Create sdxl.ipynb
Browse files- sdxl.ipynb +222 -0
sdxl.ipynb
ADDED
|
@@ -0,0 +1,222 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"cells": [
|
| 3 |
+
{
|
| 4 |
+
"cell_type": "markdown",
|
| 5 |
+
"metadata": {},
|
| 6 |
+
"source": [
|
| 7 |
+
"# SDXL with Custom LoRA on T4 GPU\n",
|
| 8 |
+
"\n",
|
| 9 |
+
"This notebook sets up Stable Diffusion XL (SDXL) on a T4 GPU in Google Colab with Python 3.11, downloads the base model from Hugging Face, and applies a custom LoRA model from Hugging Face or Civitai. It generates images using the configured pipeline.\n",
|
| 10 |
+
"\n",
|
| 11 |
+
"**Prerequisites:**\n",
|
| 12 |
+
"- Hugging Face account and token for gated model access (e.g., `stabilityai/stable-diffusion-xl-base-1.0`).\n",
|
| 13 |
+
"- Civitai API key if downloading LoRA from Civitai.\n",
|
| 14 |
+
"- Ensure Colab is set to T4 GPU (Runtime > Change runtime type > T4 GPU).\n",
|
| 15 |
+
"\n",
|
| 16 |
+
"**Note:** Replace placeholders (e.g., `YOUR_HF_TOKEN`, `YOUR_CIVITAI_API_KEY`) with your actual credentials."
|
| 17 |
+
]
|
| 18 |
+
},
|
| 19 |
+
{
|
| 20 |
+
"cell_type": "code",
|
| 21 |
+
"metadata": {},
|
| 22 |
+
"source": [
|
| 23 |
+
"# Install dependencies for Python 3.11\n",
|
| 24 |
+
"!pip install torch==2.2.0 torchvision==0.17.0 --index-url https://download.pytorch.org/whl/cu118\n",
|
| 25 |
+
"!pip install diffusers==0.29.2 transformers==4.44.2 accelerate==0.33.0 safetensors==0.4.5\n",
|
| 26 |
+
"!pip install requests\n",
|
| 27 |
+
"\n",
|
| 28 |
+
"# Verify Python version\n",
|
| 29 |
+
"import sys\n",
|
| 30 |
+
"print(sys.version)\n",
|
| 31 |
+
"\n",
|
| 32 |
+
"# Check GPU availability\n",
|
| 33 |
+
"import torch\n",
|
| 34 |
+
"print(f\"CUDA Available: {torch.cuda.is_available()}\")\n",
|
| 35 |
+
"print(f\"GPU: {torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'No GPU'}\")"
|
| 36 |
+
],
|
| 37 |
+
"execution_count": null,
|
| 38 |
+
"outputs": []
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"cell_type": "markdown",
|
| 42 |
+
"metadata": {},
|
| 43 |
+
"source": [
|
| 44 |
+
"## Step 1: Authenticate and Download Models\n",
|
| 45 |
+
"Authenticate with Hugging Face to download the SDXL base model. Optionally, provide a Civitai API key for LoRA download."
|
| 46 |
+
]
|
| 47 |
+
},
|
| 48 |
+
{
|
| 49 |
+
"cell_type": "code",
|
| 50 |
+
"metadata": {},
|
| 51 |
+
"source": [
|
| 52 |
+
"from huggingface_hub import login\n",
|
| 53 |
+
"import os\n",
|
| 54 |
+
"\n",
|
| 55 |
+
"# Log in to Hugging Face\n",
|
| 56 |
+
"HF_TOKEN = \"YOUR_HF_TOKEN\" # Replace with your Hugging Face token\n",
|
| 57 |
+
"login(HF_TOKEN)\n",
|
| 58 |
+
"\n",
|
| 59 |
+
"# Set Civitai API key (if downloading from Civitai)\n",
|
| 60 |
+
"CIVITAI_API_KEY = \"YOUR_CIVITAI_API_KEY\" # Replace with your Civitai API key or set to None\n",
|
| 61 |
+
"os.environ[\"CIVITAI_API_KEY\"] = CIVITAI_API_KEY if CIVITAI_API_KEY else \"\"\n",
|
| 62 |
+
"\n",
|
| 63 |
+
"# Download SDXL base model\n",
|
| 64 |
+
"from diffusers import StableDiffusionXLPipeline\n",
|
| 65 |
+
"base_model = \"stabilityai/stable-diffusion-xl-base-1.0\"\n",
|
| 66 |
+
"pipeline = StableDiffusionXLPipeline.from_pretrained(\n",
|
| 67 |
+
" base_model,\n",
|
| 68 |
+
" torch_dtype=torch.float16,\n",
|
| 69 |
+
" variant=\"fp16\",\n",
|
| 70 |
+
" use_safetensors=True\n",
|
| 71 |
+
").to(\"cuda\")"
|
| 72 |
+
],
|
| 73 |
+
"execution_count": null,
|
| 74 |
+
"outputs": []
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"cell_type": "markdown",
|
| 78 |
+
"metadata": {},
|
| 79 |
+
"source": [
|
| 80 |
+
"## Step 2: Download Custom LoRA\n",
|
| 81 |
+
"Choose to download a LoRA model from Hugging Face or Civitai. Replace the URLs/IDs with your desired LoRA model."
