mlp.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch \n",
    "import torch.nn.functional as F\n",
    "import random\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In this notebook, we'll be going a step further from the bigram approach:\n",
    "* training an MLP on inputs with longer context lengths (3 or more characters for now)\n",
    "* including an embedding layer to learn a better representation of the characters\n",
    "* creating a hidden layer with a tanh activation\n",
    "* output layer with a softmax to create probablity distribution "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "words = open('names.txt').read().splitlines()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['emma', 'olivia', 'ava', 'isabella', 'sophia']"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "words[:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [],
   "source": [
    "# set up the vocab size\n",
    "chars = sorted(list(set(''.join(words))))\n",
    "stoi = {c:i+1 for i,c in enumerate(chars)}\n",
    "stoi['.'] = 0 # special token\n",
    "itos = {i:c for c,i in stoi.items()}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "metadata": {},
   "outputs": [],
   "source": [
    "BLOCK_SIZE = 3 # context\n",
    "\n",
    "def build_dataset(words):\n",
    "    X, Y = [], []\n",
    "    for w in words:\n",
    "        context = [0] * BLOCK_SIZE\n",
    "        for c in w + '.':\n",
    "            idx = stoi[c]\n",
    "            X.append(context)\n",
    "            Y.append(idx)\n",
    "            context = context[1:] + [idx] # rm first char index\n",
    "    X = torch.tensor(X)\n",
    "    Y = torch.tensor(Y)\n",
    "    return X, Y\n",
    "\n",
    "random.seed(42)\n",
    "random.shuffle(words)\n",
    "n1 = int(0.8 * len(words))\n",
    "n2 = int(0.9 * len(words))\n",
    "\n",
    "X_train, y_train = build_dataset(words[:n1])\n",
    "X_val, y_val = build_dataset(words[n1:n2])\n",
    "X_test, y_test = build_dataset(words[n2:])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 137,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([182691, 3])"
      ]
     },
     "execution_count": 137,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.size()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 240,
   "metadata": {},
   "outputs": [],
   "source": [
    "EMB_DIM = 10\n",
    "\n",
    "g = torch.Generator().manual_seed(2147483647) \n",
    "C = torch.randn((27, EMB_DIM), generator=g)\n",
    "W1 = torch.randn((EMB_DIM * BLOCK_SIZE, 200), generator=g)\n",
    "b1 = torch.randn(200, generator=g)\n",
    "W2 = torch.randn((200, 27), generator=g) * 0.1\n",
    "b2 = torch.randn(27, generator=g) * 0\n",
    "params = [C, W1, b1, W2, b2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 241,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "11897"
      ]
     },
     "execution_count": 241,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sum(p.nelement() for p in params) # param count"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 242,
   "metadata": {},
   "outputs": [],
   "source": [
    "for p in params:\n",
    "    p.requires_grad = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 243,
   "metadata": {},
   "outputs": [],
   "source": [
    "stepi = []\n",
    "lossi = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 244,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([64, 3, 10])\n",
      "torch.Size([64, 30])\n"
     ]
    }
   ],
   "source": [
    "idx = torch.randint(0, X_train.shape[0], (64,)) # creating mini batches over random indices\n",
    "emb = C[X_train[idx]]\n",
    "print(emb.shape)\n",
    "print(emb.view(emb.size(0),-1).shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 245,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 0 | loss: 4.1973\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "4.197343826293945"
      ]
     },
     "execution_count": 245,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "for i in range(1):\n",
    "    # forward pass\n",
    "    idx = torch.randint(0, X_train.shape[0], (64,)) # creating mini batches over random indices\n",
    "    emb = C[X_train[idx]].view(-1, EMB_DIM * BLOCK_SIZE)\n",
    "    h = torch.tanh(emb @ W1 + b1)\n",
    "    logits = h @ W2 + b2\n",
    "\n",
    "    # l2 = 0.01 * (torch.sum(W1**2) + torch.sum(W2**2))\n",
    "    loss = F.cross_entropy(logits, y_train[idx])\n",
    "\n",
    "    # h1 = torch.tanh(emb @ W1 + b1)\n",
    "    # h2 = torch.tanh(h1 @ W2 + b2)\n",
    "    # logits = h2 @ W3 + b3\n",
    "    # loss = F.cross_entropy(logits, y_train[idx])\n",
    "\n",
    "    # zero grad\n",
    "    for p in params:\n",
    "        p.grad = None\n",
    "\n",
    "    # backward\n",
    "    loss.backward()\n",
    "\n",
    "    # update\n",
    "    lr = 0.1 if i < 100000 else 0.01\n",
    "    for p in params:\n",
    "        p.data += -lr * p.grad\n",
    "\n",
    "    if i % 10000 == 0: \n",
    "        print(f\"epoch {i} | loss: {loss:.4f}\")\n",
    "\n",
    "    # stepi.append(i)\n",
