas=(args.adam_beta1, args.adam_beta2),
+        weight_decay=args.adam_weight_decay,
+        eps=args.adam_epsilon,
+    )
+
+    # Dataset and DataLoaders creation:
+    train_dataset = RealFillDataset(
+        train_data_root=args.train_data_dir,
+        tokenizer=tokenizer,
+        size=args.resolution,
+    )
+
+    train_dataloader = torch.utils.data.DataLoader(
+        train_dataset,
+        batch_size=args.train_batch_size,
+        shuffle=True,
+        collate_fn=collate_fn,
+        num_workers=1,
+    )
+
+    # Scheduler and math around the number of training steps.
+    overrode_max_train_steps = False
+    num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
+    if args.max_train_steps is None:
+        args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
+        overrode_max_train_steps = True
+
+    lr_scheduler = get_scheduler(
+        args.lr_scheduler,
+        optimizer=optimizer,
+        num_warmup_steps=args.lr_warmup_steps * args.gradient_accumulation_steps,
+        num_training_steps=args.max_train_steps * args.gradient_accumulation_steps,
+        num_cycles=args.lr_num_cycles,
+        power=args.lr_power,
+    )
+
+    # Prepare everything with our `accelerator`.
+    unet, text_encoder, optimizer, train_dataloader = accelerator.prepare(
+        unet, text_encoder, optimizer, train_dataloader
+    )
+
+    # For mixed precision training we cast all non-trainable weigths (vae, non-lora text_encoder and non-lora unet) to half-precision
+    # as these weights are only used for inference, keeping weights in full precision is not required.
+    weight_dtype = torch.float32
+    if accelerator.mixed_precision == "fp16":
+        weight_dtype = torch.float16
+    elif accelerator.mixed_precision == "bf16":
+        weight_dtype = torch.bfloat16
+
+    # Move vae to device and cast to weight_dtype
+    vae.to(accelerator.device, dtype=weight_dtype)
+
+    # We need to recalculate our total training steps as the size of the training dataloader may have changed.
+    num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
+    if overrode_max_train_steps:
+        args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
+    # Afterwards we recalculate our number of training epochs
+    args.num_train_epochs = math.ceil(args.max_train_steps / num_update_steps_per_epoch)
+
+    # We need to initialize the trackers we use, and also store our configuration.
+    # The trackers initializes automatically on the main process.
+    if accelerator.is_main_process:
+        tracker_config = vars(copy.deepcopy(args))
+        accelerator.init_trackers("realfill", config=tracker_config)
+
+    # Train!
+    total_batch_size = args.train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps
+
+    logger.info("***** Running training *****")
+    logger.info(f"  Num examples = {len(train_dataset)}")
+    logger.info(f"  Num batches each epoch = {len(train_dataloader)}")
+    logger.info(f"  Num Epochs = {args.num_train_epochs}")
+    logger.info(f"  Instantaneous batch size per device = {args.train_batch_size}")
+    logger.info(f"  Total train batch size (w. parallel, distributed & accumulation) = {total_batch_size}")
+    logger.info(f"  Gradient Accumulation steps = {args.gradient_accumulation_steps}")
+    logger.info(f"  Total optimization steps = {args.max_train_steps}")
+    global_step = 0
+    first_epoch = 0
+
+    # Potentially load in the weights and states from a previous save
+    if args.resume_from_checkpoint:
+        if args.resume_from_checkpoint != "latest":
+            path = os.path.basename(args.resume_from_checkpoint)
+        else:
+            # Get the mos recent checkpoint
+            dirs = os.listdir(args.output_dir)
+            dirs = [d for d in dirs if d.startswith("checkpoint")]
+            dirs = sorted(dirs, key=lambda x: int(x.split("-")[1]))
+            path = dirs[-1] if len(dirs) > 0 else None
+
+        if path is None:
+            accelerator.print(
+                f"Checkpoint '{args.resume_from_checkpoint}' does not exist. Starting a new training run."
