Duplicate from smangrul/peft-lora-sd-dreambooth

Co-authored-by: Sourab Mangrulkar <smangrul@users.noreply.huggingface.co>

.gitattributes

@@ -0,0 +1,34 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.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
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,14 @@
+---
+title: Peft Lora Sd Dreambooth
+emoji: 🎨
+colorFrom: purple
+colorTo: green
+sdk: gradio
+sdk_version: 3.16.2
+app_file: app.py
+pinned: false
+license: openrail
+duplicated_from: smangrul/peft-lora-sd-dreambooth
+---
+
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,371 @@
+#!/usr/bin/env python
+"""
+Demo showcasing parameter-efficient fine-tuning of Stable Dissfusion via Dreambooth leveraging 🤗 PEFT (https://github.com/huggingface/peft)
+
+The code in this repo is partly adapted from the following repositories:
+https://huggingface.co/spaces/hysts/LoRA-SD-training
+https://huggingface.co/spaces/multimodalart/dreambooth-training
+"""
+from __future__ import annotations
+
+import os
+import pathlib
+
+import gradio as gr
+import torch
+from typing import List
+
+from inference import InferencePipeline
+from trainer import Trainer
+from uploader import upload
+
+
+TITLE = "# LoRA + Dreambooth Training and Inference Demo 🎨"
+DESCRIPTION = "Demo showcasing parameter-efficient fine-tuning of Stable Dissfusion via Dreambooth leveraging 🤗 PEFT (https://github.com/huggingface/peft)."
+
+
+ORIGINAL_SPACE_ID = "smangrul/peft-lora-sd-dreambooth"
+
+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>
+"""
+
+
+def show_warning(warning_text: str) -> gr.Blocks:
+    with gr.Blocks() as demo:
+        with gr.Box():
+            gr.Markdown(warning_text)
+    return demo
+
+
+def update_output_files() -> dict:
+    paths = sorted(pathlib.Path("results").glob("*.pt"))
+    config_paths = sorted(pathlib.Path("results").glob("*.json"))
+    paths = paths + config_paths
+    paths = [path.as_posix() for path in paths]  # type: ignore
+    return gr.update(value=paths or None)
+
+
+def create_training_demo(trainer: Trainer, pipe: InferencePipeline) -> gr.Blocks:
+    with gr.Blocks() as demo:
+        base_model = gr.Dropdown(
+            choices=[
+                "CompVis/stable-diffusion-v1-4",
+                "runwayml/stable-diffusion-v1-5",
+                "stabilityai/stable-diffusion-2-1-base",
+            ],
+            value="runwayml/stable-diffusion-v1-5",
+            label="Base Model",
+            visible=True,
+        )
+        resolution = gr.Dropdown(choices=["512"], value="512", label="Resolution", visible=False)
+
+        with gr.Row():
+            with gr.Box():
+                gr.Markdown("Training Data")
+                concept_images = gr.Files(label="Images for your concept")
+                concept_prompt = gr.Textbox(label="Concept Prompt", max_lines=1)
+                gr.Markdown(
+                    """
+                    - Upload images of the style you are planning on training on.
+                    - For a concept prompt, use a unique, made up word to avoid collisions.
+                    - Guidelines for getting good results:
+                        - Dreambooth for an `object` or `style`:
+                            - 5-10 images of the object from different angles
+                            - 500-800 iterations should be good enough. 
+                            - Prior preservation is recommended.
+                            - `class_prompt`:
+                                - `a photo of object`
+                                - `style`
+                            - `concept_prompt`:
+                                - `<concept prompt> object`
+                                - `<concept prompt> style`
+                                - `a photo of <concept prompt> object`
+                                - `a photo of <concept prompt> style`
+                        - Dreambooth for a `Person/Face`:
+                            - 15-50 images of the person from different angles, lighting, and expressions. 
+                            Have considerable photos with close up faces.
+                            - 800-1200 iterations should be good enough.
+                            - good defaults for hyperparams
+                                - Model - `runwayml/stable-diffusion-v1-5` or `stabilityai/stable-diffusion-2-1-base`
+                                - Use/check Prior preservation.
+                                - Number of class images to use - 200
+                                - Prior Loss Weight - 1
+                                - LoRA Rank for unet - 16
+                                - LoRA Alpha for unet - 20
+                                - lora dropout - 0
+                                - LoRA Bias for unet - `all`
+                                - LoRA Rank for CLIP - 16
+                                - LoRA Alpha for CLIP - 17
+                                - LoRA Bias for CLIP - `all`
+                                - lora dropout for CLIP - 0
+                                - Uncheck `FP16` and `8bit-Adam` (don't use them for faces)
+                            - `class_prompt`: Use the gender related word of the person
+                                - `man`
+                                - `woman`
+                                - `boy`
+                                - `girl`
+                            - `concept_prompt`: just the unique, made up word, e.g., `srm`
+                            - Choose `all` for `lora_bias` and `text_encode_lora_bias`
+                        - Dreambooth for a `Scene`:
+                            - 15-50 images of the scene from different angles, lighting, and expressions.
+                            - 800-1200 iterations should be good enough.
+                            - Prior preservation is recommended.
+                            - `class_prompt`:
+                                - `scene`
+                                - `landscape`
+                                - `city`
+                                - `beach`
+                                - `mountain`
+                            - `concept_prompt`:
+                                - `<concept prompt> scene`
+                                - `<concept prompt> landscape`
+                        - Experiment with various values for lora dropouts, enabling/disabling fp16 and 8bit-Adam
+                    """
+                )
+            with gr.Box():
+                gr.Markdown("Training Parameters")
+                num_training_steps = gr.Number(label="Number of Training Steps", value=1000, precision=0)
+                learning_rate = gr.Number(label="Learning Rate", value=0.0001)
+                gradient_checkpointing = gr.Checkbox(label="Whether to use gradient checkpointing", value=True)
+                train_text_encoder = gr.Checkbox(label="Train Text Encoder", value=True)
+                with_prior_preservation = gr.Checkbox(label="Prior Preservation", value=True)
+                class_prompt = gr.Textbox(
+                    label="Class Prompt", max_lines=1, placeholder='Example: "a photo of object"'
+                )
+                num_class_images = gr.Number(label="Number of class images to use", value=50, precision=0)
+                prior_loss_weight = gr.Number(label="Prior Loss Weight", value=1.0, precision=1)
+                # use_lora = gr.Checkbox(label="Whether to use LoRA", value=True)
+                lora_r = gr.Number(label="LoRA Rank for unet", value=4, precision=0)
+                lora_alpha = gr.Number(
+                    label="LoRA Alpha for unet. scaling factor = lora_alpha/lora_r", value=4, precision=0
+                )
+                lora_dropout = gr.Number(label="lora dropout", value=0.00)
+                lora_bias = gr.Dropdown(
+                    choices=["none", "all", "lora_only"],
+                    value="none",
+                    label="LoRA Bias for unet. This enables bias params to be trainable based on the bias type",