    "    # lossi.append(loss.log10().item())\n",
    "\n",
    "loss.item()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 246,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x11df87be0>"
      ]
     },
     "execution_count": 246,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 2000x1000 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(20,10))\n",
    "plt.imshow(h.abs() > 0.99, cmap='grey', interpolation='nearest')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 238,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(h.view(-1).tolist(), 50);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 225,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(h.view(-1).tolist(), 50);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x11e385fa0>]"
      ]
     },
     "execution_count": 156,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# batch64\n",
    "plt.plot(stepi, lossi)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x11d31b610>]"
      ]
     },
     "execution_count": 150,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# batch32\n",
    "plt.plot(stepi, lossi)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 170,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2.0567357540130615"
      ]
     },
     "execution_count": 170,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# training loss over the full training set\n",
    "emb = C[X_train].view(-1, EMB_DIM * BLOCK_SIZE)\n",
    "h = torch.tanh(emb @ W1 + b1)\n",
    "logits = h @ W2 + b2\n",
    "loss = F.cross_entropy(logits, y_train)\n",
    "loss.item()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 171,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2.1300973892211914"
      ]
     },
     "execution_count": 171,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# val loss\n",
    "emb = C[X_val].view(-1, EMB_DIM * BLOCK_SIZE)\n",
    "h = torch.tanh(emb @ W1 + b1)\n",
    "logits = h @ W2 + b2\n",
    "loss = F.cross_entropy(logits, y_val)\n",
    "loss.item()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 172,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2.1434171199798584"
      ]
     },
     "execution_count": 172,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# test loss\n",
    "emb = C[X_test].view(-1, EMB_DIM * BLOCK_SIZE)\n",
    "h = torch.tanh(emb @ W1 + b1)\n",
    "logits = h @ W2 + b2\n",
    "loss = F.cross_entropy(logits, y_test)\n",
    "loss.item()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "carlah.\n",
      "amirick.\n",
      "hiimleigta.\n",
      "caskan.\n",
      "saeja.\n",
      "hutnder.\n",
      "sarti.\n",
      "geri.\n",
      "neronia.\n",
      "chaiir.\n",
      "kaleigh.\n",
      "hamon.\n",
      "cadesian.\n",
      "shoil.\n",
      "alian.\n",
      "quinze.\n",
      "paijarixi.\n",
      "jaxen.\n",
      "duran.\n",
      "edde.\n"
     ]
    }
   ],
   "source": [
    "# sample from the model\n",
    "g = torch.Generator().manual_seed(2147483647 + 10)\n",
    "\n",
    "for _ in range(20):\n",
    "    \n",
    "    out = []\n",
    "    context = [0] * BLOCK_SIZE # initialize with all ...\n",
    "    while True:\n",
    "      emb = C[torch.tensor([context])] # (1,block_size,d)\n",
    "      h = torch.tanh(emb.view(1, -1) @ W1 + b1)\n",
    "      logits = h @ W2 + b2\n",
    "      probs = F.softmax(logits, dim=1)\n",
    "      ix = torch.multinomial(probs, num_samples=1, generator=g).item()\n",
    "      context = context[1:] + [ix]\n",
    "      out.append(ix)\n",
    "      if ix == 0:\n",
    "        break\n",
    "    \n",
    "    print(''.join(itos[i] for i in out))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Learning rate initialization technique"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 209,
   "metadata": {},
   "outputs": [],
   "source": [
    "EMB_DIM = 10\n",
    "\n",
    "g = torch.Generator().manual_seed(2147483647) \n",
    "C = torch.randn((27, EMB_DIM), generator=g)\n",
    "W1 = torch.randn((EMB_DIM * BLOCK_SIZE, 200), generator=g)\n",
    "b1 = torch.randn(200, generator=g)\n",
    "W2 = torch.randn((200, 27), generator=g)\n",
    "b2 = torch.randn(27, generator=g) * 0\n",
    "params = [C, W1, b1, W2, b2]\n",
    "\n",
    "for p in params:\n",
    "    p.requires_grad = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 210,
   "metadata": {},
   "outputs": [],
   "source": [
    "lre = torch.linspace(-3, 0, 1000) # tensor of size [1000] spread out from -3 to 0\n",
    "lrs = 10**lre # learning rates spread out from 0.001 to 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 211,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "25.412076950073242\n",
      "24.960102081298828\n",
      "22.334226608276367\n",
      "26.939409255981445\n",
      "30.658588409423828\n",
      "25.029556274414062\n",
      "24.317527770996094\n",
      "25.63251304626465\n",
      "23.563217163085938\n",
      "24.785871505737305\n",
      "26.71943473815918\n",
      "29.78579330444336\n",
      "25.641244888305664\n",
      "26.279422760009766\n",
