{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "adbd9598",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\Users\\Chris\\miniconda3\\envs\\deep_learning\\Lib\\site-packages\\torch\\utils\\_device.py:103: UserWarning: Aten Op fallback from XPU to CPU happends. This may have performance implications. If need debug the fallback ops please set environment variable `PYTORCH_DEBUG_XPU_FALLBACK=1`  (Triggered internally at C:\\actions-runner\\_work\\pytorch\\pytorch\\pytorch\\build\\xpu\\ATen\\RegisterXPU_0.cpp:54528.)\n",
      "  return func(*args, **kwargs)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH 0\n",
      "\tCurrent Loss: 8.951058387756348\n",
      "EPOCH 10\n",
      "\tCurrent Loss: 8.913984298706055\n",
      "EPOCH 20\n",
      "\tCurrent Loss: 8.911956787109375\n",
      "EPOCH 30\n",
      "\tCurrent Loss: 8.911856651306152\n",
      "EPOCH 40\n",
      "\tCurrent Loss: 8.911840438842773\n",
      "EPOCH 50\n",
      "\tCurrent Loss: 8.911835670471191\n",
      "EPOCH 60\n",
      "\tCurrent Loss: 8.911831855773926\n",
      "EPOCH 70\n",
      "\tCurrent Loss: 8.91179084777832\n",
      "EPOCH 80\n",
      "\tCurrent Loss: 8.899038314819336\n",
      "EPOCH 90\n",
      "\tCurrent Loss: 8.898558616638184\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[31m---------------------------------------------------------------------------\u001b[39m",
      "\u001b[31mKeyboardInterrupt\u001b[39m                         Traceback (most recent call last)",
      "\u001b[36mCell\u001b[39m\u001b[36m \u001b[39m\u001b[32mIn[1]\u001b[39m\u001b[32m, line 133\u001b[39m\n\u001b[32m    130\u001b[39m LOSS_HISTORY.append(loss)\n\u001b[32m    132\u001b[39m loss.backward()\n\u001b[32m--> \u001b[39m\u001b[32m133\u001b[39m \u001b[43moptimizer\u001b[49m\u001b[43m.\u001b[49m\u001b[43mstep\u001b[49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m    134\u001b[39m \u001b[38;5;66;03m# scheduler.step()\u001b[39;00m\n\u001b[32m    136\u001b[39m most_probable_tokens = torch.argmax(logits, \u001b[32m2\u001b[39m)\n",
      "\u001b[36mFile \u001b[39m\u001b[32mc:\\Users\\Chris\\miniconda3\\envs\\deep_learning\\Lib\\site-packages\\torch\\optim\\optimizer.py:516\u001b[39m, in \u001b[36mOptimizer.profile_hook_step.<locals>.wrapper\u001b[39m\u001b[34m(*args, **kwargs)\u001b[39m\n\u001b[32m    511\u001b[39m         \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[32m    512\u001b[39m             \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mRuntimeError\u001b[39;00m(\n\u001b[32m    513\u001b[39m                 \u001b[33mf\u001b[39m\u001b[33m\"\u001b[39m\u001b[38;5;132;01m{\u001b[39;00mfunc\u001b[38;5;132;01m}\u001b[39;00m\u001b[33m must return None or a tuple of (new_args, new_kwargs), but got \u001b[39m\u001b[38;5;132;01m{\u001b[39;00mresult\u001b[38;5;132;01m}\u001b[39;00m\u001b[33m.\u001b[39m\u001b[33m\"\u001b[39m\n\u001b[32m    514\u001b[39m             )\n\u001b[32m--> \u001b[39m\u001b[32m516\u001b[39m out = \u001b[43mfunc\u001b[49m\u001b[43m(\u001b[49m\u001b[43m*\u001b[49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43m*\u001b[49m\u001b[43m*\u001b[49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m    517\u001b[39m \u001b[38;5;28mself\u001b[39m._optimizer_step_code()\n\u001b[32m    519\u001b[39m \u001b[38;5;66;03m# call optimizer step post hooks\u001b[39;00m\n",
