TinyQuadMNIST.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "mount_file_id": "1RslENvDNQtzjfUYe_Xweq_o9z0cd6BKY",
      "authorship_tag": "ABX9TyN/Y0w6ih7nO997cPJNFtGx",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/diro5t/deep_quadric_learning/blob/main/TinyQuadMNIST.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "(C) Dirk Roeckmann 2023\n",
        "\n",
        "Simple tinygrad MNIST ~25K parameter classifier with quadric layers 784-16-10,batch size 64 modeled after the tinygrad quickstart.md\n",
        "\n",
        "NOTE: The goal is NOT to find the minimum model or a certain target accuracy, but to compare it with a ~100K linear model after the same number of epochs (see TinyLinMNIST.ipynb).\n",
        "\n",
        "Just by introducing quadric layers on a high level, the model size shrank to a quarter. This stuff has huge potential to significantly reduce model size in all kinds of applications.\n",
        "\n",
        "More about quadric models here:\n",
        "\n",
        "https://github.com/diro5t/deep_quadric_learning\n",
        "\n",
        "This paradigm is based on my research here:\n",
        "\n",
        "https://scholar.google.com/citations?hl=en&user=hamuGh4AAAAJ"
      ],
      "metadata": {
        "id": "grsZyPZ5VvFE"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Install fresh tinygrad"
      ],
      "metadata": {
        "id": "K2cj-VLY_Xvm"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "%cd /content"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0S3mazdk6DMy",
        "outputId": "c18baf24-9a60-43f5-b813-9472eb4509fd"
      },
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "/content\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import shutil\n",
        "def handler(func, path, exc_info):\n",
        "    print(exc_info)\n",
        "shutil.rmtree('tinygrad', onerror=handler)"
      ],
      "metadata": {
        "id": "PEDAFyxQ5RP4",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "57206b6a-a695-4e38-a07f-3c4e19182393"
      },
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(<class 'FileNotFoundError'>, FileNotFoundError(2, 'No such file or directory'), <traceback object at 0x784457f24640>)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "!git clone https://github.com/geohot/tinygrad.git"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8m3Bn7mE0yfe",
        "outputId": "93cbebcb-552a-4915-aeaf-1c5e854390b2"
      },
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Cloning into 'tinygrad'...\n",
            "remote: Enumerating objects: 14774, done.\u001b[K\n",
            "remote: Counting objects: 100% (4284/4284), done.\u001b[K\n",
            "remote: Compressing objects: 100% (450/450), done.\u001b[K\n",
            "remote: Total 14774 (delta 3951), reused 3871 (delta 3830), pack-reused 10490\u001b[K\n",
            "Receiving objects: 100% (14774/14774), 18.43 MiB | 21.64 MiB/s, done.\n",
            "Resolving deltas: 100% (10541/10541), done.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "%cd tinygrad"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "u4Lt8PJV1jA5",
        "outputId": "201f1fba-1c8d-42de-8258-ba2920f28a20"
      },
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "/content/tinygrad\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "!python3 -m pip install -e ."
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "xRqwqTsE1mhM",
        "outputId": "a1504fbf-3b41-4084-a1dd-da3ff9cab2c4"
      },
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Obtaining file:///content/tinygrad\n",
            "  Preparing metadata (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.10/dist-packages (from tinygrad==0.6.0) (1.22.4)\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.10/dist-packages (from tinygrad==0.6.0) (2.27.1)\n",
            "Requirement already satisfied: pillow in /usr/local/lib/python3.10/dist-packages (from tinygrad==0.6.0) (8.4.0)\n",
            "Requirement already satisfied: tqdm in /usr/local/lib/python3.10/dist-packages (from tinygrad==0.6.0) (4.65.0)\n",
            "Requirement already satisfied: networkx in /usr/local/lib/python3.10/dist-packages (from tinygrad==0.6.0) (3.1)\n",
            "Collecting pyopencl (from tinygrad==0.6.0)\n",
            "  Downloading pyopencl-2023.1.1-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (919 kB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m919.2/919.2 kB\u001b[0m \u001b[31m24.5 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hRequirement already satisfied: PyYAML in /usr/local/lib/python3.10/dist-packages (from tinygrad==0.6.0) (6.0)\n",
            "Collecting pytools>=2021.2.7 (from pyopencl->tinygrad==0.6.0)\n",
