{ "metadata": { "colab": { "collapsed_sections": [], "provenance": [] }, "kernelspec": { "name": "python", "display_name": "Python (Pyodide)", "language": "python" }, "language_info": { "codemirror_mode": { "name": "python", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.8" }, "vscode": { "interpreter": { "hash": "b0fa6594d8f4cbf19f97940f81e996739fb7646882a419484c72d19e05852a7e" } } }, "nbformat_minor": 5, "nbformat": 4, "cells": [ { "cell_type": "markdown", "source": "# Dreambooth\n### Notebook implementation by Joe Penna (@MysteryGuitarM on Twitter) - Improvements by David Bielejeski\nhttps://github.com/JoePenna/Dreambooth-Stable-Diffusion\n\n### If on runpod / vast.ai / etc, spin up an A6000 or A100 pod using a Stable Diffusion template with Jupyter pre-installed.", "metadata": { "id": "aa2c1ada" }, "id": "aa2c1ada" }, { "cell_type": "markdown", "source": "## Build Environment", "metadata": { "id": "7b971cc0" }, "id": "7b971cc0" }, { "cell_type": "code", "source": "#BUILD ENV\n!pip install omegaconf\n!pip install einops\n!pip install pytorch-lightning==1.6.5\n!pip install test-tube\n!pip install transformers\n!pip install kornia\n!pip install -e git+https://github.com/CompVis/taming-transformers.git@master#egg=taming-transformers\n!pip install -e git+https://github.com/openai/CLIP.git@main#egg=clip\n!pip install setuptools==59.5.0\n!pip install pillow==9.0.1\n!pip install torchmetrics==0.6.0\n!pip install -e .\n!pip install protobuf==3.20.1\n!pip install gdown\n!pip install pydrive\n!pip install -qq diffusers[\"training\"]==0.3.0 transformers ftfy\n!pip install -qq \"ipywidgets>=7,<8\"\n!pip install huggingface_hub\n!pip install ipywidgets", "metadata": { "id": "9e1bc458-091b-42f4-a125-c3f0df20f29d", "scrolled": true }, "execution_count": null, "outputs": [], "id": "9e1bc458-091b-42f4-a125-c3f0df20f29d" }, { "cell_type": "code", "source": "## Move the sd-v1-4.ckpt to the root of this directory as \"model.ckpt\"\n#actual_locations_of_model_blob = !readlink -f {downloaded_model_path}\n#!cp {actual_locations_of_model_blob[-1]} model.ckpt\n!apt-get update ; apt-get install wget\n!wget 'https://prodesk.home.thijn.ovh/sd-v1-4.ckpt'\n!cp sd-v1-4.ckpt model.ckpt", "metadata": { "collapsed": false, "jupyter": { "outputs_hidden": false } }, "execution_count": null, "outputs": [], "id": "ddf7a43d" }, { "cell_type": "markdown", "source": "## Download pre-generated regularization images\n\nWe've created the following image sets\n\n* man_euler - provided by Niko Pueringer (Corridor Digital) - euler @ 40 steps, CFG 7.5\n* man_unsplash - pictures from various photographers\n* person_ddim\n* woman_ddim - provided by David Bielejeski - ddim @ 50 steps, CFG 10.0\n\n`person_ddim` is recommended", "metadata": { "id": "mxPL2O0OLvBW" }, "id": "mxPL2O0OLvBW" }, { "cell_type": "code", "source": "# Grab the existing regularization images\n# Choose the dataset that best represents what you are trying to do and matches what you used for your token\n# man_euler, man_unsplash, person_ddim, woman_ddim\ndataset=\"man_euler\"\n!rm -rf Stable-Diffusion-Regularization-Images-{dataset}\n!git clone https://github.com/djbielejeski/Stable-Diffusion-Regularization-Images-{dataset}.git\n\n!mkdir -p outputs/txt2img-samples/samples/{dataset}\n!mv -v Stable-Diffusion-Regularization-Images-{dataset}/{dataset}/*.