Basic_learner.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from fastai.imports import *\n",
    "from fastai.torch_imports import *\n",
    "from fastai.column_data import *\n",
    "import torch.nn as nn"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Generate synthetic data of x1^2 + x2^2 function with random noise"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "input_range = np.arange(-1,1,.05)\n",
    "x1,x2 = np.meshgrid(input_range, input_range) \n",
    "x1 = x1.reshape([-1,1])\n",
    "x2 = x2.reshape([-1,1])\n",
    "X =np.concatenate([x1,x2], axis=1)\n",
    "Y = x1 * x1 + x2 * x2 + np.random.normal(.1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Take out random 20% for validation set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "shuff = np.random.permutation(len(X))\n",
    "train_size = int(len(X) * 0.8)\n",
    "train_idx = shuff[:train_size]\n",
    "val_idx = shuff[train_size:]\n",
    "trainX, trainY, valX, valY = X[train_idx], Y[train_idx], X[val_idx], Y[val_idx]\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Use ArraysIndexDataset put arrays into datasets and create ModelData out of it"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_aid = ArraysIndexDataset(trainX, trainY, None)\n",
    "val_aid = ArraysIndexDataset(valX, valY, None)\n",
    "md = ModelData.from_dls('.', DataLoader(train_aid,batch_size=16), DataLoader(val_aid,batch_size=16))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Tiny neural net pytorch model (no .cuda() so I can run it on CPU)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "mininet = nn.Sequential(\n",
    "    nn.Linear(2, 10),\n",
    "    nn.ReLU(),\n",
    "    nn.Linear(10, 10),\n",
    "    nn.ReLU(),\n",
    "    nn.Linear(10, 1),\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Create Learner for my model, need to wrap it in fastai BasicModel.\n",
    "This allows me to use the learning rate finder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "74094c2b143740d88c65fcfc7faaf9b8",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/80 [00:00<?, ?it/s]\n",
      "                                      \r"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn = Learner(md,BasicModel(mininet),crit=nn.MSELoss())\n",
    "learn.lr_find()\n",
    "learn.sched.plot()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And finally train with cyclical learning rate."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7a744b85ad1a485b88687a1e7103a8b1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch', max=14), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch      trn_loss   val_loss        \n",
      "    0      0.522018   0.154176  \n",
      "    1      0.202892   0.135903        \n",
      "    2      0.106134   0.052498        \n",
      "    3      0.044772   0.024611        \n",
      "    4      0.023902   0.019703        \n",
      "    5      0.0185     0.019254        \n",
      "    6      0.015372   0.015297        \n",
      "    7      0.012699   0.012472        \n",
      "    8      0.010711   0.010909        \n",
      "    9      0.009289   0.009893        \n",
      "    10     0.008443   0.009093        \n",
      "    11     0.007955   0.008642        \n",
      "    12     0.007693   0.008548        \n",
      "    13     0.007584   0.008538        \n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[array([0.00854])]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "learn.fit(1e-2,3,cycle_len=2, cycle_mult=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.008537675789061524"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "preds = mininet(V(valX)).data.cpu().numpy()\n",
    "mse = ((preds-valY)**2).mean()\n",
    "mse"
   ]
  }
 ],
 "metadata": {
  "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": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}