CoreGen.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "16c969e2",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import torch.nn.functional as F\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "064c6b63",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['emma', 'olivia', 'ava', 'isabella', 'sophia', 'charlotte', 'mia', 'amelia']"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "words = open(\"names.txt\",\"r\").read().splitlines()\n",
    "words[:8]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "d250f360",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "32033"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(words)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "7a20a811",
   "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": [
    "#build the vocabulary of characters and mappings to/form integers\n",
    "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": 5,
   "id": "eb18b7b1",
   "metadata": {},
   "outputs": [],
   "source": [
    "#build the dataset\n",
    "    \n",
    "block_size = 3\n",
    "X,Y = [],[]\n",
    "for w in words:\n",
    "    #print(\"\\n\\n\"+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]\n",
    "        \n",
    "X = torch.tensor(X)\n",
    "Y = torch.tensor(Y)\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "0ef6efc4",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([228146, 3])"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "02edfc33",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([228146])"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Y.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "039d9b21",
   "metadata": {},
   "outputs": [],
   "source": [
    "#build the dataset\n",
    " \n",
    "def build_dataset(words):\n",
    "    block_size = 3\n",
    "    X,Y = [],[]\n",
    "    for w in words:\n",
    "        #print(\"\\n\\n\"+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]\n",
    "\n",
    "    X = torch.tensor(X)\n",
    "    Y = torch.tensor(Y)\n",
    "    return X,Y\n",
    "\n",
    "import random\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": 9,
   "id": "899d0dc2",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[-0.3180, -1.8183],\n",
       "        [-0.3628,  1.0987],\n",
       "        [ 0.1184, -0.1279],\n",
       "        [ 0.2544, -1.5219],\n",
       "        [ 0.2560, -0.4845],\n",
       "        [-0.4812, -0.5172],\n",
       "        [-0.2158, -1.3272],\n",
       "        [ 0.4478, -0.8338],\n",
       "        [ 0.8250,  0.9677],\n",
       "        [-0.1496,  0.8728],\n",
       "        [-0.4391,  0.2419],\n",
       "        [ 0.4167, -0.7212],\n",
       "        [-1.4451, -0.4055],\n",
       "        [-0.6108,  0.5628],\n",
       "        [-0.7402, -2.2544],\n",
       "        [-1.4550, -0.0243],\n",
       "        [ 0.8448,  1.0562],\n",
       "        [ 0.2020,  0.0203],\n",
       "        [-0.2973,  0.1061],\n",
       "        [ 0.3426,  0.7916],\n",
       "        [ 1.0461,  0.3871],\n",
       "        [ 0.6111, -1.0064],\n",
       "        [-0.2679,  0.9292],\n",
       "        [ 0.7866,  0.3807],\n",
       "        [ 0.4841, -1.4623],\n",
       "        [ 0.2188,  0.3872],\n",
       "        [-2.0868, -0.9311]])"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "C = torch.randn((27,2))\n",
    "C"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "82bbbbfc",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([-0.4812, -0.5172])"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "C[5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "c244a3f8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([-0.4812, -0.5172])"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "F.one_hot(torch.tensor(5), num_classes=27).float() @ C"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "7eb0689d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([228146, 3, 2])"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "emb = C[X]\n",
    "emb.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "d8cf145a",
   "metadata": {},
   "outputs": [],
   "source": [
    "W1 = torch.randn((6,100))\n",
    "b1 = torch.randn(100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "f9100342",
   "metadata": {},
   "outputs": [],
   "source": [
    "h = torch.tanh(emb.view(-1,6) @W1+b1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "bce8036d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[-0.9959,  0.9783,  0.9822,  ..., -0.4461,  0.5109, -0.0633],\n",
       "        [-0.9956,  0.9669,  0.9963,  ..., -0.9907, -0.4197,  0.7057],\n",
       "        [-0.8344,  0.9853,  0.9946,  ..., -0.9996, -0.8855,  0.9673],\n",
       "        ...,\n",
       "        [-0.2707,  1.0000,  0.9989,  ..., -0.9996,  0.5803,  0.9965],\n",
       "        [-1.0000,  0.9995, -0.9991,  ..., -0.9999, -0.7327,  0.9493],\n",
       "        [ 0.7914,  0.0898,  0.9809,  ...,  0.8900,  0.8474,  0.2724]])"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "h"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "90c98e7b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([228146, 100])"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "h.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "72b1024d",
   "metadata": {},
   "outputs": [],
   "source": [
    "W2 = torch.randn((100,27))\n",
    "b2 = torch.randn(27)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "e03a242e",
   "metadata": {},
   "outputs": [],
   "source": [
    "logits = h @ W2 + b2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "2398dd3e",
   "metadata": {},
   "outputs": [],
   "source": [
    "counts = logits.exp()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "4b6c9042",
   "metadata": {},
   "outputs": [],
