notebooks/ToyRNN.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "32325127",
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "import sys\n",
    "sys.path.append(\"../\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "58519934",
   "metadata": {},
   "outputs": [],
   "source": [
    "import jax\n",
    "import jax.numpy as jnp\n",
    "from jax import vmap\n",
    "\n",
    "import optax\n",
    "\n",
    "import numpy as np\n",
    "from functools import partial"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "a2fd3f37",
   "metadata": {},
   "outputs": [],
   "source": [
    "# copied from https://github.com/DSKSD/DeepNLP-models-Pytorch/tree/master\n",
    "\n",
    "flatten = lambda l: [item for sublist in l for item in sublist]\n",
    "\n",
    "def prepare_sequence(seq, to_index):\n",
    "    idxs = list(map(lambda w: to_index[w] if to_index.get(w) is not None else to_index[\"<unk>\"], seq))\n",
    "    return idxs\n",
    "\n",
    "def prepare_ptb_dataset(filename, word2index=None):\n",
    "    corpus = open(filename, 'r', encoding='utf-8').readlines()\n",
    "    corpus = flatten([co.strip().split() + ['</s>'] for co in corpus])\n",
    "    \n",
    "    if word2index == None:\n",
    "        vocab = list(set(corpus))\n",
    "        word2index = {'<unk>': 0}\n",
    "        for vo in vocab:\n",
    "            if word2index.get(vo) is None:\n",
    "                word2index[vo] = len(word2index)\n",
    "    \n",
    "    return prepare_sequence(corpus, word2index), word2index\n",
    "\n",
    "\n",
    "train_data, word2index = prepare_ptb_dataset('../ironhide/datasets/ptb/ptb.train.txt',)\n",
    "dev_data , _ = prepare_ptb_dataset('../ironhide/datasets/ptb/ptb.valid.txt', word2index)\n",
    "test_data, _ = prepare_ptb_dataset('../ironhide/datasets/ptb/ptb.test.txt', word2index)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "aa3ee059",
   "metadata": {},
   "outputs": [],
   "source": [
    "BATCH_SIZE = 100\n",
    "\n",
    "VOCAB_SIZE = 10000\n",
    "EMBED_SIZE = 64\n",
    "HIDDEN_SIZE = 32"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "64a71b41",
   "metadata": {},
   "outputs": [],
   "source": [
    "# borrowed code from https://github.com/pytorch/examples/tree/master/word_language_model\n",
    "def batchify(data, bsz):\n",
    "    # Work out how cleanly we can divide the dataset into bsz parts.\n",
    "    nbatch = int(len(data) / bsz)\n",
    "    \n",
    "    # Trim off any extra elements that wouldn't cleanly fit (remainders).\n",
    "    data = data[0: nbatch * bsz]\n",
    "    \n",
    "    # Evenly divide the data across the bsz batches.\n",
    "    data = np.reshape(data, (bsz, -1))\n",
    "    return data\n",
    "\n",
    "\n",
    "train_data = batchify(train_data, BATCH_SIZE)\n",
    "dev_data = batchify(dev_data, BATCH_SIZE//2)\n",
    "test_data = batchify(test_data, BATCH_SIZE//2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "19f14a36",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(100, 9292)"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_data.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "9bd93765",
   "metadata": {},
   "outputs": [],
   "source": [
    "from ironhide.models.rnn import ToyRNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "b6661bd1",
   "metadata": {},
   "outputs": [],
   "source": [
    "layer_sizes = [VOCAB_SIZE, EMBED_SIZE, HIDDEN_SIZE]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "c0ccd557",
   "metadata": {},
   "outputs": [],
   "source": [
    "toy_rnn = ToyRNN(layer_sizes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "dfd4ce0d",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "9.210332\n",