      "26.47990608215332\n",
      "26.02263641357422\n",
      "26.013179779052734\n",
      "27.652616500854492\n",
      "25.50037956237793\n",
      "25.03707504272461\n",
      "25.149675369262695\n",
      "25.206119537353516\n",
      "22.527353286743164\n",
      "24.901792526245117\n",
      "26.26892852783203\n",
      "24.417070388793945\n",
      "26.672155380249023\n",
      "22.309532165527344\n",
      "26.76412010192871\n",
      "28.425386428833008\n",
      "24.076807022094727\n",
      "23.918909072875977\n",
      "23.7745418548584\n",
      "24.464183807373047\n",
      "25.09384536743164\n",
      "24.667644500732422\n",
      "25.207714080810547\n",
      "24.812259674072266\n",
      "23.694332122802734\n",
      "26.859052658081055\n",
      "23.684200286865234\n",
      "24.482751846313477\n",
      "22.9853572845459\n",
      "24.293968200683594\n",
      "28.071264266967773\n",
      "26.180349349975586\n",
      "23.60114097595215\n",
      "23.825298309326172\n",
      "23.40693473815918\n",
      "22.697710037231445\n",
      "24.01326560974121\n",
      "21.243688583374023\n",
      "24.486515045166016\n",
      "21.077266693115234\n",
      "24.218137741088867\n",
      "24.225915908813477\n",
      "24.649723052978516\n",
      "24.54364585876465\n",
      "22.662874221801758\n",
      "26.01569938659668\n",
      "22.5478458404541\n",
      "26.115646362304688\n",
      "24.857358932495117\n",
      "24.06843376159668\n",
      "24.761165618896484\n",
      "22.7179012298584\n",
      "22.621479034423828\n",
      "25.054698944091797\n",
      "25.79100227355957\n",
      "20.579980850219727\n",
      "23.490156173706055\n",
      "22.910066604614258\n",
      "24.300113677978516\n",
      "21.374860763549805\n",
      "22.938039779663086\n",
      "25.3114070892334\n",
      "21.87074089050293\n",
      "19.91212272644043\n",
      "23.785676956176758\n",
      "24.609216690063477\n",
      "24.984865188598633\n",
      "25.664133071899414\n",
      "24.96719741821289\n",
      "22.452802658081055\n",
      "23.29185676574707\n",
      "21.026193618774414\n",
      "23.969619750976562\n",
      "22.709333419799805\n",
      "24.010528564453125\n",
      "20.407655715942383\n",
      "27.27540397644043\n",
      "22.033748626708984\n",
      "23.930500030517578\n",
      "24.298917770385742\n",
      "23.091581344604492\n",
      "22.986770629882812\n",
      "26.0553035736084\n",
      "23.885360717773438\n",
      "21.99724769592285\n",
      "20.93625259399414\n",
      "19.732139587402344\n",
      "25.92780113220215\n",
      "21.56834602355957\n",
      "21.735733032226562\n",
      "21.86883544921875\n",
      "25.21951675415039\n",
      "23.192670822143555\n",
      "20.744199752807617\n",
      "22.381324768066406\n",
      "22.601465225219727\n",
      "20.883943557739258\n",
      "21.031652450561523\n",
      "20.856182098388672\n",
      "20.521297454833984\n",
      "23.92245864868164\n",
      "23.92931365966797\n",
      "20.796812057495117\n",
      "24.07891845703125\n",
      "20.114595413208008\n",
      "21.569000244140625\n",
      "20.981544494628906\n",
      "20.944002151489258\n",
      "21.136157989501953\n",
      "20.58318519592285\n",
      "21.906509399414062\n",
      "24.040630340576172\n",
      "21.515844345092773\n",
      "22.79462242126465\n",
      "21.211334228515625\n",
      "21.591062545776367\n",
      "18.901649475097656\n",
      "20.921977996826172\n",
      "22.386350631713867\n",
      "21.679533004760742\n",
      "22.759302139282227\n",
      "22.541156768798828\n",
      "19.979663848876953\n",
      "22.779743194580078\n",
      "21.479570388793945\n",
      "19.94373893737793\n",
      "22.073406219482422\n",
      "23.639766693115234\n",
      "21.12409210205078\n",
      "21.842817306518555\n",
      "19.121868133544922\n",
      "20.34286117553711\n",
      "24.560232162475586\n",
      "25.388574600219727\n",
      "21.093542098999023\n",
      "20.014984130859375\n",
      "20.087474822998047\n",
      "18.951723098754883\n",
      "19.3341121673584\n",
      "20.83563995361328\n",
      "22.361900329589844\n",
      "22.417041778564453\n",
      "20.614208221435547\n",
      "17.353736877441406\n",
      "21.426483154296875\n",
      "19.118881225585938\n",
      "19.912796020507812\n",
      "21.17316246032715\n",
      "21.998027801513672\n",
      "20.88184356689453\n",
      "24.609981536865234\n",
      "21.972469329833984\n",
      "22.199567794799805\n",
      "18.39086151123047\n",
      "18.33717918395996\n",
      "18.503246307373047\n",
      "18.136743545532227\n",
      "20.790327072143555\n",
      "21.607702255249023\n",
      "21.781497955322266\n",
      "20.049449920654297\n",
      "22.129812240600586\n",
      "22.84461212158203\n",
      "19.812135696411133\n",
      "21.492902755737305\n",
      "19.339122772216797\n",
      "20.31093978881836\n",
      "20.900270462036133\n",
      "21.410472869873047\n",
      "21.30817985534668\n",
      "21.293777465820312\n",
