nb/ae/AEOneHot.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "c1b184d4",
   "metadata": {},
   "source": [
    "# Autoencoder w/ One Hot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "19d0f047",
   "metadata": {
    "slideshow": {
     "slide_type": ""
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import torch\n",
    "from torch.utils.data import DataLoader, SubsetRandomSampler\n",
    "from torchvision.transforms.functional import to_pil_image\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "f42bf158",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load pallets library from project root\n",
    "import sys, os\n",
    "sys.path.append(os.path.abspath(os.path.join(os.getcwd(), '..', '..')))\n",
    "\n",
    "from pallets import (\n",
    "    images as I,\n",
    "    datasets as DS,\n",
    "    models as M,\n",
    "    logging as L\n",
    ")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fdb35bdb",
   "metadata": {},
   "source": [
    "# Settings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "93b89466",
   "metadata": {},
   "outputs": [],
   "source": [
    "SAVE_NAME = 'ae.onehot'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "7887e5ae",
   "metadata": {},
   "outputs": [],
   "source": [
    "USE_GPU = True\n",
    "LOG_LEVEL = 'INFO'\n",
    "LABELS_FILE = 'pallets_labels.json'\n",
    "\n",
    "TEST_SIZE = 1000\n",
    "EPOCHS = 20\n",
    "LR = 1e-03\n",
    "BATCH_SIZE = 32\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "189cf27e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Logger pallets (INFO)>"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# To GPU, or not to GPU\n",
    "device = M.get_device(require_gpu=USE_GPU)\n",
    "\n",
    "# Logging\n",
    "L.init_logger(level=LOG_LEVEL, notebook=True)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "adfd5648",
   "metadata": {},
   "source": [
    "# Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "864aacb2",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO | loading punk images\n",
      "INFO | - image 0\n",
      "INFO | - image 1000\n",
      "INFO | - image 2000\n",
      "INFO | - image 3000\n",
      "INFO | - image 4000\n",
      "INFO | - image 5000\n",
      "INFO | - image 6000\n",
      "INFO | - image 7000\n",
      "INFO | - image 8000\n",
      "INFO | - image 9000\n",
      "INFO | punk images complete\n",
      "INFO | loading punk labels: pallets_labels.json\n"
     ]
    }
   ],
   "source": [
    "all_colors = I.get_punk_colors()\n",
    "mapper = DS.ColorOneHotMapper(all_colors)\n",
    "dataset = DS.OneHotCPunksDataset(mapper, labels_file=LABELS_FILE, test_size=TEST_SIZE)\n",
    "\n",
    "# Uncomment to put entire dataset in GPU RAM\n",
    "# dataset = DS.FastOneHotCPunksDataset(\n",
    "#     device, mapper, labels_file=LABELS_FILE, test_size=TEST_SIZE\n",
    "# )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "f3afb633",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_sampler = SubsetRandomSampler(dataset.train_idx)\n",
    "test_sampler = SubsetRandomSampler(dataset.test_idx)\n",