      "9.196235\n",
      "9.1442585\n",
      "9.021426\n",
      "8.843581\n",
      "8.566142\n",
      "8.279136\n",
      "8.057715\n",
      "7.8017063\n",
      "7.544209\n",
      "7.4141912\n",
      "7.2681303\n",
      "7.130321\n",
      "6.976433\n",
      "6.891679\n",
      "6.930152\n",
      "6.954357\n",
      "6.9107823\n",
      "6.9933214\n",
      "6.9568186\n",
      "7.0555573\n",
      "6.9841003\n",
      "7.052563\n",
      "6.9837475\n",
      "7.2074533\n",
      "7.0360703\n",
      "7.1297913\n",
      "7.2954264\n",
      "7.1510925\n",
      "6.8956223\n",
      "7.169015\n",
      "7.1527796\n",
      "6.959667\n",
      "6.9115567\n",
      "6.9632096\n",
      "7.095815\n",
      "7.1992373\n",
      "7.2657056\n",
      "6.9459934\n",
      "7.0386815\n",
      "7.1011987\n",
      "7.0226693\n",
      "7.2957115\n",
      "6.9610863\n",
      "6.9926744\n",
      "6.9223294\n",
      "6.86336\n",
      "6.9864597\n",
      "7.1788597\n",
      "6.882137\n",
      "6.953119\n",
      "6.9041076\n",
      "6.862512\n",
      "6.8227634\n",
      "6.86561\n",
      "6.877152\n",
      "6.8058324\n",
      "6.7857914\n",
      "6.814853\n",
      "6.6911697\n",
      "6.8264227\n",
      "6.7979565\n",
      "6.6703563\n",
      "6.7713976\n",
      "6.696469\n",
      "6.933451\n",
      "6.6649413\n",
      "6.782641\n",
      "6.8120165\n",
      "6.7059264\n",
      "6.676217\n",
      "6.785879\n",
      "6.9199467\n",
      "6.8755655\n",
      "6.831278\n",
      "6.6198163\n",
      "6.7888885\n",
      "6.505158\n",
      "6.5362267\n",
      "6.7496786\n",
      "6.676518\n",
      "6.7374883\n",
      "6.638535\n",
      "6.601127\n",
      "6.6576414\n",
      "6.6096478\n",
      "6.608624\n",
      "6.6618285\n",
      "6.6408243\n",
      "6.617655\n",
      "6.557988\n",
      "6.654204\n",
      "6.5584116\n",
      "6.426489\n",
      "6.5937634\n",
      "6.490703\n",
      "6.476086\n",
      "6.4601035\n",
      "6.613132\n",
      "6.450916\n",
      "6.58074\n",
      "6.5737305\n",
      "6.5629344\n",
      "6.699541\n",
      "6.5791016\n",
      "6.489977\n",
      "6.438784\n",
      "6.4397655\n",
      "6.5974236\n",
      "6.416807\n",
      "6.672091\n",
      "6.6099057\n",
      "6.5857153\n",
      "6.571959\n",
      "6.429189\n",
      "6.371197\n",
      "6.432085\n",
      "6.4228497\n",
      "6.449795\n",
      "6.5055466\n",
      "6.4095325\n",
      "6.378902\n",
      "6.6173344\n",
      "6.506464\n",
      "6.42562\n",
      "6.55071\n",
      "6.6237464\n",
      "6.428731\n",
      "6.499007\n",
      "6.504085\n",
      "6.6020617\n",
      "6.494121\n",
      "6.2909484\n",
      "6.2776666\n",
      "6.4437966\n",
      "6.431745\n",
      "6.298474\n",
      "6.30724\n",
      "6.1538672\n",
      "6.4847693\n",
      "6.404713\n",
      "6.344017\n",
      "6.294642\n",
      "6.415467\n",
      "6.3244066\n",
      "6.159973\n",
      "6.318835\n",
      "6.3147025\n",
      "6.3395905\n",
      "6.397005\n",
      "6.111226\n",
      "6.3488193\n",
      "6.1624317\n",
      "6.34449\n",
      "6.3058534\n",
      "6.22514\n",
      "6.3231688\n",
      "6.0932755\n",
      "6.3290677\n",
      "6.1326647\n",
      "6.269503\n",
      "6.3551526\n",
      "6.3917747\n",
      "6.1947546\n",
      "6.2700067\n",
      "6.2946825\n",
      "6.2498984\n",
      "5.934281\n",
      "6.362466\n",
      "6.353697\n",
      "6.1387763\n",
      "6.107413\n",
      "6.1624756\n",
      "6.1093736\n",
      "5.988038\n",
      "6.1865716\n",
      "6.2454357\n",
      "6.214922\n",
      "6.1473584\n",