      "19.997282028198242\n",
      "18.585939407348633\n",
      "16.7863712310791\n",
      "19.56409454345703\n",
      "20.19472312927246\n",
      "22.68025016784668\n",
      "19.4115047454834\n",
      "19.91450309753418\n",
      "20.253524780273438\n",
      "16.427459716796875\n",
      "19.852468490600586\n",
      "19.594709396362305\n",
      "18.523387908935547\n",
      "20.93562126159668\n",
      "19.359378814697266\n",
      "20.420263290405273\n",
      "20.08905029296875\n",
      "17.85565948486328\n",
      "20.90118980407715\n",
      "19.722169876098633\n",
      "19.884620666503906\n",
      "20.177078247070312\n",
      "22.228740692138672\n",
      "20.12813949584961\n",
      "19.070703506469727\n",
      "20.847087860107422\n",
      "23.568098068237305\n",
      "20.014039993286133\n",
      "20.47722625732422\n",
      "18.701631546020508\n",
      "22.462875366210938\n",
      "18.20603370666504\n",
      "18.511171340942383\n",
      "16.906835556030273\n",
      "20.14168930053711\n",
      "20.45790672302246\n",
      "17.686382293701172\n",
      "15.651216506958008\n",
      "19.585403442382812\n",
      "17.005592346191406\n",
      "20.003738403320312\n",
      "18.41323471069336\n",
      "21.82874870300293\n",
      "19.65104103088379\n",
      "18.86302947998047\n",
      "20.63150978088379\n",
      "20.480695724487305\n",
      "19.48807144165039\n",
      "20.280237197875977\n",
      "16.250137329101562\n",
      "19.408004760742188\n",
      "16.313365936279297\n",
      "18.53536605834961\n",
      "18.121183395385742\n",
      "19.051712036132812\n",
      "14.619728088378906\n",
      "17.823585510253906\n",
      "16.287187576293945\n",
      "18.55343246459961\n",
      "18.58717918395996\n",
      "18.70114517211914\n",
      "17.293088912963867\n",
      "18.098161697387695\n",
      "19.17051887512207\n",
      "18.019577026367188\n",
      "18.06667137145996\n",
      "16.11016845703125\n",
      "16.04570770263672\n",
      "18.872987747192383\n",
      "18.101318359375\n",
      "17.535888671875\n",
      "15.821523666381836\n",
      "18.516578674316406\n",
      "19.228424072265625\n",
      "20.57151222229004\n",
      "17.611934661865234\n",
      "18.276411056518555\n",
      "18.05716323852539\n",
      "21.99453353881836\n",
      "18.4362850189209\n",
      "16.58999252319336\n",
      "14.569233894348145\n",
      "19.921337127685547\n",
      "18.57185935974121\n",
      "18.630455017089844\n",
      "17.531970977783203\n",
      "16.10009002685547\n",
      "16.914934158325195\n",
      "18.394819259643555\n",
      "16.715526580810547\n",
      "19.661739349365234\n",
      "16.972129821777344\n",
      "17.19000244140625\n",
      "16.543045043945312\n",
      "19.38549041748047\n",
      "17.7441349029541\n",
      "17.351198196411133\n",
      "15.65994644165039\n",
      "13.708699226379395\n",
      "17.56841468811035\n",
      "17.734350204467773\n",
      "17.793764114379883\n",
      "20.022489547729492\n",
      "16.43182945251465\n",
      "18.759347915649414\n",
      "16.16208839416504\n",
      "16.34282875061035\n",
      "18.380674362182617\n",
      "17.698707580566406\n",
      "15.328229904174805\n",
      "17.187864303588867\n",
      "15.247488021850586\n",
      "16.32949447631836\n",
      "19.732276916503906\n",
      "14.543469429016113\n",
      "16.814476013183594\n",
      "17.95030975341797\n",
      "17.921648025512695\n",
      "17.538827896118164\n",
      "14.737112045288086\n",
      "18.33069610595703\n",
      "15.099164009094238\n",
      "15.296079635620117\n",
      "17.085773468017578\n",
      "14.387886047363281\n",
      "18.22842025756836\n",
      "18.842798233032227\n",
      "15.05219841003418\n",
      "15.894901275634766\n",
      "18.234848022460938\n",
      "20.534191131591797\n",
      "14.828778266906738\n",
      "16.779258728027344\n",
      "14.761775970458984\n",
      "15.209671974182129\n",
      "17.082077026367188\n",
      "18.15104866027832\n",
      "15.367799758911133\n",
      "15.392168998718262\n",
      "16.344161987304688\n",
      "14.080928802490234\n",
      "14.84919261932373\n",
      "16.07406234741211\n",
      "13.548918724060059\n",
      "15.584797859191895\n",
      "14.425164222717285\n",
      "14.49711799621582\n",
      "16.71271324157715\n",
      "16.88355827331543\n",
      "14.53210163116455\n",
      "14.628009796142578\n",
      "13.245295524597168\n",
      "16.787263870239258\n",
      "13.900520324707031\n",
      "17.553207397460938\n",
      "14.286356925964355\n",
      "15.938310623168945\n",
      "11.7998628616333\n",
      "15.087955474853516\n",
      "16.53461265563965\n",
      "15.89911937713623\n",
      "13.502167701721191\n",
      "13.767096519470215\n",
      "17.020036697387695\n",
      "15.415323257446289\n",
      "13.63786506652832\n",
      "14.719406127929688\n",
      "16.46636962890625\n",
      "17.418466567993164\n",
      "15.340770721435547\n",
      "16.771636962890625\n",