      "\u001b[36mFile \u001b[39m\u001b[32mc:\\Users\\Chris\\miniconda3\\envs\\deep_learning\\Lib\\site-packages\\torch\\optim\\optimizer.py:81\u001b[39m, in \u001b[36m_use_grad_for_differentiable.<locals>._use_grad\u001b[39m\u001b[34m(*args, **kwargs)\u001b[39m\n\u001b[32m     79\u001b[39m     torch.set_grad_enabled(\u001b[38;5;28mself\u001b[39m.defaults[\u001b[33m\"\u001b[39m\u001b[33mdifferentiable\u001b[39m\u001b[33m\"\u001b[39m])\n\u001b[32m     80\u001b[39m     torch._dynamo.graph_break()\n\u001b[32m---> \u001b[39m\u001b[32m81\u001b[39m     ret = \u001b[43mfunc\u001b[49m\u001b[43m(\u001b[49m\u001b[43m*\u001b[49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43m*\u001b[49m\u001b[43m*\u001b[49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m     82\u001b[39m \u001b[38;5;28;01mfinally\u001b[39;00m:\n\u001b[32m     83\u001b[39m     torch._dynamo.graph_break()\n",
      "\u001b[36mFile \u001b[39m\u001b[32mc:\\Users\\Chris\\miniconda3\\envs\\deep_learning\\Lib\\site-packages\\torch\\optim\\adam.py:247\u001b[39m, in \u001b[36mAdam.step\u001b[39m\u001b[34m(self, closure)\u001b[39m\n\u001b[32m    235\u001b[39m     beta1, beta2 = group[\u001b[33m\"\u001b[39m\u001b[33mbetas\u001b[39m\u001b[33m\"\u001b[39m]\n\u001b[32m    237\u001b[39m     has_complex = \u001b[38;5;28mself\u001b[39m._init_group(\n\u001b[32m    238\u001b[39m         group,\n\u001b[32m    239\u001b[39m         params_with_grad,\n\u001b[32m   (...)\u001b[39m\u001b[32m    244\u001b[39m         state_steps,\n\u001b[32m    245\u001b[39m     )\n\u001b[32m--> \u001b[39m\u001b[32m247\u001b[39m     \u001b[43madam\u001b[49m\u001b[43m(\u001b[49m\n\u001b[32m    248\u001b[39m \u001b[43m        \u001b[49m\u001b[43mparams_with_grad\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    249\u001b[39m \u001b[43m        \u001b[49m\u001b[43mgrads\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    250\u001b[39m \u001b[43m        \u001b[49m\u001b[43mexp_avgs\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    251\u001b[39m \u001b[43m        \u001b[49m\u001b[43mexp_avg_sqs\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    252\u001b[39m \u001b[43m        \u001b[49m\u001b[43mmax_exp_avg_sqs\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    253\u001b[39m \u001b[43m        \u001b[49m\u001b[43mstate_steps\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    254\u001b[39m \u001b[43m        \u001b[49m\u001b[43mamsgrad\u001b[49m\u001b[43m=\u001b[49m\u001b[43mgroup\u001b[49m\u001b[43m[\u001b[49m\u001b[33;43m\"\u001b[39;49m\u001b[33;43mamsgrad\u001b[39;49m\u001b[33;43m\"\u001b[39;49m\u001b[43m]\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    255\u001b[39m \u001b[43m        \u001b[49m\u001b[43mhas_complex\u001b[49m\u001b[43m=\u001b[49m\u001b[43mhas_complex\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    256\u001b[39m \u001b[43m        \u001b[49m\u001b[43mbeta1\u001b[49m\u001b[43m=\u001b[49m\u001b[43mbeta1\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    257\u001b[39m \u001b[43m        \u001b[49m\u001b[43mbeta2\u001b[49m\u001b[43m=\u001b[49m\u001b[43mbeta2\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    258\u001b[39m \u001b[43m        \u001b[49m\u001b[43mlr\u001b[49m\u001b[43m=\u001b[49m\u001b[43mgroup\u001b[49m\u001b[43m[\u001b[49m\u001b[33;43m\"\u001b[39;49m\u001b[33;43mlr\u001b[39;49m\u001b[33;43m\"\u001b[39;49m\u001b[43m]\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    