      "6.1211386\n",
      "6.133562\n",
      "6.2888913\n",
      "6.5228024\n",
      "6.1481113\n",
      "6.3098164\n",
      "6.2441854\n",
      "6.218231\n",
      "6.1661143\n",
      "5.9940743\n",
      "6.174498\n",
      "6.0979896\n",
      "6.1342688\n",
      "6.168496\n",
      "6.0988975\n",
      "6.0808873\n",
      "6.037755\n",
      "6.121489\n",
      "6.096301\n",
      "6.029386\n",
      "6.319808\n",
      "6.1463165\n",
      "6.0545373\n",
      "6.177245\n",
      "6.2633157\n",
      "6.3260508\n",
      "6.0102406\n",
      "6.022345\n",
      "6.2514567\n",
      "5.9934235\n",
      "6.0468225\n",
      "6.2031317\n",
      "6.184138\n",
      "6.1282177\n",
      "6.140301\n",
      "6.086175\n",
      "5.9451013\n",
      "5.979778\n",
      "6.2259054\n",
      "6.0484447\n",
      "5.968528\n",
      "5.801412\n",
      "6.1528034\n",
      "5.973572\n",
      "5.9234595\n",
      "5.869315\n",
      "5.9849577\n",
      "6.153066\n",
      "5.990333\n",
      "5.962215\n",
      "6.0154037\n",
      "6.1816063\n",
      "6.00637\n",
      "5.938558\n",
      "6.0841465\n",
      "6.286182\n",
      "5.935166\n",
      "6.091692\n",
      "6.1142416\n",
      "6.1249394\n",
      "5.967766\n",
      "5.9469213\n",
      "6.0170093\n",
      "6.0971823\n",
      "6.0977597\n",
      "6.0907555\n",
      "6.0125585\n",
      "6.032136\n",
      "6.119221\n",
      "6.086304\n",
      "6.1024346\n",
      "6.1143875\n",
      "5.8881993\n",
      "5.8251085\n",
      "5.9760327\n",
      "6.206515\n",
      "5.9603972\n",
      "5.8168125\n",
      "5.8600535\n",
      "5.8503923\n",
      "5.9174004\n",
      "6.195488\n",
      "5.8775244\n",
      "6.0845947\n",
      "5.950658\n",
      "5.94971\n",
      "5.9018087\n",
      "6.0831847\n",
      "6.013934\n",
      "5.981296\n",
      "6.101008\n",
      "5.684571\n",
      "5.937145\n",
      "5.9350853\n",
      "5.864093\n",
      "6.0363755\n",
      "5.993302\n",
      "5.864431\n",
      "5.8339486\n",
      "5.8999076\n",
      "5.9809103\n",
      "5.971721\n",
      "5.952134\n",
      "5.8970532\n",
      "6.092819\n",
      "6.0979333\n",
      "5.9677324\n",
      "5.9766674\n",
      "5.9456506\n",
      "5.8484387\n",
      "5.788141\n",
      "5.9735737\n",
      "5.971218\n",
      "5.8490534\n",
      "5.9064355\n",
      "6.023979\n",
      "6.1129184\n",
      "6.049415\n",
      "5.9252853\n",
      "5.7400208\n",
      "5.8114886\n",
      "5.7449536\n",
      "5.8799534\n",
      "5.889571\n",
      "5.9269495\n",
      "6.053692\n",
      "5.8565845\n",
      "5.7966075\n",
      "5.9734473\n",
      "5.831302\n",
      "5.676985\n",
      "5.853305\n",
      "5.9426403\n",
      "5.8583393\n",
      "5.7930746\n",
      "5.7822385\n",
      "5.7715464\n",
      "5.8349314\n",
      "5.715023\n",
      "5.7647057\n",
      "5.8007617\n",
      "5.759304\n",
      "5.8326488\n",
      "5.936921\n",
      "5.899065\n",
      "5.7296863\n",
      "5.868474\n",
      "5.7656064\n",
      "5.843275\n",
      "5.873087\n",
      "5.9342685\n",
      "5.942824\n",
      "5.9617186\n",
      "6.020813\n",
      "6.0229554\n",
      "5.8431015\n",
      "5.828364\n",
      "5.9773083\n",
      "5.8157372\n",
      "5.8268323\n",
      "5.840089\n",
      "5.7318206\n",
      "5.5885262\n",
      "5.8110166\n",
      "5.5506554\n",
      "5.897616\n",
      "5.7388077\n",
      "5.5648866\n",
      "5.7076364\n",
      "5.7997117\n",
      "5.693063\n",
      "6.056035\n",
      "5.9293365\n",
      "5.962409\n",
      "6.041075\n",
      "5.7135806\n",
      "5.865376\n",
      "5.867125\n",
      "5.8646135\n",