* outputs/txt2img-samples/samples/{dataset}", "metadata": { "id": "e7EydXCjOV1v" }, "execution_count": null, "outputs": [], "id": "e7EydXCjOV1v" }, { "cell_type": "markdown", "source": "# Upload your training images\nUpload 10-20 images of someone to\n\n```\n/workspace/Dreambooth-Stable-Diffusion/training_samples\n```\n\nWARNING: Be sure to upload an *even* amount of images, otherwise the training inexplicably stops at 1500 steps.\n\n* 2-3 full body\n* 3-5 upper body \n* 5-12 close-up on face", "metadata": { "id": "zshrC_JuMXmM" }, "id": "zshrC_JuMXmM" }, { "cell_type": "code", "source": "#@markdown Add here the URLs to the images of the concept you are adding\nurls = [\n\"https://prodesk.home.thijn.ovh/gijs/IMG_20220926_121625.png\",\n\"https://prodesk.home.thijn.ovh/gijs/IMG_20220926_121630.png\",\n\"https://prodesk.home.thijn.ovh/gijs/IMG_20220926_121632.png\",\n\"https://prodesk.home.thijn.ovh/gijs/IMG_20220926_121636.png\",\n\"https://prodesk.home.thijn.ovh/gijs/IMG_20220926_121637.png\",\n\"https://prodesk.home.thijn.ovh/gijs/IMG_20220926_121640.png\",\n\"https://prodesk.home.thijn.ovh/gijs/IMG_20220926_121642.png\",\n\"https://prodesk.home.thijn.ovh/gijs/IMG_20220926_121644.png\",\n\"https://prodesk.home.thijn.ovh/gijs/IMG_20220926_121647.png\",\n\"https://prodesk.home.thijn.ovh/gijs/IMG_20220926_121649.png\",\n\"https://prodesk.home.thijn.ovh/gijs/IMG_20220926_121653.png\",\n\"https://prodesk.home.thijn.ovh/gijs/IMG_20220926_121656.png\",\n\"https://prodesk.home.thijn.ovh/gijs/IMG_20220926_121659.png\",\n\"https://prodesk.home.thijn.ovh/gijs/IMG_20220926_121702.png\"\n ## You can add additional images here\n]", "metadata": { "collapsed": false, "jupyter": { "outputs_hidden": false } }, "execution_count": null, "outputs": [], "id": "60e37ee0" }, { "cell_type": "code", "source": "#@title Download and check the images you have just added\nimport os\nimport requests\nfrom io import BytesIO\nfrom PIL import Image\n\n\ndef image_grid(imgs, rows, cols):\n assert len(imgs) == rows*cols\n\n w, h = imgs[0].size\n grid = Image.new('RGB', size=(cols*w, rows*h))\n grid_w, grid_h = grid.size\n\n for i, img in enumerate(imgs):\n grid.paste(img, box=(i%cols*w, i//cols*h))\n return grid\n\ndef download_image(url):\n try:\n response = requests.get(url)\n except:\n return None\n return Image.open(BytesIO(response.content)).convert(\"RGB\")\n\nimages = list(filter(None,[download_image(url) for url in urls]))\nsave_path = \"./training_samples\"\nif not os.path.exists(save_path):\n os.mkdir(save_path)\n[image.save(f\"{save_path}/{i}.png\", format=\"png\") for i, image in enumerate(images)]\nimage_grid(images, 1, len(images))", "metadata": { "collapsed": false, "jupyter": { "outputs_hidden": false } }, "execution_count": null, "outputs": [], "id": "df314e2e" }, { "cell_type": "markdown", "source": "## Training\n\nIf training a person or subject, keep an eye on your project's `logs/{folder}/images/train/samples_scaled_gs-00xxxx` generations.