259\u001b[39m \u001b[43m        \u001b[49m\u001b[43mweight_decay\u001b[49m\u001b[43m=\u001b[49m\u001b[43mgroup\u001b[49m\u001b[43m[\u001b[49m\u001b[33;43m\"\u001b[39;49m\u001b[33;43mweight_decay\u001b[39;49m\u001b[33;43m\"\u001b[39;49m\u001b[43m]\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    260\u001b[39m \u001b[43m        \u001b[49m\u001b[43meps\u001b[49m\u001b[43m=\u001b[49m\u001b[43mgroup\u001b[49m\u001b[43m[\u001b[49m\u001b[33;43m\"\u001b[39;49m\u001b[33;43meps\u001b[39;49m\u001b[33;43m\"\u001b[39;49m\u001b[43m]\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    261\u001b[39m \u001b[43m        \u001b[49m\u001b[43mmaximize\u001b[49m\u001b[43m=\u001b[49m\u001b[43mgroup\u001b[49m\u001b[43m[\u001b[49m\u001b[33;43m\"\u001b[39;49m\u001b[33;43mmaximize\u001b[39;49m\u001b[33;43m\"\u001b[39;49m\u001b[43m]\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    262\u001b[39m \u001b[43m        \u001b[49m\u001b[43mforeach\u001b[49m\u001b[43m=\u001b[49m\u001b[43mgroup\u001b[49m\u001b[43m[\u001b[49m\u001b[33;43m\"\u001b[39;49m\u001b[33;43mforeach\u001b[39;49m\u001b[33;43m\"\u001b[39;49m\u001b[43m]\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    263\u001b[39m \u001b[43m        \u001b[49m\u001b[43mcapturable\u001b[49m\u001b[43m=\u001b[49m\u001b[43mgroup\u001b[49m\u001b[43m[\u001b[49m\u001b[33;43m\"\u001b[39;49m\u001b[33;43mcapturable\u001b[39;49m\u001b[33;43m\"\u001b[39;49m\u001b[43m]\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    264\u001b[39m \u001b[43m        \u001b[49m\u001b[43mdifferentiable\u001b[49m\u001b[43m=\u001b[49m\u001b[43mgroup\u001b[49m\u001b[43m[\u001b[49m\u001b[33;43m\"\u001b[39;49m\u001b[33;43mdifferentiable\u001b[39;49m\u001b[33;43m\"\u001b[39;49m\u001b[43m]\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    265\u001b[39m \u001b[43m        \u001b[49m\u001b[43mfused\u001b[49m\u001b[43m=\u001b[49m\u001b[43mgroup\u001b[49m\u001b[43m[\u001b[49m\u001b[33;43m\"\u001b[39;49m\u001b[33;43mfused\u001b[39;49m\u001b[33;43m\"\u001b[39;49m\u001b[43m]\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    266\u001b[39m \u001b[43m        \u001b[49m\u001b[43mgrad_scale\u001b[49m\u001b[43m=\u001b[49m\u001b[38;5;28;43mgetattr\u001b[39;49m\u001b[43m(\u001b[49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[33;43m\"\u001b[39;49m\u001b[33;43mgrad_scale\u001b[39;49m\u001b[33;43m\"\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;28;43;01mNone\u001b[39;49;00m\u001b[43m)\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    267\u001b[39m \u001b[43m        \u001b[49m\u001b[43mfound_inf\u001b[49m\u001b[43m=\u001b[49m\u001b[38;5;28;43mgetattr\u001b[39;49m\u001b[43m(\u001b[49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[33;43m\"\u001b[39;49m\u001b[33;43mfound_inf\u001b[39;49m\u001b[33;43m\"\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;28;43;01mNone\u001b[39;49;00m\u001b[43m)\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    268\u001b[39m \u001b[43m        \u001b[49m\u001b[43mdecoupled_weight_decay\u001b[49m\u001b[43m=\u001b[49m\u001b[43mgroup\u001b[49m\u001b[43m[\u001b[49m\u001b[33;43m\"\u001b[39;49m\u001b[33;43mdecoupled_weight_decay\u001b[39;49m\u001b[33;43m\"\u001b[39;49m\u001b[43m]\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    269\u001b[39m \u001b[43m    \u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m    271\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m loss\n",