+                    visible=True,
+                )
+                lora_text_encoder_r = gr.Number(label="LoRA Rank for CLIP", value=4, precision=0)
+                lora_text_encoder_alpha = gr.Number(
+                    label="LoRA Alpha for CLIP. scaling factor = lora_alpha/lora_r", value=4, precision=0
+                )
+                lora_text_encoder_dropout = gr.Number(label="lora dropout for CLIP", value=0.00)
+                lora_text_encoder_bias = gr.Dropdown(
+                    choices=["none", "all", "lora_only"],
+                    value="none",
+                    label="LoRA Bias for CLIP. This enables bias params to be trainable based on the bias type",
+                    visible=True,
+                )
+                gradient_accumulation = gr.Number(label="Number of Gradient Accumulation", value=1, precision=0)
+                fp16 = gr.Checkbox(label="FP16", value=True)
+                use_8bit_adam = gr.Checkbox(label="Use 8bit Adam", value=True)
+                gr.Markdown(
+                    """
+                    - It will take about 20-30 minutes to train for 1000 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.
+                    - Note that your trained models will be deleted when the second training is started. You can upload your trained model in the "Upload" tab.
+                    """
+                )
+
+        run_button = gr.Button("Start Training")
+        with gr.Box():
+            with gr.Row():
+                check_status_button = gr.Button("Check Training Status")
+                with gr.Column():
+                    with gr.Box():
+                        gr.Markdown("Message")
+                        training_status = gr.Markdown()
+                    output_files = gr.Files(label="Trained Weight Files and Configs")
+
+        run_button.click(fn=pipe.clear)
+
+        run_button.click(
+            fn=trainer.run,
+            inputs=[
+                base_model,
+                resolution,
+                num_training_steps,
+                concept_images,
+                concept_prompt,
+                learning_rate,
+                gradient_accumulation,
+                fp16,
+                use_8bit_adam,
+                gradient_checkpointing,
+                train_text_encoder,
+                with_prior_preservation,
+                prior_loss_weight,
+                class_prompt,
+                num_class_images,
+                lora_r,
+                lora_alpha,
+                lora_bias,
+                lora_dropout,
+                lora_text_encoder_r,
+                lora_text_encoder_alpha,
+                lora_text_encoder_bias,
+                lora_text_encoder_dropout,
+            ],
+            outputs=[
+                training_status,
+                output_files,
+            ],
+            queue=False,
+        )
+        check_status_button.click(fn=trainer.check_if_running, inputs=None, outputs=training_status, queue=False)
+        check_status_button.click(fn=update_output_files, inputs=None, outputs=output_files, queue=False)
+    return demo
+
+
+def find_weight_files() -> List[str]:
+    curr_dir = pathlib.Path(__file__).parent
+    paths = sorted(curr_dir.rglob("*.pt"))
+    return [path.relative_to(curr_dir).as_posix() for path in paths]
+
+
+def reload_lora_weight_list() -> dict:
+    return gr.update(choices=find_weight_files())
+
+
+def create_inference_demo(pipe: InferencePipeline) -> gr.Blocks:
+    with gr.Blocks() as demo:
+        with gr.Row():
+            with gr.Column():
+                base_model = gr.Dropdown(
+                    choices=[
+                        "CompVis/stable-diffusion-v1-4",
+                        "runwayml/stable-diffusion-v1-5",
+                        "stabilityai/stable-diffusion-2-1-base",
+                    ],
+                    value="runwayml/stable-diffusion-v1-5",
+                    label="Base Model",
+                    visible=True,
+                )
+                reload_button = gr.Button("Reload Weight List")
+                lora_weight_name = gr.Dropdown(
+                    choices=find_weight_files(), value="lora/lora_disney.pt", label="LoRA Weight File"
+                )
+                prompt = gr.Textbox(label="Prompt", max_lines=1, placeholder='Example: "style of sks, baby lion"')
+                negative_prompt = gr.Textbox(
+                    label="Negative Prompt", max_lines=1, placeholder='Example: "blurry, botched, low quality"'
+                )
+                seed = gr.Slider(label="Seed", minimum=0, maximum=100000, step=1, value=1)
+                with gr.Accordion("Other Parameters", open=False):
+                    num_steps = gr.Slider(label="Number of Steps", minimum=0, maximum=1000, step=1, value=50)
+                    guidance_scale = gr.Slider(label="CFG Scale", minimum=0, maximum=50, step=0.1, value=7)
+
+                run_button = gr.Button("Generate")
+
+                gr.Markdown(
+                    """
+                - After training, you can press "Reload Weight List" button to load your trained model names.
+                - Few repos to refer for ideas:
+                    - https://huggingface.co/smangrul/smangrul
+                    - https://huggingface.co/smangrul/painting-in-the-style-of-smangrul
+                    - https://huggingface.co/smangrul/erenyeager
+                """
+                )
+            with gr.Column():
+                result = gr.Image(label="Result")
+
+        reload_button.click(fn=reload_lora_weight_list, inputs=None, outputs=lora_weight_name)
+        prompt.submit(
+            fn=pipe.run,
+            inputs=[
+                base_model,
+                lora_weight_name,
+                prompt,
+                negative_prompt,
+                seed,
+                num_steps,
+                guidance_scale,
+            ],
+            outputs=result,
+            queue=False,
+        )
+        run_button.click(
+            fn=pipe.run,
+            inputs=[
+                base_model,
+                lora_weight_name,
+                prompt,
+                negative_prompt,
+                seed,
+                num_steps,
+                guidance_scale,
+            ],
+            outputs=result,
+            queue=False,
+        )
+        seed.change(
+            fn=pipe.run,
+            inputs=[
+                base_model,
+                lora_weight_name,
+                prompt,
+                negative_prompt,
+                seed,
+                num_steps,
+                guidance_scale,
+            ],
+            outputs=result,
+            queue=False,
+        )
+    return demo
+
+
+def create_upload_demo() -> gr.Blocks:
+    with gr.Blocks() as demo:
+        model_name = gr.Textbox(label="Model Name")
+        hf_token = gr.Textbox(label="Hugging Face Token (with write permission)")
+        upload_button = gr.Button("Upload")
+        with gr.Box():
+            gr.Markdown("Message")
+            result = gr.Markdown()
+        gr.Markdown(
+            """
+            - You can upload your trained model to your private Model repo (i.e. https://huggingface.co/{your_username}/{model_name}).
+            - You can find your Hugging Face token [here](https://huggingface.co/settings/tokens).
+            """
+        )
+
+    upload_button.click(fn=upload, inputs=[model_name, hf_token], outputs=result)
+
+    return demo
+
+
+pipe = InferencePipeline()
+trainer = Trainer()
+
+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)
+
+    gr.Markdown(TITLE)
+    gr.Markdown(DESCRIPTION)
+
+    with gr.Tabs():
+        with gr.TabItem("Train"):
+            create_training_demo(trainer, pipe)
+        with gr.TabItem("Test"):
+            create_inference_demo(pipe)
+        with gr.TabItem("Upload"):
+            create_upload_demo()
+
+demo.queue(default_enabled=False).launch(share=False)