      "15.024168014526367\n",
      "13.988763809204102\n",
      "14.980829238891602\n",
      "12.873682022094727\n",
      "18.477502822875977\n",
      "13.93608283996582\n",
      "16.914813995361328\n",
      "13.365339279174805\n",
      "15.406152725219727\n",
      "13.866235733032227\n",
      "13.467511177062988\n",
      "12.921154975891113\n",
      "15.749809265136719\n",
      "14.784042358398438\n",
      "15.656085014343262\n",
      "13.702481269836426\n",
      "14.036688804626465\n",
      "15.025775909423828\n",
      "13.530773162841797\n",
      "13.747352600097656\n",
      "14.281195640563965\n",
      "14.245448112487793\n",
      "14.281939506530762\n",
      "16.452945709228516\n",
      "13.598546028137207\n",
      "12.709338188171387\n",
      "12.929962158203125\n",
      "13.799050331115723\n",
      "16.210649490356445\n",
      "14.409868240356445\n",
      "13.846454620361328\n",
      "14.11350154876709\n",
      "12.367510795593262\n",
      "13.505223274230957\n",
      "12.50576400756836\n",
      "12.119500160217285\n",
      "15.414996147155762\n",
      "11.752182006835938\n",
      "12.02121353149414\n",
      "14.821345329284668\n",
      "12.739481925964355\n",
      "12.400092124938965\n",
      "12.569028854370117\n",
      "12.616069793701172\n",
      "14.967374801635742\n",
      "12.74454116821289\n",
      "13.583574295043945\n",
      "10.343524932861328\n",
      "13.002232551574707\n",
      "14.120434761047363\n",
      "12.979312896728516\n",
      "10.554070472717285\n",
      "11.057208061218262\n",
      "13.560822486877441\n",
      "11.201860427856445\n",
      "11.787495613098145\n",
      "14.149736404418945\n",
      "11.991908073425293\n",
      "10.630630493164062\n",
      "11.97218132019043\n",
      "12.475896835327148\n",
      "12.816884994506836\n",
      "12.275853157043457\n",
      "13.504979133605957\n",
      "11.183762550354004\n",
      "12.927671432495117\n",
      "13.422664642333984\n",
      "13.105731010437012\n",
      "13.605217933654785\n",
      "12.211427688598633\n",
      "13.354777336120605\n",
      "11.298724174499512\n",
      "12.700770378112793\n",
      "12.353736877441406\n",
      "12.165904998779297\n",
      "13.488523483276367\n",
      "11.940277099609375\n",
      "13.212433815002441\n",
      "11.548725128173828\n",
      "13.380653381347656\n",
      "13.115131378173828\n",
      "13.563642501831055\n",
      "14.135555267333984\n",
      "12.150054931640625\n",
      "9.747739791870117\n",
      "13.056198120117188\n",
      "12.02258014678955\n",
      "12.791686058044434\n",
      "9.999019622802734\n",
      "13.858255386352539\n",
      "11.94572639465332\n",
      "11.293937683105469\n",
      "13.145262718200684\n",
      "9.7647066116333\n",
      "12.216604232788086\n",
      "12.544425964355469\n",
      "11.461893081665039\n",
      "10.88576889038086\n",
      "12.471015930175781\n",
      "11.090293884277344\n",
      "10.906200408935547\n",
      "10.32209587097168\n",
      "14.001911163330078\n",
      "9.908681869506836\n",
      "12.698397636413574\n",
      "8.920905113220215\n",
      "12.525598526000977\n",
      "11.815277099609375\n",
      "12.134218215942383\n",
      "10.576627731323242\n",
      "9.72913932800293\n",
      "14.29831314086914\n",
      "13.701780319213867\n",
      "12.890429496765137\n",
      "9.900461196899414\n",
      "12.77249813079834\n",
      "13.584546089172363\n",
      "11.020340919494629\n",
      "9.994226455688477\n",
      "8.91348934173584\n",
      "10.70274829864502\n",
      "10.091477394104004\n",
      "9.846320152282715\n",
      "9.419800758361816\n",
      "10.155145645141602\n",
      "10.813528060913086\n",
      "9.387022018432617\n",
      "9.723989486694336\n",
      "10.105583190917969\n",
      "11.16166877746582\n",
      "9.707433700561523\n",
      "12.234739303588867\n",
      "10.687987327575684\n",
      "9.165336608886719\n",
      "11.297906875610352\n",
      "10.445420265197754\n",
      "10.94372844696045\n",
      "10.581990242004395\n",
      "10.183206558227539\n",
      "8.191448211669922\n",
      "12.479743003845215\n",
      "9.272814750671387\n",
      "10.575050354003906\n",
      "8.781550407409668\n",
      "10.250506401062012\n",
      "8.720808982849121\n",
      "10.727657318115234\n",
      "10.930768013000488\n",
      "12.540977478027344\n",
      "10.003340721130371\n",
      "9.32773494720459\n",
      "9.1211519241333\n",
      "11.562442779541016\n",
      "9.703831672668457\n",
      "11.106463432312012\n",
      "10.313117980957031\n",
      "12.004536628723145\n",
      "8.367897987365723\n",
      "10.642087936401367\n",
      "9.171318054199219\n",
      "12.14456558227539\n",
      "11.254575729370117\n",
      "9.690874099731445\n",
      "8.857476234436035\n",
      "10.9574556350708\n",
      "11.679807662963867\n",
      "10.136260032653809\n",
      "10.409748077392578\n",
      "7.79173469543457\n",
      "10.859745025634766\n",
      "9.721339225769043\n",