\n\nIf training a style, keep an eye on your project's `logs/{folder}/images/train/samples_gs-00xxxx` generations.", "metadata": { "id": "ad4e50df" }, "id": "ad4e50df" }, { "cell_type": "code", "source": "# START THE TRAINING\nproject_name = \"gijsbert\"\nbatch_size = 1000\nclass_word = \"man\" # << match this word to the class word from regularization images above\nreg_data_root = \"/workspace/Dreambooth-Stable-Diffusion/outputs/txt2img-samples/samples/\" + dataset\n\n!rm -rf training_samples/.ipynb_checkpoints\n!python \"main.py\" \\\n --base configs/stable-diffusion/v1-finetune_unfrozen.yaml \\\n -t \\\n --actual_resume \"model.ckpt\" \\\n --reg_data_root {reg_data_root} \\\n -n {project_name} \\\n --gpus 0, \\\n --data_root \"/workspace/Dreambooth-Stable-Diffusion/training_samples\" \\\n --batch_size {batch_size} \\\n --class_word class_word", "metadata": { "id": "6fa5dd66-2ca0-4819-907e-802e25583ae6", "tags": [] }, "execution_count": null, "outputs": [], "id": "6fa5dd66-2ca0-4819-907e-802e25583ae6" }, { "cell_type": "markdown", "source": "## Pruning (12GB to 2GB)\nWe are working on having this happen automatically (TODO: PR's welcome)", "metadata": {}, "id": "dc49d0bd" }, { "cell_type": "code", "source": "directory_paths = !ls -d logs/*", "metadata": { "collapsed": false, "jupyter": { "outputs_hidden": false } }, "execution_count": null, "outputs": [], "id": "27cea333" }, { "cell_type": "code", "source": "# This version should automatically prune around 10GB from the ckpt file\nlast_checkpoint_file = directory_paths[-1] + \"/checkpoints/last.ckpt\"\n!python \"prune_ckpt.py\" --ckpt {last_checkpoint_file}", "metadata": { "collapsed": false, "jupyter": { "outputs_hidden": false } }, "execution_count": null, "outputs": [], "id": "965b4654" }, { "cell_type": "code", "source": "last_checkpoint_file_pruned = directory_paths[-1] + \"/checkpoints/last-pruned.ckpt\"\ntraining_samples = !ls training_samples\ndate_string = !date +\"%Y-%m-%dT%H-%M-%S\"\nfile_name = date_string[-1] + \"_\" + project_name + \"_\" + str(len(training_samples)) + \"_training_images_\" + str(batch_size) + \"_batch_size_\" + class_word + \"_class_word.ckpt\"\n!mkdir -p trained_models\n!mv {last_checkpoint_file_pruned} trained_models/{file_name}", "metadata": { "collapsed": false, "jupyter": { "outputs_hidden": false } }, "execution_count": null, "outputs": [], "id": "b7a8cec3" }, { "cell_type": "code", "source": "# Download your trained model file from `trained_models` and use in your favorite Stable Diffusion repo!", "metadata": {}, "execution_count": null, "outputs": [], "id": "ff1a46d9" }, { "cell_type": "markdown", "source": "## Generate Images With Your Trained Model!", "metadata": {}, "id": "d28d0139" }, { "cell_type": "code", "source": "!echo python scripts/stable_txt2img.py \\\n --ddim_eta 0.0 \\\n --n_samples 1 \\\n --n_iter 4 \\\n --scale 7.0 \\\n --ddim_steps 50 \\\n --ckpt \"/workspace/Dreambooth-Stable-Diffusion/trained_models/\" + {file_name} \\\n --prompt \"gijsbert person as a masterpiece portrait painting by John Singer Sargent in the style of Rembrandt\"", "metadata": { "trusted": true }, "execution_count": 2, "outputs": [ { "ename": "", "evalue": "module 'pexpect' has no attribute 'TIMEOUT'", "traceback": [ "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m", "\u001b[0;31mAttributeError\u001b[0m Traceback (most recent call last)", "Cell \u001b[0;32mIn [2], line 1\u001b[0m\n\u001b[0;32m----> 1\u001b[0m get_ipython()\u001b[38;5;241m.