colab.py

@@ -0,0 +1,371 @@
+#!/usr/bin/env python
+"""
+Demo showcasing parameter-efficient fine-tuning of Stable Dissfusion via Dreambooth leveraging 🤗 PEFT (https://github.com/huggingface/peft)
+
+The code in this repo is partly adapted from the following repositories:
+https://huggingface.co/spaces/hysts/LoRA-SD-training
+https://huggingface.co/spaces/multimodalart/dreambooth-training
+"""
+from __future__ import annotations
+
+import os
+import pathlib
+
+import gradio as gr
+import torch
+from typing import List
+
+from inference import InferencePipeline
+from trainer import Trainer
+from uploader import upload
+
+
+TITLE = "# LoRA + Dreambooth Training and Inference Demo 🎨"
+DESCRIPTION = "Demo showcasing parameter-efficient fine-tuning of Stable Dissfusion via Dreambooth leveraging 🤗 PEFT (https://github.com/huggingface/peft)."
+
+
+ORIGINAL_SPACE_ID = "smangrul/peft-lora-sd-dreambooth"
+
+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>
+"""
+
+
+def show_warning(warning_text: str) -> gr.Blocks:
+    with gr.Blocks() as demo:
+        with gr.Box():
+            gr.Markdown(warning_text)
+    return demo
+
+
+def update_output_files() -> dict:
+    paths = sorted(pathlib.Path("results").glob("*.pt"))
+    config_paths = sorted(pathlib.Path("results").glob("*.json"))
+    paths = paths + config_paths
+    paths = [path.as_posix() for path in paths]  # type: ignore
+    return gr.update(value=paths or None)
+
+
+def create_training_demo(trainer: Trainer, pipe: InferencePipeline) -> gr.Blocks:
+    with gr.Blocks() as demo:
+        base_model = gr.Dropdown(
+            choices=[
+                "CompVis/stable-diffusion-v1-4",
+                "runwayml/stable-diffusion-v1-5",
+                "stabilityai/stable-diffusion-2-1-base",
+            ],
+            value="runwayml/stable-diffusion-v1-5",
+            label="Base Model",
+            visible=True,
+        )
+        resolution = gr.Dropdown(choices=["512"], value="512", label="Resolution", visible=False)
+
+        with gr.Row():
+            with gr.Box():
+                gr.Markdown("Training Data")
+                concept_images = gr.Files(label="Images for your concept")
+                concept_prompt = gr.Textbox(label="Concept Prompt", max_lines=1)
+                gr.Markdown(
+                    """
+                    - Upload images of the style you are planning on training on.
+                    - For a concept prompt, use a unique, made up word to avoid collisions.
+                    - Guidelines for getting good results:
+                        - Dreambooth for an `object` or `style`:
+                            - 5-10 images of the object from different angles
+                            - 500-800 iterations should be good enough. 
+                            - Prior preservation is recommended.
+                            - `class_prompt`:
+                                - `a photo of object`
+                                - `style`
+                            - `concept_prompt`:
+                                - `<concept prompt> object`
+                                - `<concept prompt> style`
+                                - `a photo of <concept prompt> object`
+                                - `a photo of <concept prompt> style`
+                        - Dreambooth for a `Person/Face`:
+                            - 15-50 images of the person from different angles, lighting, and expressions. 
+                            Have considerable photos with close up faces.
+                            - 800-1200 iterations should be good enough.
+                            - good defaults for hyperparams
+                                - Model - `runwayml/stable-diffusion-v1-5` or `stabilityai/stable-diffusion-2-1-base`
+                                - Use/check Prior preservation.
+                                - Number of class images to use - 200
+                                - Prior Loss Weight - 1
+                                - LoRA Rank for unet - 16
+                                - LoRA Alpha for unet - 20
+                                - lora dropout - 0
+                                - LoRA Bias for unet - `all`
+                                - LoRA Rank for CLIP - 16
+                                - LoRA Alpha for CLIP - 17
+                                - LoRA Bias for CLIP - `all`
+                                - lora dropout for CLIP - 0
+                                - Uncheck `FP16` and `8bit-Adam` (don't use them for faces)
+                            - `class_prompt`: Use the gender related word of the person
+                                - `man`
+                                - `woman`
+                                - `boy`
+                                - `girl`
+                            - `concept_prompt`: just the unique, made up word, e.g., `srm`
+                            - Choose `all` for `lora_bias` and `text_encode_lora_bias`
+                        - Dreambooth for a `Scene`:
+                            - 15-50 images of the scene from different angles, lighting, and expressions.
+                            - 800-1200 iterations should be good enough.
+                            - Prior preservation is recommended.
+                            - `class_prompt`:
+                                - `scene`
+                                - `landscape`
+                                - `city`
+                                - `beach`
+                                - `mountain`
+                            - `concept_prompt`:
+                                - `<concept prompt> scene`
+                                - `<concept prompt> landscape`
+                        - Experiment with various values for lora dropouts, enabling/disabling fp16 and 8bit-Adam
+                    """
+                )
+            with gr.Box():
+                gr.Markdown("Training Parameters")
+                num_training_steps = gr.Number(label="Number of Training Steps", value=1000, precision=0)
+                learning_rate = gr.Number(label="Learning Rate", value=0.0001)
+                gradient_checkpointing = gr.Checkbox(label="Whether to use gradient checkpointing", value=True)
+                train_text_encoder = gr.Checkbox(label="Train Text Encoder", value=True)
+                with_prior_preservation = gr.Checkbox(label="Prior Preservation", value=True)
+                class_prompt = gr.Textbox(
+                    label="Class Prompt", max_lines=1, placeholder='Example: "a photo of object"'
+                )
+                num_class_images = gr.Number(label="Number of class images to use", value=50, precision=0)
+                prior_loss_weight = gr.Number(label="Prior Loss Weight", value=1.0, precision=1)
+                # use_lora = gr.Checkbox(label="Whether to use LoRA", value=True)
+                lora_r = gr.Number(label="LoRA Rank for unet", value=4, precision=0)
+                lora_alpha = gr.Number(
+                    label="LoRA Alpha for unet. scaling factor = lora_r/lora_alpha", value=4, precision=0
+                )
+                lora_dropout = gr.Number(label="lora dropout", value=0.00)
+                lora_bias = gr.Dropdown(
+                    choices=["none", "all", "lora_only"],
+                    value="none",
+                    label="LoRA Bias for unet. This enables bias params to be trainable based on the bias type",
+                    visible=True,
+                )
+                lora_text_encoder_r = gr.Number(label="LoRA Rank for CLIP", value=4, precision=0)
+                lora_text_encoder_alpha = gr.Number(
+                    label="LoRA Alpha for CLIP. scaling factor = lora_r/lora_alpha", value=4, precision=0
+                )
+                lora_text_encoder_dropout = gr.Number(label="lora dropout for CLIP", value=0.00)
+                lora_text_encoder_bias = gr.Dropdown(
+                    choices=["none", "all", "lora_only"],
+                    value="none",
+                    label="LoRA Bias for CLIP. This enables bias params to be trainable based on the bias type",
+                    visible=True,
+                )
+                gradient_accumulation = gr.Number(label="Number of Gradient Accumulation", value=1, precision=0)
+                fp16 = gr.Checkbox(label="FP16", value=True)
+                use_8bit_adam = gr.Checkbox(label="Use 8bit Adam", value=True)
+                gr.Markdown(
+                    """
+                    - It will take about 20-30 minutes to train for 1000 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.
+                    - Note that your trained models will be deleted when the second training is started. You can upload your trained model in the "Upload" tab.
+                    """
+                )
+
+        run_button = gr.Button("Start Training")
+        with gr.Box():
+            with gr.Row():
+                check_status_button = gr.Button("Check Training Status")
+                with gr.Column():
+                    with gr.Box():
+                        gr.Markdown("Message")
+                        training_status = gr.Markdown()
+                    output_files = gr.Files(label="Trained Weight Files and Configs")
+
+        run_button.click(fn=pipe.clear)
+
+        run_button.click(
+            fn=trainer.run,
+            inputs=[
+                base_model,
+                resolution,
+                num_training_steps,
+                concept_images,
+                concept_prompt,
+                learning_rate,
+                gradient_accumulation,
+                fp16,
+                use_8bit_adam,
+                gradient_checkpointing,
+                train_text_encoder,
+                with_prior_preservation,
+                prior_loss_weight,
+                class_prompt,
+                num_class_images,
+                lora_r,
+                lora_alpha,
+                lora_bias,
+                lora_dropout,
+                lora_text_encoder_r,
+                lora_text_encoder_alpha,
+                lora_text_encoder_bias,
+                lora_text_encoder_dropout,
+            ],
+            outputs=[
+                training_status,
+                output_files,
+            ],
+            queue=False,
+        )
+        check_status_button.click(fn=trainer.check_if_running, inputs=None, outputs=training_status, queue=False)
+        check_status_button.click(fn=update_output_files, inputs=None, outputs=output_files, queue=False)
+    return demo
+
+
+def find_weight_files() -> List[str]:
+    curr_dir = pathlib.Path(__file__).parent
+    paths = sorted(curr_dir.rglob("*.pt"))
+    return [path.relative_to(curr_dir).as_posix() for path in paths]
+
+
+def reload_lora_weight_list() -> dict:
+    return gr.update(choices=find_weight_files())
+
+
+def create_inference_demo(pipe: InferencePipeline) -> gr.Blocks:
+    with gr.Blocks() as demo:
+        with gr.Row():
+            with gr.Column():
+                base_model = gr.Dropdown(
+                    choices=[
+                        "CompVis/stable-diffusion-v1-4",
+                        "runwayml/stable-diffusion-v1-5",
+                        "stabilityai/stable-diffusion-2-1-base",
+                    ],
+                    value="runwayml/stable-diffusion-v1-5",
+                    label="Base Model",
+                    visible=True,
+                )
+                reload_button = gr.Button("Reload Weight List")
+                lora_weight_name = gr.Dropdown(
+                    choices=find_weight_files(), value="lora/lora_disney.pt", label="LoRA Weight File"
+                )
+                prompt = gr.Textbox(label="Prompt", max_lines=1, placeholder='Example: "style of sks, baby lion"')
+                negative_prompt = gr.Textbox(
+                    label="Negative Prompt", max_lines=1, placeholder='Example: "blurry, botched, low quality"'
+                )
+                seed = gr.Slider(label="Seed", minimum=0, maximum=100000, step=1, value=1)
+                with gr.Accordion("Other Parameters", open=False):
+                    num_steps = gr.Slider(label="Number of Steps", minimum=0, maximum=1000, step=1, value=50)
+                    guidance_scale = gr.Slider(label="CFG Scale", minimum=0, maximum=50, step=0.1, value=7)
+
+                run_button = gr.Button("Generate")
+
+                gr.Markdown(
+                    """
+                - After training, you can press "Reload Weight List" button to load your trained model names.
+                - Few repos to refer for ideas:
+                    - https://huggingface.co/smangrul/smangrul
+                    - https://huggingface.co/smangrul/painting-in-the-style-of-smangrul
+                    - https://huggingface.co/smangrul/erenyeager
+                """
+                )
+            with gr.Column():
+                result = gr.Image(label="Result")
+
+        reload_button.click(fn=reload_lora_weight_list, inputs=None, outputs=lora_weight_name)
+        prompt.submit(
+            fn=pipe.run,
+            inputs=[
+                base_model,
+                lora_weight_name,
+                prompt,
+                negative_prompt,
+                seed,
+                num_steps,
+                guidance_scale,
+            ],
+            outputs=result,
+            queue=False,
+        )
+        run_button.click(
+            fn=pipe.run,
+            inputs=[
+                base_model,
+                lora_weight_name,
+                prompt,
+                negative_prompt,
+                seed,
+                num_steps,
+                guidance_scale,
+            ],
+            outputs=result,
+            queue=False,
+        )
+        seed.change(
+            fn=pipe.run,
+            inputs=[
+                base_model,
+                lora_weight_name,
+                prompt,
+                negative_prompt,
+                seed,
+                num_steps,
+                guidance_scale,
+            ],
+            outputs=result,
+            queue=False,
+        )
+    return demo
+
+
+def create_upload_demo() -> gr.Blocks:
+    with gr.Blocks() as demo:
+        model_name = gr.Textbox(label="Model Name")
+        hf_token = gr.Textbox(label="Hugging Face Token (with write permission)")
+        upload_button = gr.Button("Upload")
+        with gr.Box():
+            gr.Markdown("Message")
+            result = gr.Markdown()
+        gr.Markdown(
+            """
+            - You can upload your trained model to your private Model repo (i.e. https://huggingface.co/{your_username}/{model_name}).
+            - You can find your Hugging Face token [here](https://huggingface.co/settings/tokens).
+            """
+        )
+
+    upload_button.click(fn=upload, inputs=[model_name, hf_token], outputs=result)
+
+    return demo
+
+
+pipe = InferencePipeline()
+trainer = Trainer()
+
+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)
+
+    gr.Markdown(TITLE)
+    gr.Markdown(DESCRIPTION)
+
+    with gr.Tabs():
+        with gr.TabItem("Train"):
+            create_training_demo(trainer, pipe)
+        with gr.TabItem("Test"):
+            create_inference_demo(pipe)
+        with gr.TabItem("Upload"):
+            create_upload_demo()
+
+demo.queue(default_enabled=False).launch(share=True)