|
| 82 |
+
]
|
| 83 |
+
},
|
| 84 |
+
{
|
| 85 |
+
"cell_type": "code",
|
| 86 |
+
"metadata": {},
|
| 87 |
+
"source": [
|
| 88 |
+
"import requests\n",
|
| 89 |
+
"import os\n",
|
| 90 |
+
"\n",
|
| 91 |
+
"def download_hf_lora(repo_id, filename, local_dir=\"./lora\"):\n",
|
| 92 |
+
" os.makedirs(local_dir, exist_ok=True)\n",
|
| 93 |
+
" local_path = os.path.join(local_dir, filename)\n",
|
| 94 |
+
" url = f\"https://huggingface.co/{repo_id}/resolve/main/{filename}\"\n",
|
| 95 |
+
" headers = {\"Authorization\": f\"Bearer {HF_TOKEN}\"}\n",
|
| 96 |
+
" response = requests.get(url, headers=headers)\n",
|
| 97 |
+
" response.raise_for_status()\n",
|
| 98 |
+
" with open(local_path, \"wb\") as f:\n",
|
| 99 |
+
" f.write(response.content)\n",
|
| 100 |
+
" return local_path\n",
|
| 101 |
+
"\n",
|
| 102 |
+
"def download_civitai_lora(model_id, filename, local_dir=\"./lora\"):\n",
|
| 103 |
+
" os.makedirs(local_dir, exist_ok=True)\n",
|
| 104 |
+
" local_path = os.path.join(local_dir, filename)\n",
|
| 105 |
+
" url = f\"https://civitai.com/api/download/models/{model_id}\"\n",
|
| 106 |
+
" headers = {\"Authorization\": f\"Bearer {os.environ['CIVITAI_API_KEY']}\"}\n",
|
| 107 |
+
" response = requests.get(url, headers=headers)\n",
|
| 108 |
+
" response.raise_for_status()\n",
|
| 109 |
+
" with open(local_path, \"wb\") as f:\n",
|
| 110 |
+
" f.write(response.content)\n",
|
| 111 |
+
" return local_path\n",
|
| 112 |
+
"\n",
|
| 113 |
+
"# Example: Download LoRA (choose one method)\n",
|
| 114 |
+
"# Hugging Face LoRA (e.g., a hypothetical LoRA model)\n",
|
| 115 |
+
"lora_path = download_hf_lora(\n",
|
| 116 |
+
" repo_id=\"username/sdxl-lora-model\", # Replace with actual Hugging Face repo ID\n",
|
| 117 |
+
" filename=\"model.safetensors\" # Replace with actual filename\n",
|
| 118 |
+
")\n",
|
| 119 |
+
"\n",
|
| 120 |
+
"# Civitai LoRA (uncomment to use)\n",
|
| 121 |
+
"# lora_path = download_civitai_lora(\n",
|
| 122 |
+
"# model_id=\"123456\", # Replace with Civitai model ID\n",
|
| 123 |
+
"# filename=\"lora_model.safetensors\"\n",
|
| 124 |
+
"# )"
|
| 125 |
+
],
|
| 126 |
+
"execution_count": null,
|
| 127 |
+
"outputs": []
|
| 128 |
+
},
|
| 129 |
+
{
|
| 130 |
+
"cell_type": "markdown",
|
| 131 |
+
"metadata": {},
|
| 132 |
+
"source": [
|
| 133 |
+
"## Step 3: Load LoRA into Pipeline\n",
|
| 134 |
+
"Load the custom LoRA weights into the SDXL pipeline."