      "9.764034271240234\n",
      "9.506978988647461\n",
      "9.668822288513184\n",
      "10.375391006469727\n",
      "9.239137649536133\n",
      "11.102481842041016\n",
      "10.34960651397705\n",
      "9.244979858398438\n",
      "9.84671688079834\n",
      "10.49755573272705\n",
      "10.130456924438477\n",
      "9.80290699005127\n",
      "11.647847175598145\n",
      "9.787638664245605\n",
      "8.237993240356445\n",
      "8.768970489501953\n",
      "8.781014442443848\n",
      "9.743583679199219\n",
      "10.523478507995605\n",
      "8.440277099609375\n",
      "9.425712585449219\n",
      "10.290558815002441\n",
      "8.816877365112305\n",
      "9.154443740844727\n",
      "7.206700801849365\n",
      "10.588300704956055\n",
      "9.072596549987793\n",
      "10.236872673034668\n",
      "10.18338680267334\n",
      "7.730976104736328\n",
      "11.601140022277832\n",
      "9.697868347167969\n",
      "10.93351936340332\n",
      "10.030196189880371\n",
      "6.709292888641357\n",
      "9.60832691192627\n",
      "6.603701114654541\n",
      "8.758570671081543\n",
      "8.217137336730957\n",
      "11.949156761169434\n",
      "9.507940292358398\n",
      "8.58802604675293\n",
      "11.031871795654297\n",
      "9.429106712341309\n",
      "10.489008903503418\n",
      "6.803842544555664\n",
      "10.17839527130127\n",
      "8.851862907409668\n",
      "10.733983993530273\n",
      "8.737058639526367\n",
      "8.56701374053955\n",
      "8.625421524047852\n",
      "9.923121452331543\n",
      "8.539873123168945\n",
      "8.813498497009277\n",
      "9.19442081451416\n",
      "7.848241806030273\n",
      "7.884323596954346\n",
      "7.59107780456543\n",
      "8.12540054321289\n",
      "7.503921985626221\n",
      "7.976367473602295\n",
      "8.656963348388672\n",
      "10.619449615478516\n",
      "9.433002471923828\n",
      "7.847597599029541\n",
      "8.669647216796875\n",
      "7.634931564331055\n",
      "9.692808151245117\n",
      "7.565920352935791\n",
      "8.950568199157715\n",
      "7.055250644683838\n",
      "7.470794200897217\n",
      "6.7104811668396\n",
      "8.355412483215332\n",
      "8.601850509643555\n",
      "9.687093734741211\n",
      "6.467170238494873\n",
      "10.898594856262207\n",
      "7.565789699554443\n",
      "6.029592037200928\n",
      "8.350068092346191\n",
      "7.1388726234436035\n",
      "6.844168663024902\n",
      "7.017860412597656\n",
      "8.130913734436035\n",
      "7.908946514129639\n",
      "8.628708839416504\n",
      "7.699108123779297\n",
      "8.321897506713867\n",
      "6.180063724517822\n",
      "8.171321868896484\n",
      "8.117927551269531\n",
      "9.176627159118652\n",
      "7.711779594421387\n",
      "9.06550121307373\n",
      "8.057093620300293\n",
      "7.692635536193848\n",
      "8.996438980102539\n",
      "6.7504706382751465\n",
      "6.878643989562988\n",
      "10.431674003601074\n",
      "7.677513122558594\n",
      "7.933310508728027\n",
      "7.79130220413208\n",
      "6.174947738647461\n",
      "7.078901290893555\n",
      "7.51767635345459\n",
      "7.284752368927002\n",
      "8.072871208190918\n",
      "7.055579662322998\n",
      "8.398027420043945\n",
      "6.715819835662842\n",
      "8.878727912902832\n",
      "8.123669624328613\n",
      "8.768206596374512\n",
      "7.592058181762695\n",
      "7.33500862121582\n",
      "6.292102813720703\n",
      "6.8333916664123535\n",
      "6.58093786239624\n",
      "8.170300483703613\n",
      "8.534403800964355\n",
      "8.694093704223633\n",
      "8.537354469299316\n",
      "7.45054817199707\n",
      "6.881276607513428\n",
      "6.941120147705078\n",
      "8.192270278930664\n",
      "5.554057598114014\n",
      "6.733492851257324\n",
      "6.747755527496338\n",
      "4.957910537719727\n",
      "7.961345672607422\n",
      "6.65543270111084\n",
      "7.584658145904541\n",
      "7.027492523193359\n",
      "7.460750102996826\n",
      "7.673327922821045\n",
      "6.70610237121582\n",
      "8.27570915222168\n",
      "6.35024356842041\n",
      "6.272027492523193\n",
      "7.618823528289795\n",
      "8.788009643554688\n",
      "6.30364990234375\n",
      "6.77175760269165\n",
      "6.472630500793457\n",
      "8.178035736083984\n",
      "6.112629413604736\n",
      "11.290245056152344\n",
      "8.79719352722168\n",
      "7.544471263885498\n",
      "6.257068157196045\n",
      "6.012401580810547\n",
      "5.9170145988464355\n",
      "8.109383583068848\n",
      "6.770847320556641\n",
      "5.836953163146973\n",
      "7.282804489135742\n",
      "6.680734157562256\n",
      "7.413850784301758\n",
      "6.768106937408447\n",
      "6.697570323944092\n",
      "6.094171524047852\n",
      "6.471095561981201\n",
      "6.814108371734619\n",
      "8.988357543945312\n",
      "5.959596633911133\n",
      "7.76361083984375\n",
      "7.955788612365723\n",
      "7.849547863006592\n",
      "6.372339248657227\n",
      "7.224239826202393\n",
      "5.365438938140869\n",