inference.py

@@ -0,0 +1,91 @@
+from __future__ import annotations
+
+import gc
+import json
+import pathlib
+import sys
+
+import gradio as gr
+import PIL.Image
+import torch
+from diffusers import StableDiffusionPipeline
+from peft import LoraModel, LoraConfig, set_peft_model_state_dict
+
+
+class InferencePipeline:
+    def __init__(self):
+        self.pipe = None
+        self.device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
+        self.weight_path = None
+
+    def clear(self) -> None:
+        self.weight_path = None
+        del self.pipe
+        self.pipe = None
+        torch.cuda.empty_cache()
+        gc.collect()
+
+    @staticmethod
+    def get_lora_weight_path(name: str) -> pathlib.Path:
+        curr_dir = pathlib.Path(__file__).parent
+        return curr_dir / name, curr_dir / f'{name.replace(".pt", "_config.json")}'
+
+    def load_and_set_lora_ckpt(self, pipe, weight_path, config_path, dtype):
+        with open(config_path, "r") as f:
+            lora_config = json.load(f)
+        lora_checkpoint_sd = torch.load(weight_path, map_location=self.device)
+        unet_lora_ds = {k: v for k, v in lora_checkpoint_sd.items() if "text_encoder_" not in k}
+        text_encoder_lora_ds = {
+            k.replace("text_encoder_", ""): v for k, v in lora_checkpoint_sd.items() if "text_encoder_" in k
+        }
+
+        unet_config = LoraConfig(**lora_config["peft_config"])
+        pipe.unet = LoraModel(unet_config, pipe.unet)
+        set_peft_model_state_dict(pipe.unet, unet_lora_ds)
+
+        if "text_encoder_peft_config" in lora_config:
+            text_encoder_config = LoraConfig(**lora_config["text_encoder_peft_config"])
+            pipe.text_encoder = LoraModel(text_encoder_config, pipe.text_encoder)
+            set_peft_model_state_dict(pipe.text_encoder, text_encoder_lora_ds)
+
+        if dtype in (torch.float16, torch.bfloat16):
+            pipe.unet.half()
+            pipe.text_encoder.half()
+
+        pipe.to(self.device)
+        return pipe
+
+    def load_pipe(self, model_id: str, lora_filename: str) -> None:
+        weight_path, config_path = self.get_lora_weight_path(lora_filename)
+        if weight_path == self.weight_path:
+            return
+
+        pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16).to(self.device)
+        pipe = pipe.to(self.device)
+        pipe = self.load_and_set_lora_ckpt(pipe, weight_path, config_path, torch.float16)
+        self.pipe = pipe
+
+    def run(
+        self,
+        base_model: str,
+        lora_weight_name: str,
+        prompt: str,
+        negative_prompt: 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(base_model, lora_weight_name)
+
+        generator = torch.Generator(device=self.device).manual_seed(seed)
+        out = self.pipe(
+            prompt,
+            num_inference_steps=n_steps,
+            guidance_scale=guidance_scale,
+            generator=generator,
+            negative_prompt=negative_prompt if negative_prompt else None,
+        )  # type: ignore
+        return out.images[0]

requirements.txt

@@ -0,0 +1,12 @@
+torch
+torchvision
+peft
+datasets
+git+https://github.com/huggingface/accelerate
+git+https://github.com/huggingface/diffusers
+git+https://github.com/huggingface/transformers
+tqdm
+ftfy
+Pillow
+bitsandbytes
+gradio

