min-dalle-test/min_dalle.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/kuprel/min-dalle/blob/main/min_dalle.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3WL-G_f2_ld8"
      },
      "source": [
        "# min(DALL·E)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Zl_ZFisFApeh"
      },
      "source": [
        "### Download models and install dependencies"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "cellView": "code",
        "id": "ix_xt4X1_6F4"
      },
      "outputs": [],
      "source": [
        "%%shell\n",
        "\n",
        "git clone https://github.com/kuprel/min-dalle\n",
        "mkdir -p /content/min-dalle/pretrained/vqgan/\n",
        "curl https://huggingface.co/dalle-mini/vqgan_imagenet_f16_16384/resolve/main/flax_model.msgpack -L --output /content/min-dalle/pretrained/vqgan/flax_model.msgpack\n",
        "pip install torch flax==0.4.2 wandb\n",
        "wandb login --anonymously\n",
        "wandb artifact get --root=/content/min-dalle/pretrained/dalle_bart_mini dalle-mini/dalle-mini/mini-1:v0\n",
        "wandb artifact get --root=/content/min-dalle/pretrained/dalle_bart_mega dalle-mini/dalle-mini/mega-1-fp16:v14\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kViq2dMbGDKt"
      },
      "source": [
        "### Load Model\n",
        "Note: if you are using a high-RAM runtime check the \"reusable\" box.  It loads the model once and allows it to be reused, decreasing the time to generate an image from over a minute down to about 10 seconds"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8W-L2ICFGFup",
        "outputId": "2fe867bc-ee6e-49d6-f2ad-f2e25301c548"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "initializing MinDalleTorch\n",
            "reading files from pretrained/dalle_bart_mega\n",
            "converting params to torch\n",
            "initializing DalleBartEncoderTorch\n",
            "initializing DalleBartDecoderTorch\n",
            "initializing VQGanDetokenizer\n"
          ]
        }
      ],
      "source": [
        "import os\n",
        "os.chdir('/content/min-dalle')\n",
        "from min_dalle.min_dalle_torch import MinDalleTorch\n",
        "from min_dalle.min_dalle_flax import MinDalleFlax\n",
        "\n",
        "mega = True #@param {type:\"boolean\"}\n",
        "torch = True #@param {type:\"boolean\"}\n",
        "reusable = False #@param {type:\"boolean\"}\n",
        "model_class = MinDalleTorch if torch else MinDalleFlax\n",
        "model = model_class(mega, reusable)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "c52TV1GbBNgS"
      },
      "source": [
        "### Generate an Image"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 528
        },
        "id": "nQ0UG05dA4p2",
        "outputId": "a7611ef7-6546-4ecd-ba3e-7fcc83e8e4dd"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tokenizing text\n",
            "['Ġa']\n",
            "['Ġcomfy']\n",
            "['Ġchair']\n",
            "['Ġthat']\n",
            "['Ġlooks']\n",
            "['Ġlike']\n",
            "['Ġan']\n",
            "['Ġavocado']\n",
            "text tokens [0, 58, 29872, 2408, 766, 4126, 1572, 101, 16632, 2]\n",
            "encoding text tokens\n",
            "sampling image tokens\n",
            "detokenizing image\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<PIL.Image.Image image mode=RGB size=256x256 at 0x7FDE245FB1D0>"
            ],
            "image/png": "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
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "CPU times: user 10.7 s, sys: 29.3 ms, total: 10.7 s\n",
            "Wall time: 10.6 s\n"
          ]
        }
      ],
      "source": [
        "%%time\n",
        "\n",
        "text = \"a comfy chair that looks like an avocado\" #@param {type:\"string\"}\n",
        "seed = 10 #@param {type:\"integer\"}\n",
        "\n",
        "image = model.generate_image(text, seed)\n",
        "display(image)"
      ]
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "collapsed_sections": [
        "Zl_ZFisFApeh"
      ],
      "machine_shape": "hm",
      "name": "min-dalle",
      "provenance": [],
      "include_colab_link": true
    },
    "gpuClass": "standard",
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}