    "\n",
    "train_loader = DataLoader(dataset, batch_size=BATCH_SIZE, sampler=train_sampler)\n",
    "test_loader = DataLoader(dataset, batch_size=BATCH_SIZE, sampler=test_sampler)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8801d898",
   "metadata": {},
   "source": [
    "# Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "93fd96d3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO | model: pallets.models.ae.AEOneHot\n",
      "INFO | criterion: torch.nn.modules.loss.CrossEntropyLoss\n",
      "INFO | learn rate: 0.001\n",
      "INFO | epochs: 20\n",
      "INFO | epoch 1 (  0%) loss: 2.702146\n",
      "INFO | epoch 1 ( 35%) loss: 0.641268\n",
      "INFO | epoch 1 ( 70%) loss: 0.429861\n",
      "INFO | epoch 1 (100%) loss: 0.343722\n",
      "INFO | epoch 1 (test) loss: 0.025397\n",
      "INFO | epoch 2 (  0%) loss: 0.093849\n",
      "INFO | epoch 2 ( 35%) loss: 0.083924\n",
      "INFO | epoch 2 ( 70%) loss: 0.070849\n",
      "INFO | epoch 2 (100%) loss: 0.062984\n",
      "INFO | epoch 2 (test) loss: 0.010011\n",
      "INFO | epoch 3 (  0%) loss: 0.043845\n",
      "INFO | epoch 3 ( 35%) loss: 0.035484\n",
      "INFO | epoch 3 ( 70%) loss: 0.031486\n",
      "INFO | epoch 3 (100%) loss: 0.029306\n",
      "INFO | epoch 3 (test) loss: 0.005895\n",
      "INFO | epoch 4 (  0%) loss: 0.023858\n",
      "INFO | epoch 4 ( 35%) loss: 0.019508\n",
      "INFO | epoch 4 ( 70%) loss: 0.018061\n",
      "INFO | epoch 4 (100%) loss: 0.017288\n",
      "INFO | epoch 4 (test) loss: 0.003962\n",
      "INFO | epoch 5 (  0%) loss: 0.013934\n",
      "INFO | epoch 5 ( 35%) loss: 0.012579\n",
      "INFO | epoch 5 ( 70%) loss: 0.011955\n",
      "INFO | epoch 5 (100%) loss: 0.011610\n",
      "INFO | epoch 5 (test) loss: 0.003141\n",
      "INFO | epoch 6 (  0%) loss: 0.009598\n",
      "INFO | epoch 6 ( 35%) loss: 0.008874\n",
      "INFO | epoch 6 ( 70%) loss: 0.008562\n",
      "INFO | epoch 6 (100%) loss: 0.008417\n",
      "INFO | epoch 6 (test) loss: 0.002257\n",
      "INFO | epoch 7 (  0%) loss: 0.006814\n",
      "INFO | epoch 7 ( 35%) loss: 0.006343\n",
      "INFO | epoch 7 ( 70%) loss: 0.006192\n",
      "INFO | epoch 7 (100%) loss: 0.006024\n",
      "INFO | epoch 7 (test) loss: 0.001750\n",
      "INFO | epoch 8 (  0%) loss: 0.004836\n",
      "INFO | epoch 8 ( 35%) loss: 0.004854\n",
      "INFO | epoch 8 ( 70%) loss: 0.004626\n",
      "INFO | epoch 8 (100%) loss: 0.004511\n",
      "INFO | epoch 8 (test) loss: 0.001510\n",
      "INFO | epoch 9 (  0%) loss: 0.004235\n",
      "INFO | epoch 9 ( 35%) loss: 0.003408\n",
      "INFO | epoch 9 ( 70%) loss: 0.003480\n",
      "INFO | epoch 9 (100%) loss: 0.003425\n",
      "INFO | epoch 9 (test) loss: 0.001246\n",
      "INFO | epoch 10 (  0%) loss: 0.003116\n",
      "INFO | epoch 10 ( 35%) loss: 0.002802\n",
      "INFO | epoch 10 ( 70%) loss: 0.002789\n",
      "INFO | epoch 10 (100%) loss: 0.002783\n",
      "INFO | epoch 10 (test) loss: 0.001159\n",
      "INFO | epoch 11 (  0%) loss: 0.002058\n",
      "INFO | epoch 11 ( 35%) loss: 0.002389\n",
      "INFO | epoch 11 ( 70%) loss: 0.002290\n",