style.css

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

train_dreambooth.py

@@ -0,0 +1,1005 @@
+import argparse
+import gc
+import hashlib
+import itertools
+import json
+import logging
+import math
+import os
+import threading
+import warnings
+from pathlib import Path
+from typing import Optional
+
+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 torch.utils.data import Dataset
+from transformers import AutoTokenizer, PretrainedConfig
+
+import datasets
+import diffusers
+import psutil
+from diffusers import AutoencoderKL, DDPMScheduler, DiffusionPipeline, UNet2DConditionModel
+from diffusers.optimization import get_scheduler
+from diffusers.utils import check_min_version
+from diffusers.utils.import_utils import is_xformers_available
+from huggingface_hub import HfFolder, Repository, whoami
+from peft import LoraConfig, LoraModel, get_peft_model_state_dict
+from PIL import Image
+from torchvision import transforms
+from tqdm.auto import tqdm
+
+
+# Will error if the minimal version of diffusers is not installed. Remove at your own risks.
+check_min_version("0.10.0.dev0")
+
+logger = get_logger(__name__)
+
+UNET_TARGET_MODULES = ["to_q", "to_v", "query", "value"]  # , "ff.net.0.proj"]
+TEXT_ENCODER_TARGET_MODULES = ["q_proj", "v_proj"]
+
+
+def import_model_class_from_model_name_or_path(pretrained_model_name_or_path: str, revision: str):
+    text_encoder_config = PretrainedConfig.from_pretrained(
+        pretrained_model_name_or_path,
+        subfolder="text_encoder",
+        revision=revision,
+    )
+    model_class = text_encoder_config.architectures[0]
+
+    if model_class == "CLIPTextModel":
+        from transformers import CLIPTextModel
+
+        return CLIPTextModel
+    elif model_class == "RobertaSeriesModelWithTransformation":
+        from diffusers.pipelines.alt_diffusion.modeling_roberta_series import RobertaSeriesModelWithTransformation
+
+        return RobertaSeriesModelWithTransformation
+    else:
+        raise ValueError(f"{model_class} is not supported.")
+
+
+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(
+        "--instance_data_dir",
+        type=str,
+        default=None,
+        required=True,
+        help="A folder containing the training data of instance images.",
+    )
+    parser.add_argument(
+        "--class_data_dir",
+        type=str,
+        default=None,
+        required=False,
+        help="A folder containing the training data of class images.",
+    )
+    parser.add_argument(
+        "--instance_prompt",
+        type=str,
+        default=None,
+        required=True,
+        help="The prompt with identifier specifying the instance",
+    )
+    parser.add_argument(
+        "--class_prompt",
+        type=str,
+        default=None,
+        help="The prompt to specify images in the same class as provided instance images.",
+    )
+    parser.add_argument(
+        "--with_prior_preservation",
+        default=False,
+        action="store_true",
+        help="Flag to add prior preservation loss.",
+    )
+    parser.add_argument("--prior_loss_weight", type=float, default=1.0, help="The weight of prior preservation loss.")
+    parser.add_argument(
+        "--num_class_images",
+        type=int,
+        default=100,
+        help=(
+            "Minimal class images for prior preservation loss. If there are not enough images already present in"
+            " class_data_dir, additional images will be sampled with class_prompt."
+        ),
+    )
+    parser.add_argument(
+        "--output_dir",
+        type=str,
+        default="text-inversion-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(
+        "--center_crop", action="store_true", help="Whether to center crop images before resizing to resolution"
+    )
+    parser.add_argument("--train_text_encoder", action="store_true", help="Whether to train the text encoder")
+
+    # lora args
+    parser.add_argument("--use_lora", action="store_true", help="Whether to use Lora for parameter efficient tuning")
+    parser.add_argument("--lora_r", type=int, default=8, help="Lora rank, only used if use_lora is True")
+    parser.add_argument("--lora_alpha", type=int, default=32, help="Lora alpha, only used if use_lora is True")
+    parser.add_argument("--lora_dropout", type=float, default=0.0, help="Lora dropout, only used if use_lora is True")
+    parser.add_argument(
+        "--lora_bias",
+        type=str,
+        default="none",
+        help="Bias type for Lora. Can be 'none', 'all' or 'lora_only', only used if use_lora is True",
+    )
+    parser.add_argument(
+        "--lora_text_encoder_r",
+        type=int,
+        default=8,
+        help="Lora rank for text encoder, only used if `use_lora` and `train_text_encoder` are True",
+    )
+    parser.add_argument(
+        "--lora_text_encoder_alpha",
+        type=int,
+        default=32,
+        help="Lora alpha for text encoder, only used if `use_lora` and `train_text_encoder` are True",
+    )
+    parser.add_argument(
+        "--lora_text_encoder_dropout",
+        type=float,
+        default=0.0,
+        help="Lora dropout for text encoder, only used if `use_lora` and `train_text_encoder` are True",
+    )
+    parser.add_argument(
+        "--lora_text_encoder_bias",
+        type=str,
+        default="none",
+        help="Bias type for Lora. Can be 'none', 'all' or 'lora_only', only used if use_lora and `train_text_encoder` are True",
+    )
+
+    parser.add_argument(
+        "--train_batch_size", type=int, default=4, help="Batch size (per device) for the training dataloader."
+    )
+    parser.add_argument(
+        "--sample_batch_size", type=int, default=4, help="Batch size (per device) for sampling images."
+    )
+    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(
+        "--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(
+        "--learning_rate",
+        type=float,
+        default=5e-6,
+        help="Initial learning rate (after the potential warmup period) to use.",
+    )
+    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(
+        "--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(
+        "--prior_generation_precision",
+        type=str,
+        default=None,
+        choices=["no", "fp32", "fp16", "bf16"],
+        help=(
+            "Choose prior generation precision between fp32, fp16 and bf16 (bfloat16). Bf16 requires PyTorch >="
+            " 1.10.and an Nvidia Ampere GPU.  Default to  fp16 if a GPU is available else fp32."
+        ),
+    )
+    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."
+    )
+
+    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
+
+    if args.with_prior_preservation:
+        if args.class_data_dir is None:
+            raise ValueError("You must specify a data directory for class images.")
+        if args.class_prompt is None:
+            raise ValueError("You must specify prompt for class images.")
+    else:
+        # logger is not available yet
+        if args.class_data_dir is not None:
+            warnings.warn("You need not use --class_data_dir without --with_prior_preservation.")
+        if args.class_prompt is not None:
+            warnings.warn("You need not use --class_prompt without --with_prior_preservation.")
+
+    return args
+
+
+# Converting Bytes to Megabytes
+def b2mb(x):
+    return int(x / 2**20)
+
+
+# This context manager is used to track the peak memory usage of the process
+class TorchTracemalloc:
+    def __enter__(self):
+        gc.collect()
+        torch.cuda.empty_cache()
+        torch.cuda.reset_max_memory_allocated()  # reset the peak gauge to zero
+        self.begin = torch.cuda.memory_allocated()
+        self.process = psutil.Process()
+
+        self.cpu_begin = self.cpu_mem_used()
+        self.peak_monitoring = True
+        peak_monitor_thread = threading.Thread(target=self.peak_monitor_func)
+        peak_monitor_thread.daemon = True
+        peak_monitor_thread.start()
+        return self
+
+    def cpu_mem_used(self):
+        """get resident set size memory for the current process"""
+        return self.process.memory_info().rss
+
+    def peak_monitor_func(self):
+        self.cpu_peak = -1
+
+        while True:
+            self.cpu_peak = max(self.cpu_mem_used(), self.cpu_peak)
+
+            # can't sleep or will not catch the peak right (this comment is here on purpose)
+            # time.sleep(0.001) # 1msec
+
+            if not self.peak_monitoring:
+                break
+
+    def __exit__(self, *exc):
+        self.peak_monitoring = False
+
+        gc.collect()
+        torch.cuda.empty_cache()
+        self.end = torch.cuda.memory_allocated()
+        self.peak = torch.cuda.max_memory_allocated()
+        self.used = b2mb(self.end - self.begin)
+        self.peaked = b2mb(self.peak - self.begin)
+
+        self.cpu_end = self.cpu_mem_used()
+        self.cpu_used = b2mb(self.cpu_end - self.cpu_begin)
+        self.cpu_peaked = b2mb(self.cpu_peak - self.cpu_begin)
+        # print(f"delta used/peak {self.used:4d}/{self.peaked:4d}")
+
+
+def print_trainable_parameters(model):
+    """
+    Prints the number of trainable parameters in the model.
+    """
+    trainable_params = 0
+    all_param = 0
+    for _, param in model.named_parameters():
+        all_param += param.numel()
+        if param.requires_grad:
+            trainable_params += param.numel()
+    print(
+        f"trainable params: {trainable_params} || all params: {all_param} || trainable%: {100 * trainable_params / all_param}"
+    )
+
+
+class DreamBoothDataset(Dataset):
+    """
+    A dataset to prepare the instance and class images with the prompts for fine-tuning the model.
+    It pre-processes the images and the tokenizes prompts.
+    """
+
+    def __init__(
+        self,
+        instance_data_root,
+        instance_prompt,
+        tokenizer,
+        class_data_root=None,
