DiffusionModels.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/Erickrus/leetcode/blob/master/DiffusionModels.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "N04LD306wMEg"
      },
      "source": [
        "# Diffusion Models\n",
        "This is an easy-to-understand implementation of diffusion models within 100 lines of code. Different from other implementations, this code doesn't use the lower-bound formulation for sampling and strictly follows Algorithm 1 from the [DDPM](https://arxiv.org/pdf/2006.11239.pdf) paper, which makes it extremely short and easy to follow. There is only one implementation: `unconditional` in this notebook. Below you can find two explanation videos for the theory behind diffusion models and the implementation.\n",
        "\n",
        "<a href=\"https://www.youtube.com/watch?v=HoKDTa5jHvg\">\n",
        "   <img alt=\"Qries\" src=\"https://user-images.githubusercontent.com/61938694/191407922-f613759e-4bea-4ac9-9135-d053a6312421.jpg\"\n",
        "   width=\"300\">\n",
        "</a>\n",
        "\n",
        "<a href=\"https://www.youtube.com/watch?v=TBCRlnwJtZU\">\n",
        "   <img alt=\"Qries\" src=\"https://user-images.githubusercontent.com/61938694/191407849-6d0376c7-05b2-43cd-a75c-1280b0e33af1.png\"\n",
        "   width=\"300\">\n",
        "</a>\n",
        "\n",
        "<hr>\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XC6PidnDHP3M"
      },
      "source": [
        "### prepare CIFAR-10 dataset"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "N3VenSfKqL1D",
        "outputId": "8c28150c-50ce-4c45-d4c4-478b180c2555",
        "cellView": "form"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "/content\n",
            "Cloning into 'CIFAR-10-images'...\n",
            "remote: Enumerating objects: 60027, done.\u001b[K\n",
            "remote: Total 60027 (delta 0), reused 0 (delta 0), pack-reused 60027\u001b[K\n",
            "Receiving objects: 100% (60027/60027), 19.94 MiB | 2.54 MiB/s, done.\n",
            "Resolving deltas: 100% (59990/59990), done.\n",
            "Checking out files: 100% (60001/60001), done.\n"
          ]
        }
      ],
      "source": [
        "#@title download CIFAR-10 dataset\n",
        "%cd /content\n",
        "!rm -Rf /content/ddpm\n",
        "!rm -Rf /content/CIFAR-10-images\n",
        "!git clone https://github.com/YoongiKim/CIFAR-10-images\n",
        "!mkdir -p /content/ddpm/dataset\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 38,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cf_puzcw06ci",
        "outputId": "46f22058-90ed-4949-cf49-79ae661b779a",
        "cellView": "form"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "shrink CIFAR-10 dataset to 9999 per category /content/ddpm/dataset/\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 10000/10000 [14:13<00:00, 11.72it/s]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\n",
            " 10000 images left\n"
          ]
        }
      ],
      "source": [
        "#@title shrink dataset size\n",
        "#@markdown Image to keep per category\n",
        "\n",
        "import glob, os\n",
        "from tqdm import tqdm\n",
        "\n",
        "nImageToKeep = 9999 #@param {\"type\":\"integer\"}\n",
        "\n",
        "srcDir = '/content/CIFAR-10-images/test/'\n",
        "tgtDir = '/content/ddpm/dataset/'\n",
        "categoryNames = [os.path.basename(x[0]) for x in os.walk(srcDir)]\n",
        "os.system(\"rm -Rf %s\" % tgtDir)\n",
        "os.system(\"mkdir -p %s\" % tgtDir)\n",
        "for categoryName in categoryNames:\n",
        "  os.system(\"mkdir -p %s\" % os.path.join(tgtDir, categoryName))\n",
        "\n",
        "print(\"shrink CIFAR-10 dataset to %d per category %s\" %(nImageToKeep, tgtDir))\n",
        "\n",
        "for filename in tqdm(glob.glob(os.path.join(srcDir, \"**\", \"*.jpg\"))):\n",
        "  shortFilename = os.path.basename(filename).replace(\".jpg\",\"\")\n",
        "  categoryName = filename.split(\"/\")[-2]\n",
        "  fileId = int(shortFilename)\n",
        "  if fileId < nImageToKeep:\n",
        "    os.system(\"cp -f %s %s\" % (filename, os.path.join(tgtDir, categoryName)))\n",
        "print(\"\\n\",len(glob.glob(os.path.join(tgtDir, \"**\", \"*.jpg\"))), \"images left\")\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "UmFHvvM1GDUu"
      },
      "source": [
        "### modules.py"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 39,
      "metadata": {
        "cellView": "form",
        "id": "sFJsW_EVw2hR"
      },
      "outputs": [],
      "source": [
        "#@title SelfAttention\n",
        "#@markdown self attention\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.nn.functional as F\n",
        "\n",
        "class SelfAttention(nn.Module):\n",
        "    def __init__(self, channels, size):\n",
        "        super(SelfAttention, self).__init__()\n",
        "        self.channels = channels\n",
        "        self.size = size\n",
        "        self.mha = nn.MultiheadAttention(channels, 4, batch_first=True)\n",
        "        self.ln = nn.LayerNorm([channels])\n",
        "        self.ff_self = nn.Sequential(\n",
        "            nn.LayerNorm([channels]),\n",
        "            nn.Linear(channels, channels),\n",
        "            nn.GELU(),\n",
        "            nn.Linear(channels, channels),\n",
        "        )\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = x.view(-1, self.channels, self.size * self.size).swapaxes(1, 2)\n",
        "        x_ln = self.ln(x)\n",