      "7.052210807800293\n",
      "6.652589321136475\n",
      "7.206209182739258\n",
      "7.407094478607178\n",
      "7.118961811065674\n",
      "5.382530689239502\n",
      "8.094643592834473\n",
      "6.697505950927734\n",
      "7.88812255859375\n",
      "7.067236423492432\n",
      "5.263119220733643\n",
      "5.959074974060059\n",
      "6.693863868713379\n",
      "6.256936550140381\n",
      "6.872890949249268\n",
      "8.161314010620117\n",
      "5.289334774017334\n",
      "6.443451881408691\n",
      "7.372050762176514\n",
      "6.032281398773193\n",
      "7.549042224884033\n",
      "6.864811420440674\n",
      "6.6594672203063965\n",
      "6.187555313110352\n",
      "5.704341888427734\n",
      "7.286426067352295\n",
      "5.34385347366333\n",
      "6.8837761878967285\n",
      "7.930327415466309\n",
      "6.378931045532227\n",
      "7.557885646820068\n",
      "6.213971138000488\n",
      "5.779256343841553\n",
      "6.701328277587891\n",
      "7.5395965576171875\n",
      "6.01011323928833\n",
      "6.398210048675537\n",
      "5.368855953216553\n",
      "6.21635627746582\n",
      "7.418092250823975\n",
      "7.277143955230713\n",
      "7.5199408531188965\n",
      "7.292342185974121\n",
      "6.6349711418151855\n",
      "6.100396633148193\n",
      "5.996092796325684\n",
      "6.4695868492126465\n",
      "5.567414283752441\n",
      "5.912166118621826\n",
      "6.3095550537109375\n",
      "5.977556228637695\n",
      "7.31133508682251\n",
      "6.749348163604736\n",
      "4.8068389892578125\n",
      "5.142502784729004\n",
      "6.554009914398193\n",
      "6.8101606369018555\n",
      "8.060096740722656\n",
      "5.4314374923706055\n",
      "6.167068004608154\n",
      "6.672640800476074\n",
      "6.181323528289795\n",
      "5.4799723625183105\n",
      "5.307830810546875\n",
      "6.929403781890869\n",
      "5.316844940185547\n",
      "5.979116916656494\n",
      "6.421212196350098\n",
      "5.755197048187256\n",
      "6.308427333831787\n",
      "7.349857807159424\n",
      "5.610589504241943\n",
      "6.591254711151123\n",
      "6.102110862731934\n",
      "6.527836322784424\n",
      "6.779018878936768\n",
      "6.4011125564575195\n",
      "5.288815021514893\n",
      "6.191483974456787\n",
      "7.308173656463623\n",
      "5.536232948303223\n",
      "6.355053901672363\n",
      "7.202725887298584\n",
      "5.055682182312012\n",
      "6.061576843261719\n",
      "7.329807281494141\n",
      "7.74278450012207\n",
      "6.973013877868652\n",
      "6.194766521453857\n",
      "5.547595500946045\n",
      "5.2542266845703125\n",
      "4.886450290679932\n",
      "6.254723072052002\n",
      "4.584530830383301\n",
      "6.500265121459961\n",
      "6.785545825958252\n",
      "6.1528191566467285\n",
      "6.5197062492370605\n",
      "7.7036004066467285\n",
      "5.040857791900635\n",
      "4.724050998687744\n",
      "5.549619197845459\n",
      "6.404627799987793\n",
      "7.375412464141846\n",
      "6.805109977722168\n",
      "6.5323405265808105\n",
      "5.961357116699219\n",
      "5.325500965118408\n",
      "6.85811710357666\n",
      "4.867590427398682\n",
      "6.182168006896973\n",
      "5.538590908050537\n",
      "7.25268030166626\n",
      "4.692014694213867\n",
      "5.852901935577393\n",
      "6.213615417480469\n",
      "7.387576103210449\n",
      "5.802264213562012\n",
      "7.293552398681641\n",
      "7.045316219329834\n",
      "6.023037433624268\n",
      "5.44197416305542\n",
      "6.816893577575684\n",
      "6.225052833557129\n",
      "6.447576522827148\n",
      "5.912296772003174\n",
      "5.035290718078613\n",
      "6.822986602783203\n",
      "5.383571147918701\n",
      "6.9349684715271\n",
      "5.64413595199585\n",
      "5.300849914550781\n",
      "5.665816783905029\n",
      "6.111795425415039\n",
      "5.781726837158203\n",
      "6.51216459274292\n",
      "6.950692176818848\n",
      "6.870978355407715\n",
      "6.20657205581665\n",
      "5.632471084594727\n",
      "5.924872875213623\n",
      "6.947624206542969\n",
      "5.7261481285095215\n",
      "7.303789138793945\n",
      "6.410956859588623\n",
      "6.14469575881958\n",
      "6.452012538909912\n",
      "6.924924373626709\n",
      "7.599083423614502\n",
      "6.7189507484436035\n",
      "5.598670482635498\n",
      "5.471279621124268\n",
      "6.483240604400635\n",
      "6.838205814361572\n",
      "7.436802387237549\n",
      "6.984280586242676\n",
      "7.27692985534668\n",
      "5.979201316833496\n",
      "5.251873016357422\n",
      "6.352623462677002\n",
      "6.087140083312988\n",
      "7.0089874267578125\n",
      "6.941642761230469\n",
      "8.11176586151123\n",
      "8.403773307800293\n",
      "7.6296563148498535\n",
      "7.988282203674316\n",
      "7.873251438140869\n",
      "7.899036884307861\n",
      "7.507711410522461\n",
      "6.409951686859131\n",
      "6.17728328704834\n",
      "7.508401870727539\n",
      "7.03412389755249\n",
      "5.896759033203125\n",
