makemore_mlp.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch.nn.functional as F\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['emma', 'olivia', 'ava', 'isabella', 'sophia', 'charlotte', 'mia', 'amelia']"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "words  = open('names.txt', 'r').read().splitlines()\n",
    "words[:8]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "32033"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(words)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{1: 'a', 2: 'b', 3: 'c', 4: 'd', 5: 'e', 6: 'f', 7: 'g', 8: 'h', 9: 'i', 10: 'j', 11: 'k', 12: 'l', 13: 'm', 14: 'n', 15: 'o', 16: 'p', 17: 'q', 18: 'r', 19: 's', 20: 't', 21: 'u', 22: 'v', 23: 'w', 24: 'x', 25: 'y', 26: 'z', 0: '.'}\n"
     ]
    }
   ],
   "source": [
    "chars = sorted(list(set(''.join(words))))\n",
    "stoi = {s:i+1 for i,s in enumerate(chars)}\n",
    "stoi[\".\"] = 0\n",
    "itos = {i:s for s,i in stoi.items() }\n",
    "print(itos)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "block_size = 3\n",
    "X,Y = [],[]\n",
    "for w in words:\n",
    "    #print(w)\n",
    "    context = [0] * block_size\n",
    "    for ch in w + '.':\n",
    "        ix = stoi[ch]\n",
    "        X.append(context)\n",
    "        Y.append(ix)\n",
    "        context = context[1:]+ [ix]\n",
    "\n",
    "X = torch.tensor(X)\n",
    "Y = torch.tensor(Y)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(torch.Size([228146, 3]), torch.int64, torch.Size([228146]), torch.int64)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X.shape, X.dtype, Y.shape, Y.dtype"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "C = torch.randn((27, 2))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "emb = C[X]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "W1 = torch.randn((6,100))\n",
    "b1 = torch.randn(100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "h = torch.tanh(emb.view(-1,6) @ W1 + b1) # -1 identifies the underlying size aka 32, 64 etc of the tensor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "W2 = torch.randn((100, 27))\n",
    "b2 = torch.randn(27)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "logits = h @ W2 + b2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([228146, 27])"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "logits.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "counts = logits.exp()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "prob = counts/counts.sum(1, keepdims = True )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "ename": "IndexError",
     "evalue": "shape mismatch: indexing tensors could not be broadcast together with shapes [32], [228146]",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mIndexError\u001b[0m                                Traceback (most recent call last)",
      "Cell \u001b[1;32mIn[17], line 1\u001b[0m\n\u001b[1;32m----> 1\u001b[0m loss \u001b[38;5;241m=\u001b[39m \u001b[38;5;241m-\u001b[39m\u001b[43mprob\u001b[49m\u001b[43m[\u001b[49m\u001b[43mtorch\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43marange\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m32\u001b[39;49m\u001b[43m)\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mY\u001b[49m\u001b[43m]\u001b[49m\u001b[38;5;241m.\u001b[39mlog()\u001b[38;5;241m.\u001b[39mmean()\n\u001b[0;32m      2\u001b[0m loss\n",
      "\u001b[1;31mIndexError\u001b[0m: shape mismatch: indexing tensors could not be broadcast together with shapes [32], [228146]"
     ]
    }
   ],
   "source": [
    "loss = -prob[torch.arange(32), Y].log().mean()\n",
    "loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([182580, 3]) torch.Size([182580])\n",
      "torch.Size([22767, 3]) torch.Size([22767])\n",
      "torch.Size([22799, 3]) torch.Size([22799])\n"
     ]
    }
   ],
   "source": [
    "def build_dataset(words):\n",
    "    block_size = (\n",
    "        3  # context length: how many characters do we take to predict the next one?\n",
    "    )\n",
    "    X, Y = [], []\n",
    "    for w in words:\n",
    "\n",
    "        # print(w)\n",
    "        context = [0] * block_size\n",
    "        for ch in w + \".\":\n",
    "            ix = stoi[ch]\n",
    "            X.append(context)\n",
    "            Y.append(ix)\n",
    "            # print(''.join(itos[i] for i in context), ' --- >', itos[ix])\n",
    "            context = context[1:] + [ix]  # crop and append\n",
    "\n",
    "    X = torch.tensor(X)\n",
    "    Y = torch.tensor(Y)\n",