        "        attention_value, _ = self.mha(x_ln, x_ln, x_ln)\n",
        "        attention_value = attention_value + x\n",
        "        attention_value = self.ff_self(attention_value) + attention_value\n",
        "        return attention_value.swapaxes(2, 1).view(-1, self.channels, self.size, self.size)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 40,
      "metadata": {
        "cellView": "form",
        "id": "13urmSLyGR7G"
      },
      "outputs": [],
      "source": [
        "#@title DoubleConv\n",
        "#@markdown double convolution\n",
        "class DoubleConv(nn.Module):\n",
        "    def __init__(self, in_channels, out_channels, mid_channels=None, residual=False):\n",
        "        super().__init__()\n",
        "        self.residual = residual\n",
        "        if not mid_channels:\n",
        "            mid_channels = out_channels\n",
        "        self.double_conv = nn.Sequential(\n",
        "            nn.Conv2d(in_channels, mid_channels, kernel_size=3, padding=1, bias=False),\n",
        "            nn.GroupNorm(1, mid_channels),\n",
        "            nn.GELU(),\n",
        "            nn.Conv2d(mid_channels, out_channels, kernel_size=3, padding=1, bias=False),\n",
        "            nn.GroupNorm(1, out_channels),\n",
        "        )\n",
        "\n",
        "    def forward(self, x):\n",
        "        if self.residual:\n",
        "            return F.gelu(x + self.double_conv(x))\n",
        "        else:\n",
        "            return self.double_conv(x)\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 41,
      "metadata": {
        "cellView": "form",
        "id": "CUghivY4Gb1E"
      },
      "outputs": [],
      "source": [
        "#@title Down\n",
        "#@markdown downsample process\n",
        "class Down(nn.Module):\n",
        "    def __init__(self, in_channels, out_channels, emb_dim=256):\n",
        "        super().__init__()\n",
        "        self.maxpool_conv = nn.Sequential(\n",
        "            nn.MaxPool2d(2),\n",
        "            DoubleConv(in_channels, in_channels, residual=True),\n",
        "            DoubleConv(in_channels, out_channels),\n",
        "        )\n",
        "\n",
        "        self.emb_layer = nn.Sequential(\n",
        "            nn.SiLU(),\n",
        "            nn.Linear(\n",
        "                emb_dim,\n",
        "                out_channels\n",
        "            ),\n",
        "        )\n",
        "\n",
        "    def forward(self, x, t):\n",
        "        x = self.maxpool_conv(x)\n",
        "        emb = self.emb_layer(t)[:, :, None, None].repeat(1, 1, x.shape[-2], x.shape[-1])\n",
        "        return x + emb\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 42,
      "metadata": {
        "cellView": "form",
        "id": "2uH4s-bBGgbH"
      },
      "outputs": [],
      "source": [
        "#@title Up\n",
        "#@markdown upsample process\n",
        "class Up(nn.Module):\n",
        "    def __init__(self, in_channels, out_channels, emb_dim=256):\n",
        "        super().__init__()\n",
        "\n",
        "        self.up = nn.Upsample(scale_factor=2, mode=\"bilinear\", align_corners=True)\n",
        "        self.conv = nn.Sequential(\n",
        "            DoubleConv(in_channels, in_channels, residual=True),\n",
        "            DoubleConv(in_channels, out_channels, in_channels // 2),\n",
        "        )\n",
        "\n",
        "        self.emb_layer = nn.Sequential(\n",
        "            nn.SiLU(),\n",
        "            nn.Linear(\n",
        "                emb_dim,\n",
        "                out_channels\n",
        "            ),\n",
        "        )\n",
        "\n",
        "    def forward(self, x, skip_x, t):\n",
        "        x = self.up(x)\n",
        "        x = torch.cat([skip_x, x], dim=1)\n",
        "        x = self.conv(x)\n",
        "        emb = self.emb_layer(t)[:, :, None, None].repeat(1, 1, x.shape[-2], x.shape[-1])\n",
        "        return x + emb\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 43,
      "metadata": {
        "cellView": "form",
        "id": "txX02reiGmd3"
      },
      "outputs": [],
      "source": [
        "#@title UNet\n",
        "#@markdown UNet is $\\epsilon_{\\theta}(x_t,t)$\n",
        "#@markdown\n",
        "#@markdown <img src=\"https://huggingface.co/blog/assets/78_annotated-diffusion/unet_architecture.jpg\" width=\"600px\">\n",
        "#@markdown \n",
        "#@markdown The author went for a U-Net, introduced by (Ronneberger et al., 2015) (which, at the time, achieved state-of-the-art results for medical image segmentation). This network, like any autoencoder, consists of a bottleneck in the middle that makes sure the network learns only the most important information. Importantly, it introduced residual connections between the encoder and decoder, greatly improving gradient flow (inspired by ResNet in He et al., 2015).\n",
        "class UNet(nn.Module):\n",
        "    def __init__(self, c_in=3, c_out=3, time_dim=256, device=\"cuda\"):\n",
        "        super().__init__()\n",
        "        self.device = device\n",
        "        self.time_dim = time_dim\n",
        "        self.inc = DoubleConv(c_in, 64)\n",
        "        self.down1 = Down(64, 128)\n",
        "        self.sa1 = SelfAttention(128, 32)\n",
        "        self.down2 = Down(128, 256)\n",
        "        self.sa2 = SelfAttention(256, 16)\n",
        "        self.down3 = Down(256, 256)\n",