      "INFO | epoch 11 (100%) loss: 0.002274\n",
      "INFO | epoch 11 (test) loss: 0.001100\n",
      "INFO | epoch 12 (  0%) loss: 0.001880\n",
      "INFO | epoch 12 ( 35%) loss: 0.002080\n",
      "INFO | epoch 12 ( 70%) loss: 0.001958\n",
      "INFO | epoch 12 (100%) loss: 0.001918\n",
      "INFO | epoch 12 (test) loss: 0.000929\n",
      "INFO | epoch 13 (  0%) loss: 0.002213\n",
      "INFO | epoch 13 ( 35%) loss: 0.001582\n",
      "INFO | epoch 13 ( 70%) loss: 0.001562\n",
      "INFO | epoch 13 (100%) loss: 0.001586\n",
      "INFO | epoch 13 (test) loss: 0.001022\n",
      "INFO | epoch 14 (  0%) loss: 0.001627\n",
      "INFO | epoch 14 ( 35%) loss: 0.001368\n",
      "INFO | epoch 14 ( 70%) loss: 0.001350\n",
      "INFO | epoch 14 (100%) loss: 0.001399\n",
      "INFO | epoch 14 (test) loss: 0.000886\n",
      "INFO | epoch 15 (  0%) loss: 0.000836\n",
      "INFO | epoch 15 ( 35%) loss: 0.001202\n",
      "INFO | epoch 15 ( 70%) loss: 0.001364\n",
      "INFO | epoch 15 (100%) loss: 0.001427\n",
      "INFO | epoch 15 (test) loss: 0.000951\n",
      "INFO | epoch 16 (  0%) loss: 0.000804\n",
      "INFO | epoch 16 ( 35%) loss: 0.001573\n",
      "INFO | epoch 16 ( 70%) loss: 0.001353\n",
      "INFO | epoch 16 (100%) loss: 0.001393\n",
      "INFO | epoch 16 (test) loss: 0.000866\n",
      "INFO | epoch 17 (  0%) loss: 0.001031\n",
      "INFO | epoch 17 ( 35%) loss: 0.001018\n",
      "INFO | epoch 17 ( 70%) loss: 0.001009\n",
      "INFO | epoch 17 (100%) loss: 0.001065\n",
      "INFO | epoch 17 (test) loss: 0.001077\n",
      "INFO | epoch 18 (  0%) loss: 0.000959\n",
      "INFO | epoch 18 ( 35%) loss: 0.001817\n",
      "INFO | epoch 18 ( 70%) loss: 0.001774\n",
      "INFO | epoch 18 (100%) loss: 0.001723\n",
      "INFO | epoch 18 (test) loss: 0.000949\n",
      "INFO | epoch 19 (  0%) loss: 0.001045\n",
      "INFO | epoch 19 ( 35%) loss: 0.001279\n",
      "INFO | epoch 19 ( 70%) loss: 0.001074\n",
      "INFO | epoch 19 (100%) loss: 0.001015\n",
      "INFO | epoch 19 (test) loss: 0.000722\n",
      "INFO | epoch 20 (  0%) loss: 0.000374\n",
      "INFO | epoch 20 ( 35%) loss: 0.000641\n",
      "INFO | epoch 20 ( 70%) loss: 0.000590\n",
      "INFO | epoch 20 (100%) loss: 0.000581\n",
      "INFO | epoch 20 (test) loss: 0.000705\n",
      "INFO | model blob saved to /home/jmsdnns/ML/pallets/saved/ae.onehot.pkl\n",
      "INFO | training losses saved to /home/jmsdnns/ML/pallets/saved/ae.onehot.json\n"
     ]
    }
   ],
   "source": [
    "model = M.ae.AEOneHot()\n",
    "criterion = torch.nn.CrossEntropyLoss()\n",
    "\n",
    "train_losses, test_losses = M.ae.train(\n",
    "    device, model, criterion, train_loader, test_loader,\n",
    "    learn_rate=LR, epochs=EPOCHS\n",
    ")\n",
    "\n",
    "M.save(SAVE_NAME, model, train_losses, test_losses)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6bfcbdcd",
   "metadata": {},
   "source": [