            "  Downloading pytools-2023.1-py2.py3-none-any.whl (70 kB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m70.4/70.4 kB\u001b[0m \u001b[31m7.7 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hRequirement already satisfied: platformdirs>=2.2.0 in /usr/local/lib/python3.10/dist-packages (from pyopencl->tinygrad==0.6.0) (3.8.1)\n",
            "Requirement already satisfied: urllib3<1.27,>=1.21.1 in /usr/local/lib/python3.10/dist-packages (from requests->tinygrad==0.6.0) (1.26.16)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.10/dist-packages (from requests->tinygrad==0.6.0) (2023.5.7)\n",
            "Requirement already satisfied: charset-normalizer~=2.0.0 in /usr/local/lib/python3.10/dist-packages (from requests->tinygrad==0.6.0) (2.0.12)\n",
            "Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.10/dist-packages (from requests->tinygrad==0.6.0) (3.4)\n",
            "Requirement already satisfied: typing-extensions>=4.0 in /usr/local/lib/python3.10/dist-packages (from pytools>=2021.2.7->pyopencl->tinygrad==0.6.0) (4.7.1)\n",
            "Installing collected packages: pytools, pyopencl, tinygrad\n",
            "  Running setup.py develop for tinygrad\n",
            "Successfully installed pyopencl-2023.1.1 pytools-2023.1 tinygrad-0.6.0\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "!pip show tinygrad"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "PMIp_hXT6olF",
        "outputId": "afcd1ba7-a543-4438-a422-61c7eba1910d"
      },
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Name: tinygrad\n",
            "Version: 0.6.0\n",
            "Summary: You like pytorch? You like micrograd? You love tinygrad! <3\n",
            "Home-page: \n",
            "Author: George Hotz\n",
            "Author-email: \n",
            "License: MIT\n",
            "Location: /content/tinygrad\n",
            "Editable project location: /content/tinygrad\n",
            "Requires: networkx, numpy, pillow, pyopencl, PyYAML, requests, tqdm\n",
            "Required-by: \n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Train and test the model"
      ],
      "metadata": {
        "id": "4nAh4c0A_h7I"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# fetch mnist data\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "from extra.datasets import fetch_mnist\n",
        "X_train, Y_train, X_test, Y_test = fetch_mnist()\n",
        "X_train, X_test = X_train/255, X_test/255"
      ],
      "metadata": {
        "id": "MJO2ShAF_Gxh"
      },
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# samples\n",
        "figure = plt.figure()\n",
        "for i in range(1, 61):\n",
        "    plt.subplot(10, 6, i)\n",
        "    plt.axis('off')\n",
        "    plt.imshow(np.reshape(X_train[i], (28, 28)), cmap='gray_r')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 406
        },
        "id": "3yjiZ0b-0Ok7",
        "outputId": "22c9dd1d-7a33-408d-9783-065a8b526d7a"
      },
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 60 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# quadric layer\n",
        "import math\n",
        "class Quadric:\n",
        "  def __init__(self, in_features, out_features, bias=True):\n",
        "    self.qweight = Tensor.kaiming_uniform(out_features, in_features, a=math.sqrt(5))\n",
        "    self.lweight = Tensor.kaiming_uniform(out_features, in_features, a=math.sqrt(5))\n",
        "    bound = 1 / math.sqrt(self.lweight.shape[1])\n",
        "    self.bias = Tensor.uniform(out_features, low=-bound, high=bound) if bias else None\n",
        "\n",
        "  def __call__(self, x):\n",
        "    q = x.mul(x).linear(self.qweight.transpose())\n",
        "    lb = x.linear(self.lweight.transpose(), self.bias)\n",
        "    return q.add(lb)"
      ],
      "metadata": {
        "id": "_f33fdrIrjFZ"
      },
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# quadric model\n",
        "from tinygrad.tensor import Tensor\n",
        "from tinygrad.nn import Linear\n",
        "from tinygrad.state import get_parameters\n",
        "\n",
        "class TinyQuadMnist:\n",
        "  def __init__(self):\n",
        "    self.q1 = Quadric(784, 16, bias=False)\n",
        "    self.q2 = Quadric(16, 10, bias=False)\n",
        "\n",
        "  def parameters(self):\n",
        "    return get_parameters(self)\n",
        "\n",
        "  def __call__(self, x):\n",
        "    x = self.q1(x)\n",
        "    x = x.leakyrelu()\n",
        "    x = self.q2(x)\n",
        "    return x.log_softmax()\n",
        "\n",
        "model = TinyQuadMnist()\n",
        "model.parameters()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "hzN4OpARl-Jk",
        "outputId": "7e781dcf-a64d-49b9-9762-9e0facd8e53a"
      },
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[<Tensor <LB (16, 784) dtypes.float op=BinaryOps.ADD st=ShapeTracker(shape=(16, 784), views=[View(shape=(16, 784), strides=(784, 1), offset=0, mask=None, contiguous=True, shape_strides=((12544, 1),))])> on CPU with grad None>,\n",