        "        self.sa3 = SelfAttention(256, 8)\n",
        "\n",
        "        self.bot1 = DoubleConv(256, 512)\n",
        "        self.bot2 = DoubleConv(512, 512)\n",
        "        self.bot3 = DoubleConv(512, 256)\n",
        "\n",
        "        self.up1 = Up(512, 128)\n",
        "        self.sa4 = SelfAttention(128, 16)\n",
        "        self.up2 = Up(256, 64)\n",
        "        self.sa5 = SelfAttention(64, 32)\n",
        "        self.up3 = Up(128, 64)\n",
        "        self.sa6 = SelfAttention(64, 64)\n",
        "        self.outc = nn.Conv2d(64, c_out, kernel_size=1)\n",
        "\n",
        "    def pos_encoding(self, t, channels):\n",
        "        inv_freq = 1.0 / (\n",
        "            10000\n",
        "            ** (torch.arange(0, channels, 2, device=self.device).float() / channels)\n",
        "        )\n",
        "        pos_enc_a = torch.sin(t.repeat(1, channels // 2) * inv_freq)\n",
        "        pos_enc_b = torch.cos(t.repeat(1, channels // 2) * inv_freq)\n",
        "        pos_enc = torch.cat([pos_enc_a, pos_enc_b], dim=-1)\n",
        "        return pos_enc\n",
        "\n",
        "    def forward(self, x, t):\n",
        "        t = t.unsqueeze(-1).type(torch.float)\n",
        "        t = self.pos_encoding(t, self.time_dim)\n",
        "\n",
        "        x1 = self.inc(x)\n",
        "        x2 = self.down1(x1, t)\n",
        "        x2 = self.sa1(x2)\n",
        "        x3 = self.down2(x2, t)\n",
        "        x3 = self.sa2(x3)\n",
        "        x4 = self.down3(x3, t)\n",
        "        x4 = self.sa3(x4)\n",
        "\n",
        "        x4 = self.bot1(x4)\n",
        "        x4 = self.bot2(x4)\n",
        "        x4 = self.bot3(x4)\n",
        "\n",
        "        x = self.up1(x4, x3, t)\n",
        "        x = self.sa4(x)\n",
        "        x = self.up2(x, x2, t)\n",
        "        x = self.sa5(x)\n",
        "        x = self.up3(x, x1, t)\n",
        "        x = self.sa6(x)\n",
        "        output = self.outc(x)\n",
        "        return output"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BAK6POjGHCxH"
      },
      "source": [
        "###utils.py"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 44,
      "metadata": {
        "cellView": "form",
        "id": "wueS1Kv3xI5g"
      },
      "outputs": [],
      "source": [
        "#@title utils.py\n",
        "import os\n",
        "import torch\n",
        "import torchvision\n",
        "from PIL import Image\n",
        "from matplotlib import pyplot as plt\n",
        "from torch.utils.data import DataLoader\n",
        "\n",
        "#@markdown plot_images\n",
        "def plot_images(images):\n",
        "    plt.figure(figsize=(32, 32))\n",
        "    plt.imshow(torch.cat([\n",
        "        torch.cat([i for i in images.cpu()], dim=-1),\n",
        "    ], dim=-2).permute(1, 2, 0).cpu())\n",
        "    plt.show()\n",
        "\n",
        "#@markdown save_images\n",
        "def save_images(images, path, **kwargs):\n",
        "    grid = torchvision.utils.make_grid(images, **kwargs)\n",
        "    ndarr = grid.permute(1, 2, 0).to('cpu').numpy()\n",
        "    im = Image.fromarray(ndarr)\n",
        "    im.save(path)\n",
        "    display(im)\n",
        "\n",
        "#@markdown get_data\n",
        "def get_data(args):\n",
        "    transforms = torchvision.transforms.Compose([\n",
        "        torchvision.transforms.Resize(80),  # args.image_size + 1/4 *args.image_size\n",
        "        torchvision.transforms.RandomResizedCrop(args.image_size, scale=(0.8, 1.0)),\n",
        "        torchvision.transforms.ToTensor(),\n",
        "        torchvision.transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))\n",
        "    ])\n",
        "    dataset = torchvision.datasets.ImageFolder(args.dataset_path, transform=transforms)\n",
        "    dataloader = DataLoader(dataset, batch_size=args.batch_size, shuffle=True)\n",
        "    return dataloader\n",
        "\n",
        "#@markdown setup_logging\n",
        "def setup_logging(run_name):\n",
        "    os.makedirs(\"models\", exist_ok=True)\n",
        "    os.makedirs(\"results\", exist_ok=True)\n",
        "    os.makedirs(os.path.join(\"models\", run_name), exist_ok=True)\n",
        "    os.makedirs(os.path.join(\"results\", run_name), exist_ok=True)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Fpq3wLN3I_RW"
      },
      "source": [
        "###Diffusion Class"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 32,
      "metadata": {
        "cellView": "form",
        "id": "Qj6FQJfIl8eY"
      },
      "outputs": [],
      "source": [
        "#@title Reparameterization Trick\n",
        "import datetime\n",
        "import os\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "from matplotlib import pyplot as plt\n",
        "from tqdm import tqdm\n",
        "from torch import optim\n",
        "#import logging\n",
        "from torch.utils.tensorboard import SummaryWriter\n",
        "\n",
        "#logging.basicConfig(format=\"%(asctime)s - %(levelname)s: %(message)s\", level=logging.INFO, datefmt=\"%I:%M:%S\")\n",
        "\n",
        "\n",
        "class Diffusion:\n",
        "    #@markdown `Diffusion.__init__(self, noise_steps=1000, beta_start=1e-4, beta_end=0.02, img_size=256, device=\"cuda\")`\n",
        "    def __init__(self, noise_steps=1000, beta_start=1e-4, beta_end=0.02, img_size=256, device=\"cuda\"):\n",
        "        self.noise_steps = noise_steps\n",