+            )
+            args.resume_from_checkpoint = None
+            initial_global_step = 0
+        else:
+            accelerator.print(f"Resuming from checkpoint {path}")
+            accelerator.load_state(os.path.join(args.output_dir, path))
+            global_step = int(path.split("-")[1])
+
+            initial_global_step = global_step
+            first_epoch = global_step // num_update_steps_per_epoch
+    else:
+        initial_global_step = 0
+
+    progress_bar = tqdm(
+        range(0, args.max_train_steps),
+        initial=initial_global_step,
+        desc="Steps",
+        # Only show the progress bar once on each machine.
+        disable=not accelerator.is_local_main_process,
+    )
+
+    for epoch in range(first_epoch, args.num_train_epochs):
+        unet.train()
+        text_encoder.train()
+
+        for step, batch in enumerate(train_dataloader):
+            with accelerator.accumulate(unet, text_encoder):
+                # Convert images to latent space
+                latents = vae.encode(batch["images"].to(dtype=weight_dtype)).latent_dist.sample()
+                latents = latents * 0.18215
+
+                # Convert masked images to latent space
+                conditionings = vae.encode(batch["conditioning_images"].to(dtype=weight_dtype)).latent_dist.sample()
+                conditionings = conditionings * 0.18215
+
+                # Downsample mask and weighting so that they match with the latents
+                masks, size = batch["masks"].to(dtype=weight_dtype), latents.shape[2:]
+                masks = F.interpolate(masks, size=size)
+
+                weightings = batch["weightings"].to(dtype=weight_dtype)
+                weightings = F.interpolate(weightings, size=size)
+
+                # Sample noise that we'll add to the latents
+                noise = torch.randn_like(latents)
+                bsz = latents.shape[0]
+
+                # Sample a random timestep for each image
+                timesteps = torch.randint(
+                    0, noise_scheduler.config.num_train_timesteps, (bsz,), device=latents.device
+                )
+                timesteps = timesteps.long()
+
+                # Add noise to the latents according to the noise magnitude at each timestep
+                # (this is the forward diffusion process)
+                noisy_latents = noise_scheduler.add_noise(latents, noise, timesteps)
+
+                # Concatenate noisy latents, masks and conditionings to get inputs to unet
+                inputs = torch.cat([noisy_latents, masks, conditionings], dim=1)
+
+                # Get the text embedding for conditioning
+                encoder_hidden_states = text_encoder(batch["input_ids"])[0]
+
+                # Predict the noise residual
+                model_pred = unet(inputs, timesteps, encoder_hidden_states).sample
+
+                # Compute the diffusion loss
+                assert noise_scheduler.config.prediction_type == "epsilon"
+                loss = (weightings * F.mse_loss(model_pred.float(), noise.float(), reduction="none")).mean()
+
+                # Backpropagate
+                accelerator.backward(loss)
+                if accelerator.sync_gradients:
+                    params_to_clip = itertools.chain(
+                        unet.parameters(), text_encoder.parameters()
+                    )
+                    accelerator.clip_grad_norm_(params_to_clip, args.max_grad_norm)
+
+                optimizer.step()
+                lr_scheduler.step()
+                optimizer.zero_grad(set_to_none=args.set_grads_to_none)
+
+            # Checks if the accelerator has performed an optimization step behind the scenes
+            if accelerator.sync_gradients:
+                progress_bar.update(1)
+                if args.report_to == "wandb":
+                    accelerator.print(progress_bar)
+                global_step += 1
+
+                if accelerator.is_main_process:
+                    if global_step % args.checkpointing_steps == 0:
+                        # _before_ saving state, check if this save would set us over the `checkpoints_total_limit`
+                        if args.checkpoints_total_limit is not None:
+                            checkpoints = os.listdir(args.output_dir)
+                            checkpoints = [d for d in checkpoints if d.startswith("checkpoint")]
+                            checkpoints = sorted(checkpoints, key=lambda x: int(x.split("-")[1]))