+        class_prompt=None,
+        size=512,
+        center_crop=False,
+    ):
+        self.size = size
+        self.center_crop = center_crop
+        self.tokenizer = tokenizer
+
+        self.instance_data_root = Path(instance_data_root)
+        if not self.instance_data_root.exists():
+            raise ValueError("Instance images root doesn't exists.")
+
+        self.instance_images_path = list(Path(instance_data_root).iterdir())
+        self.num_instance_images = len(self.instance_images_path)
+        self.instance_prompt = instance_prompt
+        self._length = self.num_instance_images
+
+        if class_data_root is not None:
+            self.class_data_root = Path(class_data_root)
+            self.class_data_root.mkdir(parents=True, exist_ok=True)
+            self.class_images_path = list(self.class_data_root.iterdir())
+            self.num_class_images = len(self.class_images_path)
+            self._length = max(self.num_class_images, self.num_instance_images)
+            self.class_prompt = class_prompt
+        else:
+            self.class_data_root = None
+
+        self.image_transforms = transforms.Compose(
+            [
+                transforms.Resize(size, interpolation=transforms.InterpolationMode.BILINEAR),
+                transforms.CenterCrop(size) if center_crop else transforms.RandomCrop(size),
+                transforms.ToTensor(),
+                transforms.Normalize([0.5], [0.5]),
+            ]
+        )
+
+    def __len__(self):
+        return self._length
+
+    def __getitem__(self, index):
+        example = {}
+        instance_image = Image.open(self.instance_images_path[index % self.num_instance_images])
+        if not instance_image.mode == "RGB":
+            instance_image = instance_image.convert("RGB")
+        example["instance_images"] = self.image_transforms(instance_image)
+        example["instance_prompt_ids"] = self.tokenizer(
+            self.instance_prompt,
+            truncation=True,
+            padding="max_length",
+            max_length=self.tokenizer.model_max_length,
+            return_tensors="pt",
+        ).input_ids
+
+        if self.class_data_root:
+            class_image = Image.open(self.class_images_path[index % self.num_class_images])
+            if not class_image.mode == "RGB":
+                class_image = class_image.convert("RGB")
+            example["class_images"] = self.image_transforms(class_image)
+            example["class_prompt_ids"] = self.tokenizer(
+                self.class_prompt,
+                truncation=True,
+                padding="max_length",
+                max_length=self.tokenizer.model_max_length,
+                return_tensors="pt",
+            ).input_ids
+
+        return example
+
+
+def collate_fn(examples, with_prior_preservation=False):
+    input_ids = [example["instance_prompt_ids"] for example in examples]
+    pixel_values = [example["instance_images"] for example in examples]
+
+    # Concat class and instance examples for prior preservation.
+    # We do this to avoid doing two forward passes.
+    if with_prior_preservation:
+        input_ids += [example["class_prompt_ids"] for example in examples]
+        pixel_values += [example["class_images"] for example in examples]
+
+    pixel_values = torch.stack(pixel_values)
+    pixel_values = pixel_values.to(memory_format=torch.contiguous_format).float()
+
+    input_ids = torch.cat(input_ids, dim=0)
+
+    batch = {
+        "input_ids": input_ids,
+        "pixel_values": pixel_values,
+    }
+    return batch
+
+
+class PromptDataset(Dataset):
+    "A simple dataset to prepare the prompts to generate class images on multiple GPUs."
+
+    def __init__(self, prompt, num_samples):
+        self.prompt = prompt
+        self.num_samples = num_samples
+
+    def __len__(self):
+        return self.num_samples
+
+    def __getitem__(self, index):
+        example = {}
+        example["prompt"] = self.prompt
+        example["index"] = index
+        return example
+
+
+def get_full_repo_name(model_id: str, organization: Optional[str] = None, token: Optional[str] = None):
+    if token is None:
+        token = HfFolder.get_token()
+    if organization is None:
+        username = whoami(token)["name"]
+        return f"{username}/{model_id}"
+    else:
+        return f"{organization}/{model_id}"
+
+
+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,
+        logging_dir=logging_dir,
+    )
+
+    # Currently, it's not possible to do gradient accumulation when training two models with accelerate.accumulate
+    # This will be enabled soon in accelerate. For now, we don't allow gradient accumulation when training two models.
+    # TODO (patil-suraj): Remove this check when gradient accumulation with two models is enabled in accelerate.
+    if args.train_text_encoder and args.gradient_accumulation_steps > 1 and accelerator.num_processes > 1:
+        raise ValueError(
+            "Gradient accumulation is not supported when training the text encoder in distributed training. "
+            "Please set gradient_accumulation_steps to 1. This feature will be supported in the future."
+        )
+
+    # 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:
+        datasets.utils.logging.set_verbosity_warning()
+        transformers.utils.logging.set_verbosity_warning()
+        diffusers.utils.logging.set_verbosity_info()
+    else:
+        datasets.utils.logging.set_verbosity_error()
+        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)
+
+    # Generate class images if prior preservation is enabled.
+    if args.with_prior_preservation:
+        class_images_dir = Path(args.class_data_dir)
+        if not class_images_dir.exists():
+            class_images_dir.mkdir(parents=True)
+        cur_class_images = len(list(class_images_dir.iterdir()))
+
+        if cur_class_images < args.num_class_images:
+            torch_dtype = torch.float16 if accelerator.device.type == "cuda" else torch.float32
+            if args.prior_generation_precision == "fp32":
+                torch_dtype = torch.float32
+            elif args.prior_generation_precision == "fp16":
+                torch_dtype = torch.float16
+            elif args.prior_generation_precision == "bf16":
+                torch_dtype = torch.bfloat16
+            pipeline = DiffusionPipeline.from_pretrained(
+                args.pretrained_model_name_or_path,
+                torch_dtype=torch_dtype,
+                safety_checker=None,
+                revision=args.revision,
+            )
+            pipeline.set_progress_bar_config(disable=True)
+
+            num_new_images = args.num_class_images - cur_class_images
+            logger.info(f"Number of class images to sample: {num_new_images}.")
+
+            sample_dataset = PromptDataset(args.class_prompt, num_new_images)
+            sample_dataloader = torch.utils.data.DataLoader(sample_dataset, batch_size=args.sample_batch_size)
+
+            sample_dataloader = accelerator.prepare(sample_dataloader)
+            pipeline.to(accelerator.device)
+
+            for example in tqdm(
+                sample_dataloader, desc="Generating class images", disable=not accelerator.is_local_main_process
+            ):
+                images = pipeline(example["prompt"]).images
+
+                for i, image in enumerate(images):
+                    hash_image = hashlib.sha1(image.tobytes()).hexdigest()
+                    image_filename = class_images_dir / f"{example['index'][i] + cur_class_images}-{hash_image}.jpg"
+                    image.save(image_filename)
+
+            del pipeline
+            if torch.cuda.is_available():
+                torch.cuda.empty_cache()
+
+    # Handle the repository creation
+    if accelerator.is_main_process:
+        if args.push_to_hub:
+            if args.hub_model_id is None:
+                repo_name = get_full_repo_name(Path(args.output_dir).name, token=args.hub_token)
+            else:
+                repo_name = args.hub_model_id
+            repo = Repository(args.output_dir, clone_from=repo_name)  # noqa: F841
+
+            with open(os.path.join(args.output_dir, ".gitignore"), "w+") as gitignore:
+                if "step_*" not in gitignore:
+                    gitignore.write("step_*\n")
+                if "epoch_*" not in gitignore:
+                    gitignore.write("epoch_*\n")
+        elif args.output_dir is not None:
+            os.makedirs(args.output_dir, exist_ok=True)
+
+    # 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,
+        )
+
+    # import correct text encoder class
+    text_encoder_cls = import_model_class_from_model_name_or_path(args.pretrained_model_name_or_path, args.revision)
+
+    # Load scheduler and models
+    noise_scheduler = DDPMScheduler(
+        beta_start=0.00085,
+        beta_end=0.012,
+        beta_schedule="scaled_linear",
+        num_train_timesteps=1000,
+    )  # DDPMScheduler.from_pretrained(args.pretrained_model_name_or_path, subfolder="scheduler")
+    text_encoder = text_encoder_cls.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
+    )
+
+    if args.use_lora:
+        config = LoraConfig(
+            r=args.lora_r,
+            lora_alpha=args.lora_alpha,
+            target_modules=UNET_TARGET_MODULES,
+            lora_dropout=args.lora_dropout,
+            bias=args.lora_bias,
+        )
+        unet = LoraModel(config, unet)
+        print_trainable_parameters(unet)
+        print(unet)
+
+    vae.requires_grad_(False)
+    if not args.train_text_encoder:
+        text_encoder.requires_grad_(False)
+    elif args.train_text_encoder and args.use_lora:
+        config = LoraConfig(
+            r=args.lora_text_encoder_r,
+            lora_alpha=args.lora_text_encoder_alpha,
+            target_modules=TEXT_ENCODER_TARGET_MODULES,
+            lora_dropout=args.lora_text_encoder_dropout,