        "        self.beta_start = beta_start\n",
        "        self.beta_end = beta_end\n",
        "        self.img_size = img_size\n",
        "        self.device = device\n",
        "\n",
        "        #@markdown define linear beta model\n",
        "        #@markdown > use `Diffusion.prepare_noise_schedule()`\n",
        "        #@markdown > and $\\beta_{1:T}$ basically just linspace($\\beta_{start}$, $\\beta_{end}$, $T$)\n",
        "        #@markdown > <img src='https://lilianweng.github.io/posts/2021-07-11-diffusion-models/diffusion-beta.png' width=350px>\n",
        "        self.beta = self.prepare_noise_schedule().to(device)\n",
        "        #@markdown \n",
        "        #@markdown DDPM Section 2: **Background**\n",
        "\n",
        "        #@markdown > formula (2): $q(x_{1:T}|x_0) := \\displaystyle\\prod_{t=1}^{T}q(x_t|x_{t−1})$\n",
        "        #@markdown > $q(x_{t}|x_{t-1}) := N (x_t;\\sqrt{1 − \\beta_t}x_{t−1}, \\beta_tI)$ \n",
        "\n",
        "\n",
        "        #@markdown \n",
        "        #@markdown > formula (4): $q(x_t|x_0)=N(x_t;\\sqrt{\\bar\\alpha_t}x_0,(1-\\bar\\alpha_t)I)$  \n",
        "        #@markdown \n",
        "\n",
        "        #@markdown >> **reparameterization trick**\n",
        "        #@markdown\n",
        "        #@markdown >> Let $\\alpha_t=1-\\beta_t$\n",
        "        self.alpha = 1. - self.beta\n",
        "        #@markdown >> Let $\\bar\\alpha=\\displaystyle\\prod^{t}_{s=1}\\alpha_s$, accumulative product of $\\alpha$\n",
        "        #@markdown\n",
        "        #@markdown >> *Notes: In the code, $\\bar\\alpha$ is called as alpha_hat, which is not proper*\n",
        "        self.alpha_hat = torch.cumprod(self.alpha, dim=0)\n",
        "\n",
        "        #@markdown >> **derivations**\n",
        "        #@markdown\n",
        "        #@markdown >> $x_t=\\sqrt{\\alpha_t}x_{t-1} + \\sqrt{1-\\alpha_t}\\epsilon_{t-1}$\n",
        "        #@markdown >> ; where $\\epsilon_{t-1}, \\epsilon_{t-2}, ... \\sim N(0,I)$\n",
        "\n",
        "        #@markdown >> $=\\sqrt{\\alpha_t\\alpha_{t-1}}x_{t-2} + \\sqrt{1-\\alpha_t\\alpha_{t-1}}\\bar\\epsilon_{t-2}$\n",
        "        #@markdown >> ; where $\\bar\\epsilon_{t-2}$ merges 2 Gaussians $(*)$.\n",
        "\n",
        "        #@markdown\n",
        "        #@markdown >> $=\\ ...$\n",
        "        #@markdown\n",
        "        #@markdown >> $=\\sqrt{\\alpha_t\\alpha_{t-1}...\\alpha_{0}}x_0 + \\sqrt{1-\\alpha_t\\alpha_{t-1}...\\alpha_0}\\epsilon$\n",
        "        #@markdown\n",
        "        #@markdown >> $=\\sqrt{\\bar\\alpha_t}x_0 + \\sqrt{1-\\bar\\alpha_t}\\epsilon$\n",
        "\n",
        "        #@markdown >> After a series of derivations, we get: \n",
        "        #@markdown >> $x_t = \\sqrt{\\bar\\alpha_t}x_0+\\sqrt{1-\\bar\\alpha_t}\\epsilon$ ; which means \n",
        "        #@markdown >> $q(x_t|x_0)=N(x_t;\\sqrt{\\bar\\alpha_t}x_0,(1-\\bar\\alpha_t)I)$\n",
        "\n",
        "\n",
        "        #@markdown >> (*) Recall that when we merge two Gaussians with different variance, $N(0,\\sigma^2_1I)$ and $N(0,\\sigma^2_2I)$\n",
        "        #@markdown >> , the new distribution is $N(0,(\\sigma^2_1+\\sigma^2_2)I)$. Here the merged standard deviation is $\\sqrt{(1-\\alpha_t)-\\alpha_t(1-\\alpha_{t-1})}=\\sqrt{1-\\alpha_t\\alpha_{t-1}}$\n",
        "        #@markdown >> \n",
        "        #@markdown >> Usually, we can afford a larger update step when the sample gets noisier, so $\\beta_1<\\beta_2<...<\\beta_T$ and therefore $\\bar\\alpha_1 > ... > \\bar\\alpha_T$\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "        '''\n",
        "        #markdown > formula (6): $q(x_{t−1}|x_t, x_0) = N (x_{t−1}; \\tilde\\mu_t(x_t, x_0), \\tilde\\beta_tI),$\n",
        "        #markdown \n",
        "        #markdown > formula (7): where $\\tilde\\mu_t(x_t, x_0) := \\frac{\\sqrt{\\bar\\alpha_{t-1}}\\beta_t}{1-\\bar\\alpha_t}x_0 + \\frac{\\sqrt{\\alpha_t}(1-\\bar\\alpha_{t-1})}{1-\\bar\\alpha_t}x_t$\n",
        "        #markdown > and $\\tilde\\beta_t:=\\frac{{1-\\bar\\alpha_{t-1}}}{1-\\bar\\alpha_t}\\beta_t$\n",
        "        '''"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 33,
      "metadata": {
        "cellView": "form",
        "id": "mBXT0fAiH5M1"
      },
      "outputs": [],
      "source": [
        "#@title prepare_noise_schedule, noise_images, sample_timesteps\n",
        "\n",
        "\n",
        "#@markdown `Diffusion.prepare_noise_schedule()` Linear Schedule\n",
        "#@markdown\n",
        "#@markdown > simply get linspace values from $\\beta_{start}$ to $\\beta_{end}$ with $T$ noise steps\n",
        "def prepare_noise_schedule(self):\n",
        "    #@markdown > If you want to change it to cosine schedule like what OpenAI does, \n",
        "    #@markdown > it should be easy to do it here\n",
        "    return torch.linspace(self.beta_start, self.beta_end, self.noise_steps)\n",
        "\n",
        "#@markdown `Diffusion.noise_images(x, t)`\n",
        "def noise_images(self, x, t):\n",
        "    #@markdown > a) samples $\\epsilon$ from $N(0,I)$\n",
        "    e = torch.randn_like(x)\n",
        "    #@markdown > b) forward process $q(x_t|x_{0})$: calculates $T$ diffusion steps in one shot, that is $x_t=\\sqrt{\\bar\\alpha_t}x_0+\\sqrt{1-\\bar\\alpha_t}\\epsilon$\n",
        "    #@markdown >> super cool and fast\n",
        "    sqrt_alpha_hat = torch.sqrt(self.alpha_hat[t])[:, None, None, None]\n",