    "# Results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "84593a4d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAGFCAYAAAASI+9IAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjguMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/SrBM8AAAACXBIWXMAAA9hAAAPYQGoP6dpAAAGj0lEQVR4nO3dIY4cRxSA4e1oTzBgUUiIBxvlBCGWZR6WS+UM4ZZlEr5SUPDkCAv2ChUS/VoYd8+6enq/j5f6aWztryKvljHGuAOAu7u7H2YPAMB+iAIAEQUAIgoARBQAiCgAEFEAIKIAQEQBgIgCABEFACIKAEQUAIgoABBRACCiAEBEAYCIAgARBQAiCgBEFACIKAAQUQAgogBARAGAiAIAEQUAIgoARBQAiCgAEFEAIKIAQEQBgIgCABEFACIKAEQUAIgoABBRACCiAEBEAYCIAgARBQAiCgBEFACIKAAQUQAgogBARAGAiAIAEQUAIgoARBQAiCgAkPvZA8BLy7LMHmGVMcbsEeAq3BQAiCgAEFEAIKIAQEQBgIgCABEFACIKAEQUAIgoABBRACCiAEBEAYCIAgARBQDiPYWdutV3BbZ69/7D7BFWeav/Xt6ROB43BQAiCgBEFACIKAAQUQAgogBARAGAiAIAEQUAIgoARBQAiCgAEFEAIKIAQKzOfkVb1inf6grp2S5/fF199vzr+t/8rf57bfk/bu32PrkpABBRACCiAEBEAYCIAgARBQAiCgBEFACIKAAQUQAgogBARAGAiAIAEQUAYnU2h7Jl/TXgpgDAC6IAQEQBgIgCABEFACIKAEQUAIgoABBRACCiAEBEAYCIAgARBQAiCgBEFACI9xRe0Rhj9dllWXz7hr695btv+dvsj5sCABEFACIKAEQUAIgoABBRACCiAEBEAYCIAgARBQAiCgBEFACIKAAQUQAgy7D7dpe2rpCe6d37D7NH4Dv55++vq8/607NPbgoARBQAiCgAEFEAIKIAQEQBgIgCABEFACIKAEQUAIgoABBRACCiAEBEAYCIAgC5nz0Ar+Ppcpn27Yfzedq3gW3cFACIKAAQUQAgogBARAGAiAIAEQUAIgoARBQAiCgAEFEAIKIAQEQBgIgCALE6m6ububabb7dl1fkY44qTfJvn5+fVZ0+n0xUnORY3BQAiCgBEFACIKAAQUQAgogBARAGAiAIAEQUAIgoARBQAiCgAEFEAIKIAQEQBgHhPgav788uXad/+5ePHad/eYuZvBi+5KQAQUQAgogBARAGAiAIAEQUAIgoARBQAiCgAEFEAIKIAQEQBgIgCABEFALKMMcbsIY5qWZbVZ58ulytO8n09nM+rz/71+fOmb/+04dszzfzNfv70afXZmX8+np+fV589nU5XnORY3BQAiCgAEFEAIKIAQEQBgIgCABEFACIKAEQUAIgoABBRACCiAEBEAYCIAgC5nz0Ax3PLa79n8ZuxF24KAEQUAIgoABBRACCiAEBEAYCIAgARBQAiCgBEFACIKAAQUQAgogBARAGAiAIA8Z7CTj2cz5vO288PrOGmAEBEAYCIAgARBQAiCgBEFACIKAAQUQAgogBARAGAiAIAEQUAIgoARBQAiNXZr2iMsfrssixXnIT/Y+u68i2sOmcv3BQAiCgAEFEAIKIAQEQBgIgCABEFACIKAEQUAIgoABBRACCiAEBEAYCIAgARBQDiPQX4jzcNbsvpdJo9wiG5KQAQUQAgogBARAGAiAIAEQUAIgoARBQAiCgAEFEAIKIAQEQBgIgCABEFAGJ1NuzAw/k8e4RVlmWZPcIqY4zZI+yWmwIAEQUAIgoARBQAiCgAEFEAIKIAQEQBgIgCABEFACIKAEQUAIgoABBRACBWZ3Mot7qC+vHxcfYIN+fH339bfXbryu8jr952UwAgogBARAGAiAIAEQUAIgoARBQAiCgAEFEAIKIAQEQBgIgCABEFACIKAEQUAIj3FDiUp8tl9gir3Oo7ELfqyO8hbOWmAEBEAYCIAgARBQAiCgBEFACIKAAQUQAgogBARAGAiAIAEQUAIgoARBQAiNXZsANbVn5vXbttjTQvuSkAEFEAIKIAQEQBgIgCABEFACIKAEQUAIgoABBRACCiAEBEAYCIAgARBQAiCgDEewrsyta3Ad4i7yFwTW4KAEQUAIgoABBRACCiAEBEAYCIAgARBQAiCgBEFACIKAAQUQAgogBARAGAWJ3NoVgjDdu4KQAQUQAgogBARAGAiAIAEQUAIgoARBQAiCgAEFEAIKIAQEQBgIgCABEFACIKAEQUAIgoABBRACCiAEBEAYCIAgARBQAiCgBEFACIKAAQUQAgogBARAGAiAIAEQUAsowxxuwhANgHNwUAIgoARBQAiCgAEFEAIKIAQEQBgIgCABEFAPIv8q6KeAGHxAwAAAAASUVORK5CYII=",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "punk, _ = next(iter(train_loader))\n",
    "reconstructed = model(punk.to(device))\n",
    "reconstructed = reconstructed.detach().cpu()\n",
    "some_image = reconstructed[0]\n",
    "\n",
    "some_image = DS.one_hot_to_rgba(some_image, mapper)\n",
    "\n",
    "pil_image = to_pil_image(some_image.clone().detach())\n",
    "plt.imshow(pil_image)\n",
    "plt.axis('off')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "fd3506ac",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.title(\"Train & Test loss\")\n",
    "plt.plot(train_losses, label='train loss')\n",
    "plt.plot(test_losses, label='test loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}