      "5.7804356\n",
      "5.8759637\n",
      "5.9712753\n",
      "5.6745872\n",
      "5.77058\n",
      "5.742896\n",
      "5.58265\n",
      "5.8947883\n",
      "5.7988424\n",
      "5.7305346\n",
      "5.96737\n",
      "5.713569\n",
      "5.8837113\n",
      "5.6962767\n",
      "5.8784266\n",
      "5.7059045\n",
      "5.6981096\n",
      "5.9078507\n",
      "5.684701\n",
      "6.141824\n",
      "5.747271\n",
      "5.8561974\n",
      "5.6720934\n",
      "5.825915\n",
      "5.7400684\n",
      "5.7186255\n",
      "5.8593893\n",
      "5.8972716\n",
      "5.893816\n",
      "5.750078\n",
      "5.7904253\n",
      "5.8161263\n",
      "5.6507854\n",
      "5.816991\n",
      "5.671158\n",
      "5.732498\n",
      "5.713177\n",
      "5.4924126\n",
      "5.647948\n",
      "5.897496\n",
      "5.942908\n",
      "5.7274423\n",
      "5.472802\n",
      "5.872809\n",
      "5.6720104\n",
      "5.7231865\n",
      "5.878314\n",
      "5.685267\n",
      "5.913132\n",
      "5.642545\n",
      "5.498557\n",
      "5.781333\n",
      "5.7994304\n",
      "5.602355\n",
      "5.657547\n",
      "5.7690415\n",
      "5.8338046\n",
      "5.603819\n",
      "5.5650787\n",
      "5.7463465\n",
      "5.589376\n",
      "5.5651135\n",
      "5.9158983\n",
      "5.672634\n",
      "5.608286\n",
      "5.557853\n",
      "5.814476\n",
      "5.730975\n",
      "5.788697\n",
      "5.7173877\n",
      "5.849038\n",
      "5.6714563\n",
      "5.6499643\n",
      "5.5999255\n",
      "5.569803\n",
      "5.558783\n",
      "5.4398603\n",
      "5.470887\n",
      "5.721805\n",
      "5.5469127\n",
      "5.5530405\n",
      "5.4530396\n",
      "5.503889\n",
      "5.5903406\n",
      "5.873667\n",
      "5.420462\n",
      "5.920796\n",
      "5.5916953\n",
      "5.5828257\n",
      "5.618415\n",
      "5.791532\n",
      "5.5720305\n",
      "5.7479587\n",
      "5.8748856\n",
      "5.6206274\n",
      "5.513861\n",
      "5.817955\n",
      "5.844731\n",
      "5.692982\n",
      "5.6688395\n",
      "5.4413137\n",
      "5.714382\n",
      "5.521909\n",
      "5.4716806\n",
      "5.8255405\n",
      "5.674286\n",
      "5.5298204\n",
      "5.676809\n",
      "5.5103745\n",
      "5.5976486\n",
      "5.717471\n",
      "5.305032\n",
      "5.388928\n",
      "5.4867563\n",
      "5.715686\n",
      "5.6651983\n",
      "5.5487595\n",
      "5.4062047\n",
      "5.5449033\n",
      "5.569155\n",
      "5.484682\n",
      "5.6104684\n",
      "5.661173\n",
      "5.3390794\n",
      "5.503911\n",
      "5.7440705\n",
      "5.5636363\n",
      "5.560183\n",
      "5.8211913\n",
      "5.6462193\n",
      "5.3742986\n",
      "5.327314\n",
      "5.47102\n",
      "5.316797\n",
      "5.7157054\n",
      "5.534755\n",
      "5.4778857\n",
      "5.529447\n",
      "5.573502\n",
      "5.714265\n",
      "5.618193\n",
      "5.4862657\n",
      "5.378686\n"
     ]
    }
   ],
   "source": [
    "training_losses = toy_rnn.fit(train_data, seq_length=8, max_iter=500)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "a0c9f707",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "%matplotlib inline\n",
    "\n",
    "batches = range(1, len(training_losses) + 1)\n",
    "\n",
    "# Plotting the training loss\n",
    "plt.plot(batches, training_losses, 'b', label='Training Loss')\n",
    "plt.title('Training Loss Reduction')\n",
    "plt.xlabel('Batches')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9c8cd9eb",
   "metadata": {},
   "outputs": [],
   "source": [
    "ba"
   ]
  }
 ],
 "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": "3.8.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}