        "    sqrt_one_minus_alpha_hat = torch.sqrt(1 - self.alpha_hat[t])[:, None, None, None]\n",
        "    \n",
        "    return sqrt_alpha_hat * x + sqrt_one_minus_alpha_hat * e, e\n",
        "\n",
        "#@markdown `Diffusion.sample_timesteps(n)`\n",
        "#@markdown > sample $T$ from $Uniform(\\{1,...,T\\})$\n",
        "def sample_timesteps(self, n):\n",
        "    return torch.randint(low=1, high=self.noise_steps, size=(n,))\n",
        "\n",
        "\n",
        "\n",
        "Diffusion.prepare_noise_schedule = prepare_noise_schedule\n",
        "Diffusion.noise_images = noise_images\n",
        "Diffusion.sample_timesteps = sample_timesteps\n"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "**Forward process** and **Backward process**:\n",
        "\n",
        "*   a fixed (or predefined) forward diffusion process $q$ of our choosing, that gradually adds Gaussian noise to an image, until you end up with pure noise\n",
        "*   a learned reverse denoising diffusion process $p_{\\theta}$, where a neural network is trained to gradually denoise an image starting from pure noise, until you end up with an actual image.\n",
        "\n",
        "\n",
        "\n",
        "<img src=\"https://huggingface.co/blog/assets/78_annotated-diffusion/diffusion_figure.png\" />\n",
        "\n",
        "Both the **forward** and **backward process** indexed by $t$ happen for some number of finite time steps $T$ (the DDPM authors use $T=1000$). You start with $t=0$ where you sample a real image $x_0$\n",
        " from your data distribution (let's say an image of a cat from ImageNet), and the forward process samples some noise from a Gaussian distribution at each time step $t$, which is added to the image of the previous time step. Given a sufficiently large $T$ and a well behaved schedule for adding noise at each time step, you end up with what is called an [isotropic Gaussian](https://math.stackexchange.com/questions/1991961/gaussian-distribution-is-isotropic) distribution at $t=T$ via a gradual process."
      ],
      "metadata": {
        "id": "UNl0AGIPZufw"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 34,
      "metadata": {
        "cellView": "form",
        "id": "Fq3a1kL3IDY-"
      },
      "outputs": [],
      "source": [
        "#@title Algorithm 2 Sampling\n",
        "\n",
        "def sample(self, model, n):\n",
        "    print(datetime.datetime.now(), f\"sample {n} new images....\")\n",
        "    #@markdown <img src=\"https://huggingface.co/blog/assets/78_annotated-diffusion/sampling.png\" width=\"400px\"/>\n",
        "    #@markdown\n",
        "    #@markdown **Algorithm 2** Sampling\n",
        "    #@markdown \n",
        "    #@markdown ---------------     \n",
        "    model.eval()\n",
        "    with torch.no_grad():\n",
        "        #@markdown 1: $x_T \\sim N(0,I)$\n",
        "        #@markdown\n",
        "        #@markdown > sample from Gaussian distribution\n",
        "        #@markdown\n",
        "        #@markdown > the shape of $x_T$ is exactly the same as images: `(n, 3, img_size, img_size)`\n",
        "        x = torch.randn((n, 3, self.img_size, self.img_size)).to(self.device)\n",
        "        #@markdown 2: for $t=T,...,1$ do\n",
        "        #@markdown\n",
        "        #@markdown that is to execute T times the **backward process** $p(x_{t-1}|x_t)$ \n",
        "        for i in tqdm(reversed(range(1, self.noise_steps)), position=0):\n",
        "            #@markdown > 3: $z \\sim N(0,I)$ if $t>1$, else $z=0$\n",
        "            t = (torch.ones(n) * i).long().to(self.device)\n",
        "            #@markdown > 4: $x_{t-1}=\\frac{1}{\\sqrt{\\alpha_t}}(x_t-\\frac{1-\\alpha_t}{\\sqrt{1-\\alpha_t}}\\epsilon_{\\theta}(x_t,t))$\n",
        "            #@markdown \n",
        "            #@markdown >> `model(x, t)` or `predicted_noise` is acutally the $\\epsilon_{\\theta}(x_t,t)$\n",
        "            #@markdown \n",
        "            #@markdown >> Notice, we dont have $ + \\sigma_tz$ in the end, this is because the author think variance is not needed\n",
        "            predicted_noise = model(x, t)\n",
        "            alpha = self.alpha[t][:, None, None, None]\n",
        "            alpha_hat = self.alpha_hat[t][:, None, None, None]\n",
        "            beta = self.beta[t][:, None, None, None]\n",
        "            if i > 1:\n",
        "                noise = torch.randn_like(x)\n",
        "            else:\n",
        "                noise = torch.zeros_like(x)\n",
        "            \n",
        "            x = 1 / torch.sqrt(alpha) * (x - ((1 - alpha) / (torch.sqrt(1 - alpha_hat))) * predicted_noise) + torch.sqrt(beta) * noise\n",
        "            #@markdown 5: end for\n",
        "    model.train()\n",
        "    #@markdown 6: return $x_0$\n",
        "    x = (x.clamp(-1, 1) + 1) / 2\n",
        "    x = (x * 255).type(torch.uint8)\n",
        "    return x\n",
        "\n",
        "Diffusion.sample = sample"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 45,
      "metadata": {
        "cellView": "form",
        "id": "hSzWYsCAxSDt"
      },
      "outputs": [],
      "source": [
        "#@title Algorithm 1 Training\n",
        "import os\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "from matplotlib import pyplot as plt\n",
        "from tqdm import tqdm\n",
        "from torch import optim\n",
        "import logging\n",
        "from torch.utils.tensorboard import SummaryWriter\n",