|
| 135 |
+
]
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"cell_type": "code",
|
| 139 |
+
"metadata": {},
|
| 140 |
+
"source": [
|
| 141 |
+
"# Load LoRA weights\n",
|
| 142 |
+
"pipeline.load_lora_weights(\n",
|
| 143 |
+
" lora_path,\n",
|
| 144 |
+
" adapter_name=\"custom_lora\"\n",
|
| 145 |
+
")\n",
|
| 146 |
+
"\n",
|
| 147 |
+
"# Enable LoRA\n",
|
| 148 |
+
"pipeline.set_adapters([\"custom_lora\"], adapter_weights=[1.0])\n",
|
| 149 |
+
"\n",
|
| 150 |
+
"# Optimize for T4 GPU\n",
|
| 151 |
+
"pipeline.enable_model_cpu_offload()\n",
|
| 152 |
+
"pipeline.enable_vae_slicing()"
|
| 153 |
+
],
|
| 154 |
+
"execution_count": null,
|
| 155 |
+
"outputs": []
|
| 156 |
+
},
|
| 157 |
+
{
|
| 158 |
+
"cell_type": "markdown",
|
| 159 |
+
"metadata": {},
|
| 160 |
+
"source": [
|
| 161 |
+
"## Step 4: Generate Images\n",
|
| 162 |
+
"Configure the prompt and generate images using the SDXL pipeline with LoRA."
|
| 163 |
+
]
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"cell_type": "code",
|
| 167 |
+
"metadata": {},
|
| 168 |
+
"source": [
|
| 169 |
+
"prompt = \"A futuristic cityscape at sunset, cyberpunk style, highly detailed, vibrant colors\"\n",
|
| 170 |
+
"negative_prompt = \"blurry, low quality, artifacts\"\n",
|
| 171 |
+
"\n",
|
| 172 |
+
"images = pipeline(\n",
|
| 173 |
+
" prompt=prompt,\n",
|
| 174 |
+
" negative_prompt=negative_prompt,\n",
|
| 175 |
+
" num_inference_steps=30,\n",
|
| 176 |
+
" guidance_scale=7.5,\n",
|
| 177 |
+
" height=1024,\n",
|
| 178 |
+
" width=1024,\n",
|
| 179 |
+
" num_images_per_prompt=1\n",
|
| 180 |
+
").images\n",
|
| 181 |
+
"\n",
|
| 182 |
+
"# Save and display the image\n",
|
| 183 |
+
"images[0].save(\"output.png\")\n",
|
| 184 |
+
"images[0]"
|
| 185 |
+
],
|
| 186 |
+
"execution_count": null,
|
| 187 |
+
"outputs": []
|
| 188 |
+
},
|
| 189 |
+
{
|
| 190 |
+
"cell_type": "markdown",
|
| 191 |
+
"metadata": {},
|
| 192 |
+
"source": [
|
| 193 |
+
"## Notes\n",
|
| 194 |
+
"- **T4 GPU Optimization**: The notebook uses `float16` precision, VAE slicing, and model CPU offloading to fit within T4 GPU memory constraints (16GB VRAM).\n",
|
| 195 |
+
"- **LoRA Model**: Ensure the LoRA model is compatible with SDXL. Replace placeholder repo IDs or model IDs with actual values from Hugging Face or Civitai.\n",
|
| 196 |
+
"- **Performance**: Adjust `num_inference_steps` and `guidance_scale` for quality vs. speed trade-offs.\n",
|
| 197 |
+
"- **Storage**: Generated images are saved as `output.png` in the Colab environment. Download them manually if needed."
|
| 198 |
+
]
|
| 199 |
+
}
|
| 200 |
+
],
|
| 201 |
+
"metadata": {
|
| 202 |
+
"kernelspec": {
|
| 203 |
+
"display_name": "Python 3",
|
| 204 |
+
"language": "python",
|
| 205 |
+
"name": "python3"
|
| 206 |
+
},
|
| 207 |
+
"language_info": {
|
| 208 |
+
"codemirror_mode": {
|
| 209 |
+
"name": "ipython",
|
| 210 |
+
"version": 3
|
| 211 |
+
},
|
| 212 |
+
"file_extension": ".py",
|
| 213 |
+
"mimetype": "text/x-python",
|
| 214 |
+
"name": "python",
|
| 215 |
+
"nbconvert_exporter": "python",
|
| 216 |
+
"pygments_lexer": "ipython3",
|
| 217 |
+
"version": "3.11.0"
|
| 218 |
+
}
|
| 219 |
+
},
|
| 220 |
+
"nbformat": 4,
|
| 221 |
+
"nbformat_minor": 4
|
| 222 |
+
}
|