+            bias=args.lora_text_encoder_bias,
+        )
+        text_encoder = LoraModel(config, text_encoder)
+        print_trainable_parameters(text_encoder)
+        print(text_encoder)
+
+    if args.enable_xformers_memory_efficient_attention:
+        if is_xformers_available():
+            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()
+        # below fails when using lora so commenting it out
+        if args.train_text_encoder and not args.use_lora:
+            text_encoder.gradient_checkpointing_enable()
+
+    # 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.learning_rate = (
+            args.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
+    params_to_optimize = (
+        itertools.chain(unet.parameters(), text_encoder.parameters()) if args.train_text_encoder else unet.parameters()
+    )
+    optimizer = optimizer_class(
+        params_to_optimize,
+        lr=args.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 = DreamBoothDataset(
+        instance_data_root=args.instance_data_dir,
+        instance_prompt=args.instance_prompt,
+        class_data_root=args.class_data_dir if args.with_prior_preservation else None,
+        class_prompt=args.class_prompt,
+        tokenizer=tokenizer,
+        size=args.resolution,
+        center_crop=args.center_crop,
+    )
+
+    train_dataloader = torch.utils.data.DataLoader(
+        train_dataset,
+        batch_size=args.train_batch_size,
+        shuffle=True,
+        collate_fn=lambda examples: collate_fn(examples, args.with_prior_preservation),
+        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`.
+    if args.train_text_encoder:
+        unet, text_encoder, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
+            unet, text_encoder, optimizer, train_dataloader, lr_scheduler
+        )
+    else:
+        unet, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
+            unet, optimizer, train_dataloader, lr_scheduler
+        )
+
+    # For mixed precision training we cast the text_encoder and vae weights to half-precision
+    # as these models 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 and text_encoder to device and cast to weight_dtype
+    vae.to(accelerator.device, dtype=weight_dtype)
+    if not args.train_text_encoder:
+        text_encoder.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:
+        accelerator.init_trackers("dreambooth", config=vars(args))
+
+    # 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]
+        accelerator.print(f"Resuming from checkpoint {path}")
+        accelerator.load_state(os.path.join(args.output_dir, path))
+        global_step = int(path.split("-")[1])
+
+        resume_global_step = global_step * args.gradient_accumulation_steps
+        first_epoch = resume_global_step // num_update_steps_per_epoch
+        resume_step = resume_global_step % num_update_steps_per_epoch
+
+    # Only show the progress bar once on each machine.
+    progress_bar = tqdm(range(global_step, args.max_train_steps), disable=not accelerator.is_local_main_process)
+    progress_bar.set_description("Steps")
+
+    for epoch in range(first_epoch, args.num_train_epochs):
+        unet.train()
+        if args.train_text_encoder:
+            text_encoder.train()
+        with TorchTracemalloc() as tracemalloc:
+            for step, batch in enumerate(train_dataloader):
+                # Skip steps until we reach the resumed step
+                if args.resume_from_checkpoint and epoch == first_epoch and step < resume_step:
+                    if step % args.gradient_accumulation_steps == 0:
+                        progress_bar.update(1)
+                    continue
+
+                with accelerator.accumulate(unet):
+                    # Convert images to latent space
+                    latents = vae.encode(batch["pixel_values"].to(dtype=weight_dtype)).latent_dist.sample()
+                    latents = latents * 0.18215
+
+                    # 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)
+
+                    # Get the text embedding for conditioning
+                    encoder_hidden_states = text_encoder(batch["input_ids"])[0]
+
+                    # Predict the noise residual
+                    model_pred = unet(noisy_latents, timesteps, encoder_hidden_states).sample
+
+                    # Get the target for loss depending on the prediction type
+                    if noise_scheduler.config.prediction_type == "epsilon":
+                        target = noise
+                    elif noise_scheduler.config.prediction_type == "v_prediction":
+                        target = noise_scheduler.get_velocity(latents, noise, timesteps)
+                    else:
+                        raise ValueError(f"Unknown prediction type {noise_scheduler.config.prediction_type}")
+
+                    if args.with_prior_preservation:
+                        # Chunk the noise and model_pred into two parts and compute the loss on each part separately.
+                        model_pred, model_pred_prior = torch.chunk(model_pred, 2, dim=0)
+                        target, target_prior = torch.chunk(target, 2, dim=0)
+
+                        # Compute instance loss
+                        loss = F.mse_loss(model_pred.float(), target.float(), reduction="mean")
+
+                        # Compute prior loss
+                        prior_loss = F.mse_loss(model_pred_prior.float(), target_prior.float(), reduction="mean")
+
+                        # Add the prior loss to the instance loss.
+                        loss = loss + args.prior_loss_weight * prior_loss
+                    else:
+                        loss = F.mse_loss(model_pred.float(), target.float(), reduction="mean")
+
+                    accelerator.backward(loss)
+                    if accelerator.sync_gradients:
+                        params_to_clip = (
+                            itertools.chain(unet.parameters(), text_encoder.parameters())
+                            if args.train_text_encoder
+                            else unet.parameters()
+                        )
+                        accelerator.clip_grad_norm_(params_to_clip, args.max_grad_norm)
+                    optimizer.step()
+                    lr_scheduler.step()
+                    optimizer.zero_grad()
+
+                # Checks if the accelerator has performed an optimization step behind the scenes
+                if accelerator.sync_gradients:
+                    progress_bar.update(1)
+                    global_step += 1
+
+                    # if global_step % args.checkpointing_steps == 0:
+                    #     if accelerator.is_main_process:
+                    #         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}")
+
+                logs = {"loss": loss.detach().item(), "lr": lr_scheduler.get_last_lr()[0]}
+                progress_bar.set_postfix(**logs)
+                accelerator.log(logs, step=global_step)
+
+                if global_step >= args.max_train_steps:
+                    break
+        # Printing the GPU memory usage details such as allocated memory, peak memory, and total memory usage
+        accelerator.print("GPU Memory before entering the train : {}".format(b2mb(tracemalloc.begin)))
+        accelerator.print("GPU Memory consumed at the end of the train (end-begin): {}".format(tracemalloc.used))
+        accelerator.print("GPU Peak Memory consumed during the train (max-begin): {}".format(tracemalloc.peaked))
+        accelerator.print(
+            "GPU Total Peak Memory consumed during the train (max): {}".format(
+                tracemalloc.peaked + b2mb(tracemalloc.begin)
+            )
+        )
+
+        accelerator.print("CPU Memory before entering the train : {}".format(b2mb(tracemalloc.cpu_begin)))
+        accelerator.print("CPU Memory consumed at the end of the train (end-begin): {}".format(tracemalloc.cpu_used))
+        accelerator.print("CPU Peak Memory consumed during the train (max-begin): {}".format(tracemalloc.cpu_peaked))
+        accelerator.print(
+            "CPU Total Peak Memory consumed during the train (max): {}".format(
+                tracemalloc.cpu_peaked + b2mb(tracemalloc.cpu_begin)
+            )
+        )
+
+    # Create the pipeline using using the trained modules and save it.
+    accelerator.wait_for_everyone()
+    if accelerator.is_main_process:
+        if args.use_lora:
+            lora_config = {}
+            state_dict = get_peft_model_state_dict(unet, state_dict=accelerator.get_state_dict(unet))
+            lora_config["peft_config"] = unet.get_peft_config_as_dict(inference=True)
+            if args.train_text_encoder:
+                text_encoder_state_dict = get_peft_model_state_dict(
+                    text_encoder, state_dict=accelerator.get_state_dict(text_encoder)
+                )
+                text_encoder_state_dict = {f"text_encoder_{k}": v for k, v in text_encoder_state_dict.items()}
+                state_dict.update(text_encoder_state_dict)
+                lora_config["text_encoder_peft_config"] = text_encoder.get_peft_config_as_dict(inference=True)
+
+            accelerator.print(state_dict)
+            accelerator.save(state_dict, os.path.join(args.output_dir, f"{args.instance_prompt}_lora.pt"))
+            with open(os.path.join(args.output_dir, f"{args.instance_prompt}_lora_config.json"), "w") as f:
+                json.dump(lora_config, f)
+        else:
+            pipeline = DiffusionPipeline.from_pretrained(
+                args.pretrained_model_name_or_path,
+                unet=accelerator.unwrap_model(unet),
+                text_encoder=accelerator.unwrap_model(text_encoder),
+                revision=args.revision,
+            )
+            pipeline.save_pretrained(args.output_dir)
+
+    accelerator.end_training()
+
+
+if __name__ == "__main__":
+    args = parse_args()
+    main(args)