        "import IPython\n",
        "import numpy as np\n",
        "\n",
        "def train(args):\n",
        "    setup_logging(args.run_name)\n",
        "    device = args.device\n",
        "    dataloader = get_data(args)\n",
        "    model = UNet().to(device)\n",
        "    optimizer = optim.AdamW(model.parameters(), lr=args.lr)\n",
        "    mse = nn.MSELoss()\n",
        "    diffusion = Diffusion(img_size=args.image_size, device=device)\n",
        "    logger = SummaryWriter(os.path.join(\"runs\", args.run_name))\n",
        "    l = len(dataloader)\t\n",
        "    #@markdown <img src=\"https://huggingface.co/blog/assets/78_annotated-diffusion/training.png\" width=\"400px\"/>\n",
        "    #@markdown \n",
        "    #@markdown **Algorithm 1** Training\n",
        "    #@markdown\n",
        "    #@markdown ---------------\t\n",
        "    #@markdown 1: repeat\n",
        "    for epoch in range(args.epochs):\n",
        "        #IPython.display.clear_output(wait=True)\n",
        "        print(datetime.datetime.now(), f\"epoch: {epoch}:\")\n",
        "        # load data loops and augument a little bit\n",
        "        print(datetime.datetime.now(), \"train\")\n",
        "        pbar = tqdm(dataloader)\n",
        "        losses = []\n",
        "        # for each mini batch\n",
        "        for i, (images, _) in enumerate(pbar):\n",
        "            #@markdown > 2: $x_0 \\sim q(x_0)$\n",
        "            #@markdown >> sample `images` = $x_0$ from dataset\n",
        "            images = images.to(device)\n",
        "            #@markdown > 3: $t \\sim Uniform(\\{1,...,T\\})$ \n",
        "            #@markdown >> sample timesteps `t` for all images at once\n",
        "            t = diffusion.sample_timesteps(images.shape[0]).to(device)\n",
        "            \n",
        "            #@markdown > 4: $\\epsilon \\sim N(0,I)$\n",
        "            #@markdown > \n",
        "            #@markdown > `noise_images(images, t)` does 2 things here:\n",
        "            #@markdown >> \n",
        "            #@markdown >> a) samples $\\epsilon$ from $N(0,I)$\n",
        "            #@markdown >> \n",
        "            #@markdown >> b) **forward process** $q(x_t|x_{0})$: calculates $x_t=\\sqrt{\\bar\\alpha_t}x_0+\\sqrt{1-\\bar\\alpha_t}\\epsilon$\n",
        "            x_t, noise = diffusion.noise_images(images, t)\n",
        "\n",
        "            #@markdown > 5: Take gradient descent step on\n",
        "            #@markdown \n",
        "            #@markdown >> use model (we use UNet() as $\\epsilon_{\\theta}()$) to predict the noise, pass through $(x_t,t)$\n",
        "            #@markdown\n",
        "            #@markdown >> `predicted_noise = `$\\epsilon_{\\theta}(x_t,t)$ = $\\epsilon_{\\theta}(\\sqrt{\\bar\\alpha_t}x_0-\\sqrt{1-\\bar\\alpha_t}\\epsilon,t)$\n",
        "            predicted_noise = model(x_t, t)\n",
        "\n",
        "            #@markdown >> $\\nabla_{\\theta}||\\epsilon - \\epsilon_{\\theta}(\\sqrt{\\bar\\alpha_t}x_0-\\sqrt{1-\\bar\\alpha_t}\\epsilon,t) ||^2 $\n",
        "            #@markdown >> \n",
        "            #@markdown >> Then calculate a MSE loss\n",
        "            loss = mse(noise, predicted_noise)\n",
        "\n",
        "            optimizer.zero_grad()\n",
        "            loss.backward()\n",
        "            optimizer.step()\n",
        "\n",
        "            pbar.set_postfix(MSE=loss.item())\n",
        "            losses.append(loss.item())\n",
        "            logger.add_scalar(\"MSE\", loss.item(), global_step=epoch * l + i)\n",
        "        #@markdown >> average loss for current batch\n",
        "        print(datetime.datetime.now(), 'loss %.8f' % np.array(losses).mean())\n",
        "        \n",
        "        # sample process is extremely time consuming, only execute once every 10 epochs\n",
        "        if not (epoch>0 and epoch%10==0): \n",
        "            continue\n",
        "        #**algorithm 2** is called when predicting\n",
        "        sampled_images = diffusion.sample(model, n=images.shape[0])\n",
        "        save_images(sampled_images, os.path.join(\"results\", args.run_name, f\"{epoch}.jpg\"))\n",
        "        torch.save(model.state_dict(), os.path.join(\"models\", args.run_name, f\"ckpt.pt\"))\n",
        "    #@markdown 6: until converged\n",
        "    #@markdown \n",
        "    #@markdown > *we don't have this step here*\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FGfH36ObPGxx"
      },
      "source": [
        "### Train"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "-k7wG7vVxUPG",
        "cellView": "form",
        "outputId": "ff97d547-c2a3-4135-8255-6a94a661e912"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "/content/ddpm\n",
            "2022-10-06 10:44:22.040257 epoch: 0:\n",
            "2022-10-06 10:44:22.040386 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:27<00:00,  3.22it/s, MSE=0.0383]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 10:50:49.831145 loss 0.05691028\n",
            "2022-10-06 10:50:49.831615 epoch: 1:\n",
            "2022-10-06 10:50:49.831654 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:29<00:00,  3.21it/s, MSE=0.0143]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 10:57:18.984181 loss 0.02095814\n",
            "2022-10-06 10:57:18.984572 epoch: 2:\n",