              " <Tensor <LB (16, 784) dtypes.float op=BinaryOps.ADD st=ShapeTracker(shape=(16, 784), views=[View(shape=(16, 784), strides=(784, 1), offset=0, mask=None, contiguous=True, shape_strides=((12544, 1),))])> on CPU with grad None>,\n",
              " <Tensor <LB (10, 16) dtypes.float op=BinaryOps.ADD st=ShapeTracker(shape=(10, 16), views=[View(shape=(10, 16), strides=(16, 1), offset=0, mask=None, contiguous=True, shape_strides=((160, 1),))])> on CPU with grad None>,\n",
              " <Tensor <LB (10, 16) dtypes.float op=BinaryOps.ADD st=ShapeTracker(shape=(10, 16), views=[View(shape=(10, 16), strides=(16, 1), offset=0, mask=None, contiguous=True, shape_strides=((160, 1),))])> on CPU with grad None>]"
            ]
          },
          "metadata": {},
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# training\n",
        "Tensor.training = True\n",
        "from extra.training import sparse_categorical_crossentropy\n",
        "from tinygrad.nn.optim import SGD\n",
        "opt = SGD(model.parameters(), lr=0.003)\n",
        "steps = 2000\n",
        "for step in range(steps):\n",
        "  # random sample a batch\n",
        "  samp = np.random.randint(0, X_train.shape[0], size=(64))\n",
        "  batch = Tensor(X_train[samp], requires_grad=False)\n",
        "  # get the corresponding labels\n",
        "  labels = Y_train[samp]\n",
        "  # forward pass\n",
        "  out = model(batch)\n",
        "  # compute loss\n",
        "  loss = sparse_categorical_crossentropy(out, labels)\n",
        "  # zero gradients\n",
        "  opt.zero_grad()\n",
        "  # backward pass\n",
        "  loss.backward()\n",
        "  # update parameters\n",
        "  opt.step()\n",
        "  # calculate accuracy\n",
        "  pred = np.argmax(out.numpy(), axis=-1)\n",
        "  acc = (pred == labels).mean()\n",
        "  if step % 100 == 0:\n",
        "    print(f\"Step {step+1} | Loss: {loss.numpy()} | Accuracy: {acc}\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "OM6ecU6OnzuW",
        "outputId": "df90bb7b-ca6e-4a5f-812a-beebbf8376ec"
      },
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Step 1 | Loss: 2.305783987045288 | Accuracy: 0.109375\n",
            "Step 101 | Loss: 2.249108076095581 | Accuracy: 0.234375\n",
            "Step 201 | Loss: 2.0208401679992676 | Accuracy: 0.328125\n",
            "Step 301 | Loss: 1.6392288208007812 | Accuracy: 0.625\n",
            "Step 401 | Loss: 1.3168811798095703 | Accuracy: 0.65625\n",
            "Step 501 | Loss: 0.8300754427909851 | Accuracy: 0.796875\n",
            "Step 601 | Loss: 0.7402499914169312 | Accuracy: 0.796875\n",
            "Step 701 | Loss: 0.7102689743041992 | Accuracy: 0.796875\n",
            "Step 801 | Loss: 0.6042525768280029 | Accuracy: 0.796875\n",
            "Step 901 | Loss: 0.67066490650177 | Accuracy: 0.8125\n",
            "Step 1001 | Loss: 0.48179396986961365 | Accuracy: 0.84375\n",
            "Step 1101 | Loss: 0.5482581853866577 | Accuracy: 0.828125\n",
            "Step 1201 | Loss: 0.4481702744960785 | Accuracy: 0.859375\n",
            "Step 1301 | Loss: 0.5022432804107666 | Accuracy: 0.921875\n",
            "Step 1401 | Loss: 0.3423548638820648 | Accuracy: 0.90625\n",
            "Step 1501 | Loss: 0.6146490573883057 | Accuracy: 0.765625\n",
            "Step 1601 | Loss: 0.5154604315757751 | Accuracy: 0.84375\n",
            "Step 1701 | Loss: 0.5676926970481873 | Accuracy: 0.84375\n",
            "Step 1801 | Loss: 0.30041661858558655 | Accuracy: 0.90625\n",
            "Step 1901 | Loss: 0.32439661026000977 | Accuracy: 0.921875\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# evaluation\n",
        "import time\n",
        "Tensor.training = False\n",
        "st = time.perf_counter()\n",
        "avg_acc = 0\n",
        "for step in range(steps):\n",
        "  # random sample a batch\n",
        "  samp = np.random.randint(0, X_test.shape[0], size=(64))\n",
        "  batch = Tensor(X_test[samp], requires_grad=False)\n",
        "  # get the corresponding labels\n",
        "  labels = Y_test[samp]\n",
        "  # forward pass\n",
        "  out = model(batch)\n",
        "  # calculate accuracy\n",
        "  pred = np.argmax(out.numpy(), axis=-1)\n",
        "  avg_acc += (pred == labels).mean()\n",
        "print(f\"Test Accuracy: {avg_acc / steps}\")\n",
        "print(f\"Time: {time.perf_counter() - st}\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "qNZelqLpp7RP",
        "outputId": "8d32853f-70ce-4e1a-9a38-75dbaa2ee488"
      },
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Test Accuracy: 0.8968984375\n",
            "Time: 17.542694795000017\n"
          ]
        }
      ]
    }
  ]
}