      "7.536426544189453\n",
      "6.915973663330078\n",
      "7.837007999420166\n",
      "5.8463850021362305\n",
      "8.09084701538086\n",
      "8.520907402038574\n",
      "6.894617557525635\n",
      "7.223813533782959\n",
      "6.992647647857666\n",
      "7.126033306121826\n",
      "5.680670261383057\n",
      "7.090534210205078\n",
      "8.176167488098145\n",
      "6.018490791320801\n",
      "6.352689743041992\n",
      "7.854001045227051\n",
      "7.180957794189453\n",
      "7.040764808654785\n",
      "5.776238441467285\n",
      "6.497202396392822\n",
      "7.245663166046143\n",
      "6.818828105926514\n",
      "7.377751350402832\n",
      "7.576629638671875\n",
      "7.64706563949585\n",
      "7.254076957702637\n",
      "6.956527233123779\n",
      "6.904611110687256\n",
      "5.733633041381836\n",
      "4.752531051635742\n",
      "7.531887054443359\n",
      "7.35646390914917\n",
      "7.303524494171143\n",
      "5.789731979370117\n",
      "7.392685890197754\n",
      "6.6473236083984375\n",
      "6.8168864250183105\n",
      "8.778650283813477\n",
      "8.249608993530273\n",
      "7.22686767578125\n",
      "8.571677207946777\n",
      "7.845048427581787\n",
      "7.8140740394592285\n",
      "8.091533660888672\n",
      "8.377853393554688\n",
      "7.3601765632629395\n",
      "7.266949653625488\n",
      "9.756966590881348\n",
      "10.117659568786621\n",
      "7.377715110778809\n",
      "7.38897705078125\n",
      "7.8450541496276855\n",
      "6.667735576629639\n",
      "7.999963760375977\n",
      "7.753505706787109\n",
      "9.249283790588379\n",
      "5.234041690826416\n",
      "8.43841552734375\n",
      "9.797539710998535\n",
      "7.970661640167236\n",
      "6.511440753936768\n",
      "9.750182151794434\n",
      "7.963103771209717\n",
      "8.318685531616211\n",
      "8.903949737548828\n",
      "7.2438459396362305\n",
      "8.008759498596191\n",
      "6.959641933441162\n",
      "7.503759384155273\n",
      "9.999128341674805\n",
      "8.079353332519531\n",
      "9.197395324707031\n",
      "8.296069145202637\n",
      "10.03857421875\n",
      "8.720561027526855\n",
      "10.616460800170898\n",
      "12.156238555908203\n",
      "11.397790908813477\n",
      "10.894484519958496\n",
      "9.247405052185059\n",
      "8.729654312133789\n",
      "7.908536911010742\n",
      "10.625211715698242\n",
      "9.24009895324707\n",
      "8.595361709594727\n",
      "8.69510555267334\n",
      "9.07967472076416\n",
      "7.329998016357422\n",
      "10.125946998596191\n",
      "11.227228164672852\n",
      "8.417333602905273\n",
      "8.844264030456543\n",
      "8.545439720153809\n",
      "8.044591903686523\n",
      "8.987568855285645\n",
      "11.117724418640137\n",
      "8.219048500061035\n",
      "9.795764923095703\n",
      "10.872184753417969\n",
      "11.923099517822266\n",
      "6.917611122131348\n",
      "10.337305068969727\n",
      "10.039144515991211\n",
      "10.18150520324707\n",
      "9.358368873596191\n",
      "9.075199127197266\n",
      "10.794487953186035\n",
      "9.604217529296875\n",
      "9.618410110473633\n",
      "8.215758323669434\n",
      "8.744782447814941\n",
      "10.962603569030762\n",
      "9.092985153198242\n",
      "10.968177795410156\n",
      "13.033100128173828\n",
      "10.1288423538208\n",
      "8.666458129882812\n",
      "9.888543128967285\n",
      "11.766725540161133\n",
      "8.146175384521484\n",
      "10.552451133728027\n",
      "12.708191871643066\n",
      "10.988043785095215\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "10.988043785095215"
      ]
     },
     "execution_count": 211,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lri = []\n",
    "lossi = []\n",
    "\n",
    "for i in range(1000):\n",
    "    # forward pass\n",
    "    idx = torch.randint(0, X_train.shape[0], (64,)) # creating mini batches over random indices\n",
    "    emb = C[X_train[idx]].view(-1, EMB_DIM * BLOCK_SIZE)\n",
    "    h = torch.tanh(emb @ W1 + b1)\n",
    "    logits = h @ W2 + b2\n",
    "    loss = F.cross_entropy(logits, y_train[idx])\n",
    "    print(loss.item())\n",
    "\n",
    "    # zero grad\n",
    "    for p in params:\n",
    "        p.grad = None\n",
    "\n",
    "    # backward\n",
    "    loss.backward()\n",
    "\n",
    "    lr = lrs[i]\n",
    "    # update\n",
    "    for p in params:\n",
    "        p.data += -lr * p.grad\n",
    "\n",
    "    lri.append(lr)\n",
    "    lossi.append(loss.item())\n",
    "\n",
    "    # stepi.append(i)\n",
    "    # lossi.append(loss.log10().item())\n",
    "\n",
    "loss.item()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 213,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x120d04b80>]"
      ]
     },
     "execution_count": 213,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(lre, lossi)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "venv",
   "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": "3.9.16"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}