            "2022-10-06 10:57:18.984610 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:29<00:00,  3.21it/s, MSE=0.0194]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 11:03:48.299746 loss 0.01773479\n",
            "2022-10-06 11:03:48.300119 epoch: 3:\n",
            "2022-10-06 11:03:48.300158 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:28<00:00,  3.21it/s, MSE=0.0123]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 11:10:17.184838 loss 0.01621188\n",
            "2022-10-06 11:10:17.185361 epoch: 4:\n",
            "2022-10-06 11:10:17.185401 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:28<00:00,  3.22it/s, MSE=0.00663]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 11:16:45.904066 loss 0.01514003\n",
            "2022-10-06 11:16:45.904422 epoch: 5:\n",
            "2022-10-06 11:16:45.904494 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:28<00:00,  3.22it/s, MSE=0.0199]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 11:23:14.625640 loss 0.01446826\n",
            "2022-10-06 11:23:14.625973 epoch: 6:\n",
            "2022-10-06 11:23:14.626013 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:28<00:00,  3.22it/s, MSE=0.0197]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 11:29:42.842994 loss 0.01398997\n",
            "2022-10-06 11:29:42.843450 epoch: 7:\n",
            "2022-10-06 11:29:42.843493 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:28<00:00,  3.22it/s, MSE=0.00555]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 11:36:11.024964 loss 0.01326218\n",
            "2022-10-06 11:36:11.025310 epoch: 8:\n",
            "2022-10-06 11:36:11.025344 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:28<00:00,  3.22it/s, MSE=0.0107]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 11:42:39.305668 loss 0.01300647\n",
            "2022-10-06 11:42:39.306844 epoch: 9:\n",
            "2022-10-06 11:42:39.306886 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:28<00:00,  3.22it/s, MSE=0.00496]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 11:49:07.941219 loss 0.01304030\n",
            "2022-10-06 11:49:07.942974 epoch: 10:\n",
            "2022-10-06 11:49:07.943019 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:28<00:00,  3.21it/s, MSE=0.0094]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 11:55:36.783111 loss 0.01261606\n",
            "2022-10-06 11:55:36.783662 sample 8 new images....\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "999it [01:45,  9.50it/s]\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<PIL.Image.Image image mode=RGB size=530x68 at 0x7FE55C2E9110>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 11:57:22.266518 epoch: 11:\n",
            "2022-10-06 11:57:22.266681 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:29<00:00,  3.21it/s, MSE=0.0376]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 12:03:51.605021 loss 0.01214095\n",
            "2022-10-06 12:03:51.605373 epoch: 12:\n",
            "2022-10-06 12:03:51.605413 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:29<00:00,  3.21it/s, MSE=0.0192]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 12:10:20.770181 loss 0.01210485\n",
            "2022-10-06 12:10:20.770581 epoch: 13:\n",
            "2022-10-06 12:10:20.770620 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:29<00:00,  3.21it/s, MSE=0.0181]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 12:16:50.164213 loss 0.01178522\n",
            "2022-10-06 12:16:50.164586 epoch: 14:\n",
            "2022-10-06 12:16:50.164624 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:29<00:00,  3.21it/s, MSE=0.0123]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 12:23:19.612089 loss 0.01153853\n",
            "2022-10-06 12:23:19.612580 epoch: 15:\n",
            "2022-10-06 12:23:19.612616 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:29<00:00,  3.21it/s, MSE=0.0125]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 12:29:49.011124 loss 0.01146643\n",
            "2022-10-06 12:29:49.011559 epoch: 16:\n",
            "2022-10-06 12:29:49.011607 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:32<00:00,  3.18it/s, MSE=0.00546]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 12:36:22.014532 loss 0.01130828\n",
            "2022-10-06 12:36:22.015024 epoch: 17:\n",
            "2022-10-06 12:36:22.015065 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:30<00:00,  3.20it/s, MSE=0.019]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 12:42:52.919892 loss 0.01164330\n",
            "2022-10-06 12:42:52.920323 epoch: 18:\n",
            "2022-10-06 12:42:52.920368 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:31<00:00,  3.19it/s, MSE=0.00925]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 12:49:24.877288 loss 0.01107745\n",
            "2022-10-06 12:49:24.877782 epoch: 19:\n",
            "2022-10-06 12:49:24.877824 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:33<00:00,  3.18it/s, MSE=0.00276]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 12:55:58.209529 loss 0.01097001\n",
            "2022-10-06 12:55:58.210692 epoch: 20:\n",
            "2022-10-06 12:55:58.210735 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:32<00:00,  3.19it/s, MSE=0.00595]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 13:02:30.344954 loss 0.01168136\n",
            "2022-10-06 13:02:30.345345 sample 8 new images....\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "999it [01:44,  9.53it/s]\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<PIL.Image.Image image mode=RGB size=530x68 at 0x7FE55C1B7990>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 13:04:15.535913 epoch: 21:\n",