+
+                            # before we save the new checkpoint, we need to have at _most_ `checkpoints_total_limit - 1` checkpoints
+                            if len(checkpoints) >= args.checkpoints_total_limit:
+                                num_to_remove = len(checkpoints) - args.checkpoints_total_limit + 1
+                                removing_checkpoints = checkpoints[0:num_to_remove]
+
+                                logger.info(
+                                    f"{len(checkpoints)} checkpoints already exist, removing {len(removing_checkpoints)} checkpoints"
+                                )
+                                logger.info(f"removing checkpoints: {', '.join(removing_checkpoints)}")
+
+                                for removing_checkpoint in removing_checkpoints:
+                                    removing_checkpoint = os.path.join(args.output_dir, removing_checkpoint)
+                                    shutil.rmtree(removing_checkpoint)
+
+                        save_path = os.path.join(args.output_dir, f"checkpoint-{global_step}")
+                        accelerator.save_state(save_path)
+                        logger.info(f"Saved state to {save_path}")
+
+                    if global_step % args.validation_steps == 0:
+                        log_validation(
+                            text_encoder,
+                            tokenizer,
+                            unet,
+                            args,
+                            accelerator,
+                            weight_dtype,
+                            global_step,
+                        )
+
+            logs = {"loss": loss.detach().item()}
+            progress_bar.set_postfix(**logs)
+            accelerator.log(logs, step=global_step)
+
+            if global_step >= args.max_train_steps:
+                break
+
+    # Save the lora layers
+    accelerator.wait_for_everyone()
+    if accelerator.is_main_process:
+        pipeline = StableDiffusionInpaintPipeline.from_pretrained(
+            args.pretrained_model_name_or_path,
+            unet=accelerator.unwrap_model(unet.merge_and_unload(), keep_fp32_wrapper=True),
+            text_encoder=accelerator.unwrap_model(text_encoder.merge_and_unload(), keep_fp32_wrapper=True),
+            revision=args.revision,
+        )
+
+        pipeline.save_pretrained(args.output_dir)
+
+        # Final inference
+        images = log_validation(
+            text_encoder,
+            tokenizer,
+            unet,
+            args,
+            accelerator,
+            weight_dtype,
+            global_step,
+        )
+
+        if args.push_to_hub:
+            save_model_card(
+                repo_id,
+                images=images,
+                base_model=args.pretrained_model_name_or_path,
+                repo_folder=args.output_dir,
+            )
+            upload_folder(
+                repo_id=repo_id,
+                folder_path=args.output_dir,
+                commit_message="End of training",
+                ignore_patterns=["step_*", "epoch_*"],
+            )
+
+    accelerator.end_training()
+
+if __name__ == "__main__":
+    args = parse_args()
+    main(args)

trainer.py

@@ -0,0 +1,168 @@
+from __future__ import annotations
+
+import datetime
+import os
+import pathlib
+import shlex
+import shutil
+import subprocess
+
+import gradio as gr
+import PIL.Image
+import slugify
+import torch
+from huggingface_hub import HfApi
+
+from app_upload import ModelUploader
+from utils import save_model_card
+
+URL_TO_JOIN_LIBRARY_ORG = 'https://huggingface.co/organizations/realfill-library/share/WctmaLvDHWxnuWoJxagTrzVXbGwxoqoJoG'
+
+class Trainer:
+    def __init__(self, hf_token: str | None = None):
+        self.hf_token = hf_token
+        self.api = HfApi(token=hf_token)
+        self.model_uploader = ModelUploader(hf_token)
+
+    def prepare_dataset(self, reference_images: list, resolution: int,
+                        target_image: PIL.Image, target_mask: PIL.Image,
+                        train_data_dir: pathlib.Path) -> None:
+        shutil.rmtree(train_data_dir, ignore_errors=True)
+        train_data_dir.mkdir(parents=True)
+        
+        (train_data_dir / 'ref').mkdir(parents=True)
+        (train_data_dir / 'target').mkdir(parents=True)
+
+        for i, temp_path in enumerate(reference_images):
+            image = PIL.Image.open(temp_path.name)
