SDXL_base_with_refiner.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "gpuType": "T4",
      "authorship_tag": "ABX9TyMJlT1AqYb2ZEW9ogHAs7P4",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/LeandroBerlin/SDXL-base-with-refiner/blob/main/SDXL_base_with_refiner.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "pc-ZEGrWfDe9"
      },
      "outputs": [],
      "source": [
        "pip install diffusers --upgrade"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "pip install invisible_watermark transformers accelerate safetensors"
      ],
      "metadata": {
        "id": "FgpVjljffHWB"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# @title\n",
        "from diffusers import DiffusionPipeline\n",
        "import torch\n",
        "\n",
        "# load both base & refiner\n",
        "base = DiffusionPipeline.from_pretrained(\n",
        "    \"stabilityai/stable-diffusion-xl-base-1.0\", torch_dtype=torch.float16, variant=\"fp16\", use_safetensors=True,force_download=True\n",
        ")\n",
        "base.to(\"cuda\")\n",
        "refiner = DiffusionPipeline.from_pretrained(\n",
        "    \"stabilityai/stable-diffusion-xl-refiner-1.0\",\n",
        "    text_encoder_2=base.text_encoder_2,\n",
        "    vae=base.vae,\n",
        "    torch_dtype=torch.float16,\n",
        "    use_safetensors=True,\n",
        "    variant=\"fp16\",\n",
        ")\n",
        "refiner.to(\"cuda\")\n",
        "\n",
        "# Define how many steps and what % of steps to be run on each experts (80/20) here\n",
        "n_steps = 40\n",
        "high_noise_frac = 0.8\n",
        "\n",
        "prompt = \"ultra realistic portrait of dog, Canon EOS R3, nikon, f/1.4, ISO 200, 1/160s, 8K, RAW, unedited, symmetrical balance, in-frame, 8K\"\n",
        "\n",
        "# run both experts\n",
        "image = base(\n",
        "    prompt=prompt,\n",
        "    num_inference_steps=n_steps,\n",
        "    denoising_end=high_noise_frac,\n",
        "    output_type=\"latent\",\n",
        ").images\n",
        "image = refiner(\n",
        "    prompt=prompt,\n",
        "    num_inference_steps=n_steps,\n",
        "    denoising_start=high_noise_frac,\n",
        "    image=image,\n",
        ").images[0]\n",
        "\n",
        "image"
      ],
      "metadata": {
        "id": "UYHc7-p_fKBo"
      },
      "execution_count": null,
      "outputs": []
    }
  ]
}