            "2022-10-06 13:04:15.536091 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:26<00:00,  3.23it/s, MSE=0.00796]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 13:10:41.973931 loss 0.01130582\n",
            "2022-10-06 13:10:41.975394 epoch: 22:\n",
            "2022-10-06 13:10:41.975455 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:25<00:00,  3.25it/s, MSE=0.00564]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 13:17:07.083400 loss 0.01115919\n",
            "2022-10-06 13:17:07.083825 epoch: 23:\n",
            "2022-10-06 13:17:07.083866 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:23<00:00,  3.26it/s, MSE=0.0125]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 13:23:30.864243 loss 0.01111821\n",
            "2022-10-06 13:23:30.864599 epoch: 24:\n",
            "2022-10-06 13:23:30.864636 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:26<00:00,  3.23it/s, MSE=0.0101]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 13:29:57.527863 loss 0.01104765\n",
            "2022-10-06 13:29:57.528347 epoch: 25:\n",
            "2022-10-06 13:29:57.528389 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:31<00:00,  3.19it/s, MSE=0.0151]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 13:36:29.195213 loss 0.01110075\n",
            "2022-10-06 13:36:29.195587 epoch: 26:\n",
            "2022-10-06 13:36:29.195623 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:36<00:00,  3.15it/s, MSE=0.0146]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 13:43:05.935712 loss 0.01091184\n",
            "2022-10-06 13:43:05.936104 epoch: 27:\n",
            "2022-10-06 13:43:05.936150 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:32<00:00,  3.19it/s, MSE=0.00638]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 13:49:38.361304 loss 0.01129729\n",
            "2022-10-06 13:49:38.361763 epoch: 28:\n",
            "2022-10-06 13:49:38.361805 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:32<00:00,  3.19it/s, MSE=0.00386]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 13:56:10.823387 loss 0.01075631\n",
            "2022-10-06 13:56:10.823763 epoch: 29:\n",
            "2022-10-06 13:56:10.823800 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:32<00:00,  3.18it/s, MSE=0.00641]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 14:02:43.599587 loss 0.01085799\n",
            "2022-10-06 14:02:43.599941 epoch: 30:\n",
            "2022-10-06 14:02:43.599994 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|██████████| 1250/1250 [06:32<00:00,  3.19it/s, MSE=0.00585]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 14:09:16.037629 loss 0.01045670\n",
            "2022-10-06 14:09:16.038945 sample 8 new images....\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "999it [01:45,  9.48it/s]\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<PIL.Image.Image image mode=RGB size=530x68 at 0x7FE55C21A510>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2022-10-06 14:11:01.775346 epoch: 31:\n",
            "2022-10-06 14:11:01.777571 train\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            " 92%|█████████▏| 1154/1250 [06:00<00:29,  3.21it/s, MSE=0.00651]"
          ]
        }
      ],
      "source": [
        "#@title execute the training process\n",
        "%cd /content/ddpm\n",
        "\n",
        "#@markdown Input parameters:\n",
        "runName = \"DDPM_Uncondtional\" #@param {\"type\":\"string\"}\n",
        "epochs = 500 #@param {\"type\":\"integer\"}\n",
        "batchSize = 8 #@param {\"type\":\"integer\"}\n",
        "imageSize = 64 #@param {\"type\":\"integer\"}\n",
        "datasetPath = \"/content/ddpm/dataset\" #@param {\"type\":\"string\"}\n",
        "device = \"cuda\" #@param [\"cuda\", \"cpu\"]\n",
        "lr = 3e-4 #@param {\"type\":\"number\"}\n",
        "\n",
        "\n",
        "class Args():\n",
        "    def __init__(self):    \n",
        "        self.run_name = runName\n",
        "        self.epochs = epochs\n",
        "        self.batch_size = batchSize\n",
        "        self.image_size = imageSize\n",
        "        self.dataset_path = datasetPath\n",
        "        self.device = device\n",
        "        self.lr = lr\n",
        "\n",
        "train(Args())\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### References\n",
        "*  Diffusion Models\n",
        " https://github.com/dome272/Diffusion-Models-pytorch/\n",
        "*  What are Diffusion Models?\n",
        " https://lilianweng.github.io/posts/2021-07-11-diffusion-models/\n",
        "*  The Annotated Diffusion Model\n",
        " https://huggingface.co/blog/annotated-diffusion\n",
        "\n",
        "*  Gaussian distribution is isotropic? https://math.stackexchange.com/questions/1991961/gaussian-distribution-is-isotropic\n",
        "\n",
        "*  Denoising Diffusion Probabilistic Models https://arxiv.org/pdf/2006.11239.pdf\n",
        "*  Deep Unsupervised Learning using Nonequilibrium Thermodynamics https://arxiv.org/pdf/1503.03585.pdf"
      ],
      "metadata": {
        "id": "UO695Ri6cIll"
      }
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "collapsed_sections": [
        "XC6PidnDHP3M",
        "UmFHvvM1GDUu",
        "BAK6POjGHCxH"
      ],
      "provenance": [],
      "authorship_tag": "ABX9TyPIQq6oFm39DhUOJI+DTf0M",
      "include_colab_link": true
    },
    "gpuClass": "standard",
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}