trainer.py

@@ -0,0 +1,156 @@
+from __future__ import annotations
+
+import os
+import pathlib
+import shlex
+import shutil
+import subprocess
+
+import gradio as gr
+import PIL.Image
+import torch
+
+
+def pad_image(image: PIL.Image.Image) -> PIL.Image.Image:
+    w, h = image.size
+    if w == h:
+        return image
+    elif w > h:
+        new_image = PIL.Image.new(image.mode, (w, w), (0, 0, 0))
+        new_image.paste(image, (0, (w - h) // 2))
+        return new_image
+    else:
+        new_image = PIL.Image.new(image.mode, (h, h), (0, 0, 0))
+        new_image.paste(image, ((h - w) // 2, 0))
+        return new_image
+
+
+class Trainer:
+    def __init__(self):
+        self.is_running = False
+        self.is_running_message = "Another training is in progress."
+
+        self.output_dir = pathlib.Path("results")
+        self.instance_data_dir = self.output_dir / "training_data"
+
+    def check_if_running(self) -> dict:
+        if self.is_running:
+            return gr.update(value=self.is_running_message)
+        else:
+            return gr.update(value="No training is running.")
+
+    def cleanup_dirs(self) -> None:
+        shutil.rmtree(self.output_dir, ignore_errors=True)
+
+    def prepare_dataset(self, concept_images: list, resolution: int) -> None:
+        self.instance_data_dir.mkdir(parents=True)
+        for i, temp_path in enumerate(concept_images):
+            image = PIL.Image.open(temp_path.name)
+            image = pad_image(image)
+            image = image.resize((resolution, resolution))
+            image = image.convert("RGB")
+            out_path = self.instance_data_dir / f"{i:03d}.jpg"
+            image.save(out_path, format="JPEG", quality=100)
+
+    def run(
+        self,
+        base_model: str,
+        resolution_s: str,
+        n_steps: int,
+        concept_images: list | None,
+        concept_prompt: str,
+        learning_rate: float,
+        gradient_accumulation: int,
+        fp16: bool,
+        use_8bit_adam: bool,
+        gradient_checkpointing: bool,
+        train_text_encoder: bool,
+        with_prior_preservation: bool,
+        prior_loss_weight: float,
+        class_prompt: str,
+        num_class_images: int,
+        lora_r: int,
+        lora_alpha: int,
+        lora_bias: str,
+        lora_dropout: float,
+        lora_text_encoder_r: int,
+        lora_text_encoder_alpha: int,
+        lora_text_encoder_bias: str,
+        lora_text_encoder_dropout: float,
+    ) -> tuple[dict, list[pathlib.Path]]:
+        if not torch.cuda.is_available():
+            raise gr.Error("CUDA is not available.")
+
+        if self.is_running:
+            return gr.update(value=self.is_running_message), []
+
+        if concept_images is None:
+            raise gr.Error("You need to upload images.")
+        if not concept_prompt:
+            raise gr.Error("The concept prompt is missing.")
+
+        resolution = int(resolution_s)
+
+        self.cleanup_dirs()
+        self.prepare_dataset(concept_images, resolution)
+
+        command = f"""
+        accelerate launch train_dreambooth.py \
+            --pretrained_model_name_or_path={base_model}  \
+            --instance_data_dir={self.instance_data_dir} \
+            --output_dir={self.output_dir} \
+            --train_text_encoder \
+            --instance_prompt="{concept_prompt}" \
+            --resolution={resolution} \
+            --gradient_accumulation_steps={gradient_accumulation} \
+            --learning_rate={learning_rate} \
+            --max_train_steps={n_steps} \
+            --train_batch_size=1 \
+            --lr_scheduler=constant \
+            --lr_warmup_steps=0 \
+            --num_class_images={num_class_images} \
+        """
+        if train_text_encoder:
+            command += f" --train_text_encoder"
+        if with_prior_preservation:
+            command += f""" --with_prior_preservation \
+                --prior_loss_weight={prior_loss_weight} \
+                --class_prompt="{class_prompt}" \
+                --class_data_dir={self.output_dir / 'class_data'}
+                """
+
+        command += f""" --use_lora \
+            --lora_r={lora_r} \
+            --lora_alpha={lora_alpha} \
+            --lora_bias={lora_bias} \
+            --lora_dropout={lora_dropout}
+            """
+
+        if train_text_encoder:
+            command += f""" --lora_text_encoder_r={lora_text_encoder_r} \
+                --lora_text_encoder_alpha={lora_text_encoder_alpha} \
+                --lora_text_encoder_bias={lora_text_encoder_bias} \
+                --lora_text_encoder_dropout={lora_text_encoder_dropout}
+                """
+        if fp16:
+            command += " --mixed_precision fp16"
+        if use_8bit_adam:
+            command += " --use_8bit_adam"
+        if gradient_checkpointing:
+            command += " --gradient_checkpointing"
+
+        with open(self.output_dir / "train.sh", "w") as f:
+            command_s = " ".join(command.split())
+            f.write(command_s)
+
+        self.is_running = True
+        res = subprocess.run(shlex.split(command))
+        self.is_running = False
+
+        if res.returncode == 0:
+            result_message = "Training Completed!"
+        else:
+            result_message = "Training Failed!"
+        weight_paths = sorted(self.output_dir.glob("*.pt"))
+        config_paths = sorted(self.output_dir.glob("*.json"))
+        return gr.update(value=result_message), weight_paths + config_paths

uploader.py

@@ -0,0 +1,17 @@
+import gradio as gr
+from huggingface_hub import HfApi
+
+
+def upload(model_name: str, hf_token: str) -> None:
+    api = HfApi(token=hf_token)
+    user_name = api.whoami()["name"]
+    model_id = f"{user_name}/{model_name}"
+    try:
+        api.create_repo(model_id, repo_type="model", private=True)
+        api.upload_folder(repo_id=model_id, folder_path="results", path_in_repo="results", repo_type="model")
+        url = f"https://huggingface.co/{model_id}"
+        message = f"Your model was successfully uploaded to [{url}]({url})."
+    except Exception as e:
+        message = str(e)
+
+    return gr.update(value=message, visible=True)