Weight Pruning.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "WMb2Tt6lCOM-"
   },
   "source": [
    "# Default Pytorch pruning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torch.optim as optim\n",
    "from torchvision import datasets, transforms"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import fastprogress\n",
    "import fastai\n",
    "from fastai import *\n",
    "from fastai.text import *\n",
    "from fastai.vision import *"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Disable the progress bar"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "fastprogress.fastprogress.NO_BAR = True\n",
    "master_bar, progress_bar = fastprogress.force_console_behavior()\n",
    "fastai.basic_train.master_bar, fastai.basic_train.progress_bar = master_bar, progress_bar\n",
    "fastai.basic_data.master_bar, fastai.basic_data.progress_bar = master_bar, progress_bar\n",
    "dataclass.master_bar, dataclass.progress_bar = master_bar, progress_bar\n",
    "fastai.text.master_bar, fastai.text.progress_bar = master_bar, progress_bar\n",
    "fastai.text.data.master_bar, fastai.text.data.progress_bar = master_bar, progress_bar\n",
    "fastai.core.master_bar, fastai.core.progress_bar = master_bar, progress_bar"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "bs = 64\n",
    "epochs = 3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "Jdj0MFnYQN28"
   },
   "outputs": [],
   "source": [
    "class Net(nn.Module):\n",
    "    def __init__(self, mnist=True):\n",
    "        super().__init__()\n",
    "          \n",
    "        self.conv1 = nn.Conv2d(3, 20, 5, 1)\n",
    "        self.conv2 = nn.Conv2d(20, 50, 5, 1)\n",
    "        self.fc1 = nn.Linear(4*4*50, 500)\n",
    "        self.fc2 = nn.Linear(500, 10)\n",
    "    \n",
    "    def forward(self, x):\n",
    "\n",
    "        x = F.relu(self.conv1(x))\n",
    "        x = F.max_pool2d(x, 2, 2)\n",
    "        x = F.relu(self.conv2(x))\n",
    "        x = F.max_pool2d(x, 2, 2)\n",
    "        x = x.view(-1, 4*4*50)\n",
    "        x = F.relu(self.fc1(x))\n",
    "        x = self.fc2(x)\n",
    "        return F.log_softmax(x, dim=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "path = untar_data(URLs.MNIST)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = (ImageList.from_folder(path)\n",
    "                .split_by_folder(train='training', valid='testing')\n",
    "                .label_from_folder()\n",
    "                .databunch()\n",
    "                .normalize())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x1440 with 25 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "data.show_batch()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "PC-XcAqyb60B"
   },
   "source": [
    "## Compute Target Sparsity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Sparsifier():\n",
    "    def __init__(self, meta):\n",
    "        self.meta = meta\n",
    "    \n",
    "    def _compute_sparsity(self, meta):\n",
    "        return meta['final_sparsity'] + (meta['initial_sparsity'] - meta['final_sparsity'])*(1 - (meta['current_step'] - meta['starting_step'])/((meta['ending_step'] - meta['starting_step'])))**3\n",
    "\n",
    "    def prune(self, model, meta):\n",
    "\n",
    "        binary_masks = {}\n",
    "\n",
    "        sparsity = self._compute_sparsity(meta)\n",
    "    \n",
    "        for k, m in enumerate(model.modules()):\n",
    "\n",
    "            if not isinstance(m, Net):\n",
    "\n",
    "                weight = m.weight.data.view(-1).clone().abs()\n",
    "                y, i = torch.sort(weight)\n",
    "\n",
    "                spars_index = int(weight.shape[0]*sparsity/100)\n",
    "                threshold = y[spars_index]\n",
    "                mask = weight.gt(threshold).float().cuda()\n",
    "                \n",
    "                mask = mask.view(m.weight.data.shape)\n",
    "                \n",
