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": [
        "### Install"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "cellView": "code",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ix_xt4X1_6F4",
        "outputId": "c72b1def-798b-401a-fad5-25e715f60dfe"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Sun Jul 17 11:27:12 2022       \n",
            "+-----------------------------------------------------------------------------+\n",
            "| NVIDIA-SMI 460.32.03    Driver Version: 460.32.03    CUDA Version: 11.2     |\n",
            "|-------------------------------+----------------------+----------------------+\n",
            "| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |\n",
            "| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |\n",
            "|                               |                      |               MIG M. |\n",
            "|===============================+======================+======================|\n",
            "|   0  Tesla P100-PCIE...  Off  | 00000000:00:04.0 Off |                    0 |\n",
            "| N/A   36C    P0    26W / 250W |      0MiB / 16280MiB |      0%      Default |\n",
            "|                               |                      |                  N/A |\n",
            "+-------------------------------+----------------------+----------------------+\n",
            "                                                                               \n",
            "+-----------------------------------------------------------------------------+\n",
            "| Processes:                                                                  |\n",
            "|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |\n",
            "|        ID   ID                                                   Usage      |\n",
            "|=============================================================================|\n",
            "|  No running processes found                                                 |\n",
            "+-----------------------------------------------------------------------------+\n",
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting min-dalle==0.3.15\n",
            "  Downloading min-dalle-0.3.15.tar.gz (10 kB)\n",
            "Requirement already satisfied: torch>=1.11 in /usr/local/lib/python3.7/dist-packages (from min-dalle==0.3.15) (1.12.0+cu113)\n",
            "Requirement already satisfied: typing_extensions>=4.1 in /usr/local/lib/python3.7/dist-packages (from min-dalle==0.3.15) (4.1.1)\n",
            "Requirement already satisfied: numpy>=1.21 in /usr/local/lib/python3.7/dist-packages (from min-dalle==0.3.15) (1.21.6)\n",
            "Requirement already satisfied: pillow>=7.1 in /usr/local/lib/python3.7/dist-packages (from min-dalle==0.3.15) (7.1.2)\n",
            "Requirement already satisfied: requests>=2.23 in /usr/local/lib/python3.7/dist-packages (from min-dalle==0.3.15) (2.23.0)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests>=2.23->min-dalle==0.3.15) (2022.6.15)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests>=2.23->min-dalle==0.3.15) (2.10)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests>=2.23->min-dalle==0.3.15) (3.0.4)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests>=2.23->min-dalle==0.3.15) (1.24.3)\n",
            "Building wheels for collected packages: min-dalle\n",
            "  Building wheel for min-dalle (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for min-dalle: filename=min_dalle-0.3.15-py3-none-any.whl size=10562 sha256=3fa7138ec8078d7442a6a0a4c02653759b255d5853bfaa1710c5746399111aea\n",
            "  Stored in directory: /root/.cache/pip/wheels/92/58/05/36292d8d8de42c1d8afa532a27cbd026daaeab31819ac2b3c8\n",
            "Successfully built min-dalle\n",
            "Installing collected packages: min-dalle\n",
            "Successfully installed min-dalle-0.3.15\n"
          ]
        }
      ],
      "source": [
        "! nvidia-smi\n",
        "! pip install min-dalle"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kViq2dMbGDKt"
      },
      "source": [
        "### Load Model\n",
        "`float32` is faster than `float16` but uses more GPU memory.  Change the `grid_size` to 3 or less if using `float32`."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8W-L2ICFGFup",
        "outputId": "9ae4d35c-52f8-4733-9972-e46755cba899"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "using device cuda\n",
            "downloading tokenizer params\n",
            "intializing TextTokenizer\n",
            "downloading encoder params\n",
            "initializing DalleBartEncoder\n",
            "downloading decoder params\n",
            "initializing DalleBartDecoder\n",
            "downloading detokenizer params\n",
            "initializing VQGanDetokenizer\n"
          ]
        }
      ],
      "source": [
        "dtype = \"float16\" #@param [\"float32\", \"float16\", \"bfloat16\"]\n",
        "from IPython.display import display, update_display\n",
        "import torch\n",
        "from min_dalle import MinDalle\n",
        "\n",
        "model = MinDalle(\n",
        "    dtype=getattr(torch, dtype),\n",
        "    device='cuda',\n",
        "    is_mega=True, \n",
        "    is_reusable=True\n",
        ")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "c52TV1GbBNgS"
      },
      "source": [
        "### Generate Images\n",
        "\n",
        "- `grid_size` Size of the image grid.  Reduce this if you run out of GPU memory.\n",
        "\n",
        "- `progressive_outputs` Whether to show intermediate output.  Adds a small delay and increases memory usage.\n",
        "\n",
        "- `seamless` Tiles the images in token space instead of pixel space\n",
        "\n",
        "- `supercondition_factor` Higher values result in better agreement with the text but a narrower variety of generated images\n",
        "\n",
        "- `top_k` Each image token is sampled from the top $k$ most probable tokens"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "nQ0UG05dA4p2",
        "outputId": "7faf570a-3444-4e6a-d117-a4e1d0b09ce5"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<PIL.Image.Image image mode=RGB size=1024x1024 at 0x7FA55DFEEE50>"
            ],
            "image/png": "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
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "CPU times: user 55 s, sys: 771 ms, total: 55.8 s\n",
            "Wall time: 1min 3s\n"
          ]
        }
      ],
      "source": [
        "%%time\n",
        "\n",
        "text = \"Dali painting of WALL·E\" #@param {type:\"string\"}\n",
        "progressive_outputs = True #@param {type:\"boolean\"}\n",
        "seamless = True #@param {type:\"boolean\"}\n",
        "grid_size = 4 #@param {type:\"integer\"}\n",
        "temperature = 2 #@param {type:\"slider\", min:0.01, max:3, step:0.01}\n",
        "supercondition_factor = 16 #@param {type:\"number\"}\n",
        "top_k = 128 #@param {type:\"integer\"}\n",
        "\n",
        "image_stream = model.generate_image_stream(\n",
        "    text=text,\n",
        "    seed=-1,\n",
        "    grid_size=grid_size,\n",
        "    progressive_outputs=progressive_outputs,\n",
        "    is_seamless=seamless,\n",
        "    temperature=temperature,\n",
        "    top_k=int(top_k),\n",
        "    supercondition_factor=float(supercondition_factor)\n",
        ")\n",
        "\n",
        "is_first = True\n",
        "for image in image_stream:\n",
        "    display_image = display if is_first else update_display\n",
        "    display_image(image, display_id=1)\n",
        "    is_first = False"
      ]
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "collapsed_sections": [
        "Zl_ZFisFApeh"
      ],
      "name": "min-dalle",
      "provenance": [],
      "include_colab_link": true
    },
    "gpuClass": "standard",
    "kernelspec": {
      "display_name": "Python 3.9.13 64-bit",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "name": "python",
      "version": "3.9.13"
    },
    "vscode": {
      "interpreter": {
        "hash": "b0fa6594d8f4cbf19f97940f81e996739fb7646882a419484c72d19e05852a7e"
      }
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}