\u001b[39msystem(\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mchangeme python scripts/stable_txt2img.py --ddim_eta 0.0 --n_samples 1 --n_iter 4 --scale 7.0 --ddim_steps 50 --ckpt \u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m/workspace/Dreambooth-Stable-Diffusion/trained_models/\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m + \u001b[39m\u001b[38;5;132;01m{file_name}\u001b[39;00m\u001b[38;5;124m --prompt \u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mjoepenna person as a masterpiece portrait painting by John Singer Sargent in the style of Rembrandt\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m'\u001b[39m)\n", "File \u001b[0;32m/lib/python3.10/site-packages/IPython/core/interactiveshell.py:2466\u001b[0m, in \u001b[0;36mInteractiveShell.system_piped\u001b[0;34m(self, cmd)\u001b[0m\n\u001b[1;32m 2461\u001b[0m \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mOSError\u001b[39;00m(\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mBackground processes not supported.\u001b[39m\u001b[38;5;124m\"\u001b[39m)\n\u001b[1;32m 2463\u001b[0m \u001b[38;5;66;03m# we explicitly do NOT return the subprocess status code, because\u001b[39;00m\n\u001b[1;32m 2464\u001b[0m \u001b[38;5;66;03m# a non-None value would trigger :func:`sys.displayhook` calls.\u001b[39;00m\n\u001b[1;32m 2465\u001b[0m \u001b[38;5;66;03m# Instead, we store the exit_code in user_ns.\u001b[39;00m\n\u001b[0;32m-> 2466\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39muser_ns[\u001b[38;5;124m'\u001b[39m\u001b[38;5;124m_exit_code\u001b[39m\u001b[38;5;124m'\u001b[39m] \u001b[38;5;241m=\u001b[39m \u001b[43msystem\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mvar_expand\u001b[49m\u001b[43m(\u001b[49m\u001b[43mcmd\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mdepth\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;241;43m1\u001b[39;49m\u001b[43m)\u001b[49m\u001b[43m)\u001b[49m\n", "File \u001b[0;32m/lib/python3.10/site-packages/IPython/utils/_process_posix.py:129\u001b[0m, in \u001b[0;36mProcessHandler.system\u001b[0;34m(self, cmd)\u001b[0m\n\u001b[1;32m 125\u001b[0m enc \u001b[38;5;241m=\u001b[39m DEFAULT_ENCODING\n\u001b[1;32m 127\u001b[0m \u001b[38;5;66;03m# Patterns to match on the output, for pexpect. We read input and\u001b[39;00m\n\u001b[1;32m 128\u001b[0m \u001b[38;5;66;03m# allow either a short timeout or EOF\u001b[39;00m\n\u001b[0;32m--> 129\u001b[0m patterns \u001b[38;5;241m=\u001b[39m [\u001b[43mpexpect\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mTIMEOUT\u001b[49m, pexpect\u001b[38;5;241m.\u001b[39mEOF]\n\u001b[1;32m 130\u001b[0m \u001b[38;5;66;03m# the index of the EOF pattern in the list.\u001b[39;00m\n\u001b[1;32m 131\u001b[0m \u001b[38;5;66;03m# even though we know it's 1, this call means we don't have to worry if\u001b[39;00m\n\u001b[1;32m 132\u001b[0m \u001b[38;5;66;03m# we change the above list, and forget to change this value:\u001b[39;00m\n\u001b[1;32m 133\u001b[0m EOF_index \u001b[38;5;241m=\u001b[39m patterns\u001b[38;5;241m.\u001b[39mindex(pexpect\u001b[38;5;241m.\u001b[39mEOF)\n", "\u001b[0;31mAttributeError\u001b[0m: module 'pexpect' has no attribute 'TIMEOUT'" ], "output_type": "error" } ], "id": "80ddb03b" }, { "cell_type": "code", "source": "", "metadata": {}, "execution_count": null, "outputs": [], "id": "0e3c10d9-2c40-4f50-9cf4-97e88a57288c" } ] }