    "                binary_masks[k] = mask\n",
    "                m.weight.data.mul_(mask)\n",
    "\n",
    "        return binary_masks    \n",
    "        \n",
    "\n",
    "    def applyBinaryMasks(self, model, masks):\n",
    "\n",
    "        for k, m in enumerate(model.modules()):\n",
    "\n",
    "            if not isinstance(m, Net):\n",
    "                mask = masks[k]\n",
    "                m.weight.data.mul_(mask)\n",
    "    \n",
    "        return model\n",
    "\n",
    "@dataclass\n",
    "class SparsifyCallback(Callback):\n",
    "    learn:Learner\n",
    "        \n",
    "    def __init__(self, meta):\n",
    "        self.meta = meta\n",
    "        self.sparsifier = Sparsifier(self.meta)\n",
    "        self.binary_masks = None\n",
    "        \n",
    "    def on_epoch_end(self, **kwargs):\n",
    "        print(f'Sparsity: {self.sparsifier._compute_sparsity(self.meta)}')\n",
    "        \n",
    "    def on_batch_begin(self, **kwargs):\n",
    "\n",
    "        if (self.meta['current_step'] - self.meta['starting_step']) % self.meta['span'] == 0 and self.meta['current_step'] > self.meta['starting_step'] and self.meta['current_step'] < self.meta['ending_step']:\n",
    "            self.binary_masks = self.sparsifier.prune(learn.model, self.meta)\n",
    "            \n",
    "        if self.binary_masks:\n",
    "            learn.model = self.sparsifier.applyBinaryMasks(learn.model, self.binary_masks)   \n",
    "            \n",
    "        if self.meta['current_step'] < self.meta['ending_step']:\n",
    "            self.meta['current_step'] += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "prune_meta = {\n",
    "        \"starting_step\" : 0,\n",
    "        \"current_step\": 0,\n",
    "        \"ending_step\": epochs * np.ceil(len(data.train_ds)/bs),\n",
    "        \"final_sparsity\": 90,\n",
    "        \"initial_sparsity\": 0,\n",
    "        \"span\": 100\n",
    "    }"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn = Learner(data, Net().cuda(), metrics=accuracy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch     train_loss  valid_loss  accuracy  time    \n",
      "0         0.049754    0.066799    0.977900  00:13     \n",
      "Sparsity: 69.44792998081617\n",
      "1         0.035192    0.031569    0.990700  00:14     \n",
      "Sparsity: 89.00918350108631\n",
      "2         0.019769    0.022909    0.991100  00:14     \n",
      "Sparsity: 90.0\n"
     ]
    }
   ],
   "source": [
    "learn.fit(epochs, 1e-3, callbacks=SparsifyCallback(meta=prune_meta))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "def model_stats(model):\n",
    "        for k, m in enumerate(model.modules()):\n",
    "\n",
    "            if not isinstance(m, Net):\n",
    "\n",
    "                weight = m.weight.data.view(-1).clone().abs()\n",
    "                non_zero_weights = weight.gt(0).float().cuda().sum()                \n",
    "                total_weights = weight.shape[0]\n",
    "                zero_weights = 100 - int(non_zero_weights*100/total_weights)\n",
    "                print(m)\n",
    "                print(\"Percentage of zero weights: \", zero_weights)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Conv2d(3, 20, kernel_size=(5, 5), stride=(1, 1))\n",
      "Percentage of zero weights:  90\n",
      "Conv2d(20, 50, kernel_size=(5, 5), stride=(1, 1))\n",
      "Percentage of zero weights:  90\n",
      "Linear(in_features=800, out_features=500, bias=True)\n",
      "Percentage of zero weights:  90\n",
      "Linear(in_features=500, out_features=10, bias=True)\n",
      "Percentage of zero weights:  90\n"
     ]
    }
   ],
   "source": [
    "model_stats(learn.model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "Copie de pruning-torch.ipynb",
   "provenance": [],
   "version": "0.3.2"
  },
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}