+            image = image.convert('RGB')
+            out_path = train_data_dir / 'ref' / f'{i:03d}.jpg'
+            image.save(out_path, format='JPEG', quality=100)
+
+        target_image = PIL.Image.open(target_image[0].name)
+        target_image = target_image.convert('RGB')
+        out_path = train_data_dir / 'target' / f'target.jpg'
+        target_image.save(out_path, format='JPEG', quality=100)
+
+        target_mask = PIL.Image.open(target_mask[0].name)
+        target_mask = target_mask.convert('L')
+        out_path = train_data_dir / 'target' / f'mask.jpg'
+        target_mask.save(out_path, format='JPEG', quality=100)
+
+    def join_library_org(self) -> None:
+        subprocess.run(
+            shlex.split(
+                f'curl -X POST -H "Authorization: Bearer {self.hf_token}" -H "Content-Type: application/json" {URL_TO_JOIN_LIBRARY_ORG}'
+            ))
+
+    def run(
+        self,
+        reference_images: list | None,
+        target_image: PIL.ImageFile | None,
+        target_mask: PIL.ImageFile | None,
+        output_model_name: str,
+        overwrite_existing_model: bool,
+        base_model: str,
+        resolution_s: str,
+        n_steps: int,
+        unet_learning_rate: float,
+        text_encoder_learning_rate: float,
+        lora_rank: int,
+        lora_dropout: float,
+        lora_alpha: int,
+        gradient_accumulation: int,
+        seed: int,
+        fp16: bool,
+        use_8bit_adam: bool,
+        checkpointing_steps: int,
+        use_wandb: bool,
+        validation_steps: int,
+        upload_to_hub: bool,
+        use_private_repo: bool,
+        delete_existing_repo: bool,
+        upload_to: str,
+        remove_gpu_after_training: bool,
+    ) -> str:
+        if not torch.cuda.is_available():
+            raise gr.Error('CUDA is not available.')
+        if reference_images is None:
+            raise gr.Error('You need to upload reference images.')
+        if target_image is None:
+            raise gr.Error('The instance prompt is missing.')
+
+        resolution = int(resolution_s)
+
+        if not output_model_name:
+            timestamp = datetime.datetime.now().strftime('%Y-%m-%d-%H-%M-%S')
+            output_model_name = f'realfill-{timestamp}'
+        output_model_name = slugify.slugify(output_model_name)
+
+        repo_dir = pathlib.Path(__file__).parent
+        output_dir = repo_dir / 'experiments' / output_model_name
+        if overwrite_existing_model or upload_to_hub:
+            shutil.rmtree(output_dir, ignore_errors=True)
+        output_dir.mkdir(parents=True)
+
+        train_data_dir = repo_dir / 'training_data' / output_model_name
+        self.prepare_dataset(reference_images, resolution, target_image, target_mask, train_data_dir)
+
+        if upload_to_hub:
+            self.join_library_org()
+
+        command = f'''
+        python train_realfill.py \
+          --pretrained_model_name_or_path={base_model}  \
+          --train_data_dir={train_data_dir} \
+          --output_dir={output_dir} \
+          --resolution={resolution} \
+          --train_batch_size=16 \
+          --gradient_accumulation_steps={gradient_accumulation} --gradient_checkpointing \
+          --unet_learning_rate={unet_learning_rate} \
+          --text_encoder_learning_rate={text_encoder_learning_rate} \
+          --lr_scheduler=constant \
+          --lr_warmup_steps=100 \
+          --set_grads_to_none \
+          --max_train_steps={n_steps} \
+          --checkpointing_steps={checkpointing_steps} \
+          --validation_steps={validation_steps} \
+          --lora_rank={lora_rank} \
+          --lora_dropout={lora_dropout} \
+          --lora_alpha={lora_alpha} \
+          --seed={seed}
+        '''
+        if fp16:
+            command += ' --mixed_precision fp16'
+        if use_8bit_adam:
+            command += ' --use_8bit_adam'
+        if use_wandb:
+            command += ' --report_to wandb'
+
+        with open(output_dir / 'train.sh', 'w') as f:
+            command_s = ' '.join(command.split())
+            f.write(command_s)
+        subprocess.run(shlex.split(command))
+        save_model_card(save_dir=output_dir,
+                        base_model=base_model,
+                        target_image=train_data_dir / 'target' / 'target.jpg',
+                        target_mask=train_data_dir / 'target' / 'mask.jpg')
+
+        message = 'Training completed!'