   "source": [
    "prob = counts/counts.sum(1, keepdims=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "d8bf3c79",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([228146, 27])"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "prob.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "7cfb2dde",
   "metadata": {},
   "outputs": [],
   "source": [
    "#Now made respectable"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "b09be3ba",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(torch.Size([182625, 3]), torch.Size([182625]))"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Xtr.shape, Ytr.shape # dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "998769a6",
   "metadata": {},
   "outputs": [],
   "source": [
    "g = torch.Generator().manual_seed(2147483647)\n",
    "C = torch.randn((27,10), generator=g)\n",
    "W1 = torch.randn((30,500),generator=g)\n",
    "b1 = torch.randn(500,generator=g)\n",
    "W2 = torch.randn((500,27),generator=g)\n",
    "b2 = torch.randn(27,generator=g)\n",
    "parameters = [C, W1, b1, W2, b2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "670f504c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "29297"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sum(p.nelement() for p in parameters)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "c7144e58",
   "metadata": {},
   "outputs": [],
   "source": [
    "for p in parameters:\n",
    "    p.requires_grad = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "772a202d",
   "metadata": {},
   "outputs": [],
   "source": [
    "lre = torch.linspace(-3,0,1000)\n",
    "lrs = 10**lre"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "d10b80a3",
   "metadata": {},
   "outputs": [],
   "source": [
    "lri=[]\n",
    "lossi=[]\n",
    "stepi = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "e365a153",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2.65606427192688\n"
     ]
    }
   ],
   "source": [
    "\n",
    "for i in range(100000):\n",
    "    #print(i,end=\", \")\n",
    "    #forward pass\n",
    "    ix = torch.randint(0,Xtr.shape[0],(32,))\n",
    "    emb = C[Xtr[ix]] # (32,3,2)\n",
    "    h = torch.tanh(emb.view(-1,30) @ W1 + b1)\n",
    "    logits = h @ W2 + b2\n",
    "    #counts = logits.exp()\n",
    "    #prob = counts/counts.sum(1, keepdims=True)\n",
    "    #loss = -prob[torch.arange(32), Y].log().mean()\n",
    "    loss = F.cross_entropy(logits,Ytr[ix])\n",
    "    #print(loss.item())\n",
    "    #backward pass\n",
    "    for p in parameters:\n",
    "        p.grad=None\n",
    "    loss.backward()\n",
    "\n",
    "    #update\n",
    "    #lr = lrs[i]\n",
    "    lr = 0.05\n",
    "    for p in parameters:\n",
    "        p.data += -lr*p.grad\n",
    "        \n",
    "    #track stats\n",
    "    #lri.append(lre[i])\n",
    "    stepi.append(i)\n",
    "    lossi.append(loss.log10().item())\n",
    "    \n",
    "print(loss.item())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "f47bc3f3",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor(2.3755, grad_fn=<NllLossBackward0>)"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "emb = C[Xtr]\n",
    "h = torch.tanh(emb.view(-1,30)@W1 + b1)\n",
    "logits = h @ W2 + b2\n",
    "loss = F.cross_entropy(logits, Ytr)\n",
    "loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "a535b2e2",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor(2.4723, grad_fn=<NllLossBackward0>)"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "emb = C[Xdev]\n",
    "h = torch.tanh(emb.view(-1,30)@W1 + b1)\n",
    "logits = h @ W2 + b2\n",
    "loss = F.cross_entropy(logits, Ydev)\n",
    "loss\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "58b4c744",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x1ea9978e908>]"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(stepi,lossi)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "9ab35813",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tarchla.\n",
      "len.\n",
      "floritd.\n",
      "ernet.\n",
      "leanit.\n",
      "arettia.\n",
      "ginleeyancir.\n",
      "awo.\n",
      "ana.\n",
      "hile.\n",
      "yamingyelian.\n",
      "sarie.\n",
      "flanisa.\n",
      "garabe.\n",
      "haryqobenjiliyann.\n",
      "lan.\n",
      "olaovi.\n",
      "naniklieaniellah.\n",
      "betty.\n",
      "geabitana.\n",
      "solaraliferettelleeliya.\n",
      "brysto.\n",
      "bmiga.\n",
      "elan.\n",
      "haurileyna.\n",
      "gaylie.\n",
      "arolyn.\n",
      "brayla.\n",
      "jianna.\n",
      "seva.\n",
      "lenie.\n",
      "metom.\n",
      "eleylin.\n",
      "bow.\n",
      "maria.\n",
      "sya.\n",
      "paylain.\n",
      "rhan.\n",
      "sahazley.\n",
      "garistia.\n",
      "emin.\n",
      "quilin.\n",
      "lharlutabrittara.\n",
      "sandolphan.\n",
      "gyana.\n",
      "evik.\n",
      "huar.\n",
      "ima.\n",
      "pariy.\n",
      "zeyliyah.\n"
     ]
    }
   ],
   "source": [
    "# sample from the model\n",
    "g = torch.Generator().manual_seed(2147483647+2990)\n",
    "#g = torch.Generator().manual_seed(10)\n",
    "for _ in range(50):\n",
    "    out=[]\n",
    "    context = [0] * block_size\n",
    "    while True:\n",
    "        emb = C[torch.tensor([context])]\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",
    "    print(''.join(itos[i] for i in out))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "cfc304b2",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fce1180f",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "84d6ca76",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a3e5832b",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.7.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}