    "    print(X.shape, Y.shape)\n",
    "    return X, Y\n",
    "\n",
    "\n",
    "import random\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",
    "Xtr, Ytr = build_dataset(words[:n1])\n",
    "Xdev, Ydev = build_dataset(words[n1:n2])\n",
    "Xte, Yte = build_dataset(words[n2:])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(torch.Size([182580, 3]), torch.Size([182580]))"
      ]
     },
     "execution_count": 78,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Xtr.shape, Ytr.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [],
   "source": [
    "g = torch.Generator().manual_seed(2147483647) # for reproducibility\n",
    "C = torch. randn((27, 10), generator=g)\n",
    "W1 = torch.randn((30, 300), generator=g)\n",
    "b1 = torch.randn(300, generator=g)\n",
    "W2 = torch.randn((300, 27), generator=g)\n",
    "b2 = torch.randn(27, generator=g)\n",
    "parameters = [C, W1, b1, W2, b2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "17697"
      ]
     },
     "execution_count": 80,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sum(p.nelement() for p in parameters)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [],
   "source": [
    "for p in parameters:\n",
    "    p.requires_grad = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [],
   "source": [
    "lre = torch.linspace(-3,0,1000)\n",
    "lrs = 10 ** lre"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [],
   "source": [
    "lri = []\n",
    "lossi = []\n",
    "stepi = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.9795143604278564\n"
     ]
    }
   ],
   "source": [
    "\n",
    "\n",
    "for i in range(50000):\n",
    "    # minibatch construct\n",
    "    ix = torch.randint(0, Xtr.shape[0], (32,))\n",
    "\n",
    "    emb = C[Xtr[ix]]  # (32, 3, 2)\n",
    "    h = torch.tanh(emb.view(-1, 30) @ W1 + b1)  # (32, 100)\n",
    "    logits = h @ W2 + b2  # (32, 27)\n",
    "    loss = F.cross_entropy(logits, Ytr[ix])\n",
    "\n",
    "    for p in parameters:\n",
    "        p.grad = None\n",
    "        \n",
    "    loss.backward()\n",
    "    lr = 0.01\n",
    "    for p in parameters:\n",
    "        p.data += -lr * p.grad\n",
    "    \n",
    "    #lri.append(lre[i])\n",
    "    stepi.append(i)\n",
    "    lossi.append(loss.log10().item())\n",
    "print(loss.item())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x254eebcc3d0>]"
      ]
     },
     "execution_count": 93,
     "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(stepi, lossi)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor(2.1810, grad_fn=<NllLossBackward0>)"
      ]
     },
     "execution_count": 94,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "emb = C[Xtr]  # (32, 3, 2)\n",
    "h = torch.tanh(emb.view(-1, 30) @ W1 + b1)  # (32, 100)\n",
    "logits = h @ W2 + b2  # (32, 27)\n",
    "loss = F.cross_entropy(logits, Ytr)\n",
    "loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor(2.2180, grad_fn=<NllLossBackward0>)"
      ]
     },
     "execution_count": 95,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "emb = C[Xdev]  # (32, 3, 2)\n",
    "h = torch.tanh(emb.view(-1, 30) @ W1 + b1)  # (32, 100)\n",
    "logits = h @ W2 + b2  # (32, 27)\n",
    "loss = F.cross_entropy(logits, Ydev)\n",
    "loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x800 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(8,8))\n",
    "plt.scatter(C[:,0].data, C[:,1].data, s=200)\n",
    "for i in range(C.shape[0]):\n",
    "    plt.text(C[i,0].item(), C[i,1].item(), itos[i], ha=\"center\", va=\"center\", color='white')\n",
    "plt.grid('minor')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "mora.\n",
      "mayanniee.\n",
      "mad.\n",
      "rylla.\n",
      "emmadiendrari.\n",
      "adeerdieliah.\n",
      "milopaleigh.\n",
      "estanaraelyn.\n",
      "malquann.\n",
      "shurergahshie.\n",
      "trickentleigh.\n",
      "teusan.\n",
      "bented.\n",
      "ruyah.\n",
      "faeha.\n",
      "kaesh.\n",
      "sakyda.\n",
      "hal.\n",
      "calynn.\n",
      "una.\n"
     ]
    }
   ],
   "source": [
    "# sample from the model\n",
    "g = torch.Generator().manual_seed(2147483647 + 10)\n",
    "\n",
    "for _ in range(20):\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": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "pytgpu",
   "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": "3.11.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}