+        print(message)
+
+        if upload_to_hub:
+            upload_message = self.model_uploader.upload_model(
+                folder_path=output_dir.as_posix(),
+                repo_name=output_model_name,
+                upload_to=upload_to,
+                private=use_private_repo,
+                delete_existing_repo=delete_existing_repo)
+            print(upload_message)
+            message = message + '\n' + upload_message
+
+        if remove_gpu_after_training:
+            space_id = os.getenv('SPACE_ID')
+            if space_id:
+                self.api.request_space_hardware(repo_id=space_id,
+                                                hardware='cpu-basic')
+
+        return message

uploader.py

@@ -0,0 +1,42 @@
+from __future__ import annotations
+
+from huggingface_hub import HfApi
+
+
+class Uploader:
+    def __init__(self, hf_token: str | None):
+        self.api = HfApi(token=hf_token)
+
+    def get_username(self) -> str:
+        return self.api.whoami()['name']
+
+    def upload(self,
+               folder_path: str,
+               repo_name: str,
+               organization: str = '',
+               repo_type: str = 'model',
+               private: bool = True,
+               delete_existing_repo: bool = False) -> str:
+        if not folder_path:
+            raise ValueError
+        if not repo_name:
+            raise ValueError
+        if not organization:
+            organization = self.get_username()
+        repo_id = f'{organization}/{repo_name}'
+        if delete_existing_repo:
+            try:
+                self.api.delete_repo(repo_id, repo_type=repo_type)
+            except Exception:
+                pass
+        try:
+            self.api.create_repo(repo_id, repo_type=repo_type, private=private)
+            self.api.upload_folder(repo_id=repo_id,
+                                   folder_path=folder_path,
+                                   path_in_repo='.',
+                                   repo_type=repo_type)
+            url = f'https://huggingface.co/{repo_id}'
+            message = f'Your model was successfully uploaded to <a href="{url}" target="_blank">{url}</a>.'
+        except Exception as e:
+            message = str(e)
+        return message

utils.py

@@ -0,0 +1,48 @@
+from __future__ import annotations
+
+import pathlib
+
+
+def find_exp_dirs(ignore_repo: bool = False) -> list[str]:
+    repo_dir = pathlib.Path(__file__).parent
+    exp_root_dir = repo_dir / 'experiments'
+    if not exp_root_dir.exists():
+        return []
+    exp_dirs = sorted(exp_root_dir.glob('*'))
+    exp_dirs = [
+        exp_dir for exp_dir in exp_dirs
+        if (exp_dir / 'model_index.json').exists()
+    ]
+    if ignore_repo:
+        exp_dirs = [
+            exp_dir for exp_dir in exp_dirs if not (exp_dir / '.git').exists()
+        ]
+    return [path.relative_to(repo_dir).as_posix() for path in exp_dirs]
+
+
+def save_model_card(
+    save_dir: pathlib.Path,
+    base_model: str,
+    target_image: str,
+    target_mask: str,
+) -> None:
+    model_card = f'''---
+license: creativeml-openrail-m
+base_model: {base_model}
+target_image: {target_image}
+target_mask: {target_mask}
+tags:
+- stable-diffusion-inpainting
+- stable-diffusion-inpainting-diffusers
+- text-to-image
+- diffusers
+- realfill
+inference: true
+---
+# RealFill - {save_dir.name}
+
+These are RealFill weights for [{base_model}](https://huggingface.co/{base_model}). The weights were trained using [RealFill](https://realfill.github.io/).
+'''
+
+    with open(save_dir / 'README.md', 'w') as f:
+        f.write(model_card)