      "\u001b[36mFile \u001b[39m\u001b[32mc:\\Users\\Chris\\miniconda3\\envs\\deep_learning\\Lib\\site-packages\\torch\\optim\\optimizer.py:149\u001b[39m, in \u001b[36m_disable_dynamo_if_unsupported.<locals>.wrapper.<locals>.maybe_fallback\u001b[39m\u001b[34m(*args, **kwargs)\u001b[39m\n\u001b[32m    147\u001b[39m     \u001b[38;5;28;01mreturn\u001b[39;00m disabled_func(*args, **kwargs)\n\u001b[32m    148\u001b[39m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[32m--> \u001b[39m\u001b[32m149\u001b[39m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mfunc\u001b[49m\u001b[43m(\u001b[49m\u001b[43m*\u001b[49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43m*\u001b[49m\u001b[43m*\u001b[49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n",
      "\u001b[36mFile \u001b[39m\u001b[32mc:\\Users\\Chris\\miniconda3\\envs\\deep_learning\\Lib\\site-packages\\torch\\optim\\adam.py:949\u001b[39m, in \u001b[36madam\u001b[39m\u001b[34m(params, grads, exp_avgs, exp_avg_sqs, max_exp_avg_sqs, state_steps, foreach, capturable, differentiable, fused, grad_scale, found_inf, has_complex, decoupled_weight_decay, amsgrad, beta1, beta2, lr, weight_decay, eps, maximize)\u001b[39m\n\u001b[32m    946\u001b[39m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[32m    947\u001b[39m     func = _single_tensor_adam\n\u001b[32m--> \u001b[39m\u001b[32m949\u001b[39m \u001b[43mfunc\u001b[49m\u001b[43m(\u001b[49m\n\u001b[32m    950\u001b[39m \u001b[43m    \u001b[49m\u001b[43mparams\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    951\u001b[39m \u001b[43m    \u001b[49m\u001b[43mgrads\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    952\u001b[39m \u001b[43m    \u001b[49m\u001b[43mexp_avgs\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    953\u001b[39m \u001b[43m    \u001b[49m\u001b[43mexp_avg_sqs\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    954\u001b[39m \u001b[43m    \u001b[49m\u001b[43mmax_exp_avg_sqs\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    955\u001b[39m \u001b[43m    \u001b[49m\u001b[43mstate_steps\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    956\u001b[39m \u001b[43m    \u001b[49m\u001b[43mamsgrad\u001b[49m\u001b[43m=\u001b[49m\u001b[43mamsgrad\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    957\u001b[39m \u001b[43m    \u001b[49m\u001b[43mhas_complex\u001b[49m\u001b[43m=\u001b[49m\u001b[43mhas_complex\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    958\u001b[39m \u001b[43m    \u001b[49m\u001b[43mbeta1\u001b[49m\u001b[43m=\u001b[49m\u001b[43mbeta1\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    959\u001b[39m \u001b[43m    \u001b[49m\u001b[43mbeta2\u001b[49m\u001b[43m=\u001b[49m\u001b[43mbeta2\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    960\u001b[39m \u001b[43m    \u001b[49m\u001b[43mlr\u001b[49m\u001b[43m=\u001b[49m\u001b[43mlr\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    961\u001b[39m \u001b[43m    \u001b[49m\u001b[43mweight_decay\u001b[49m\u001b[43m=\u001b[49m\u001b[43mweight_decay\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    962\u001b[39m \u001b[43m    \u001b[49m\u001b[43meps\u001b[49m\u001b[43m=\u001b[49m\u001b[43meps\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    963\u001b[39m \u001b[43m    \u001b[49m\u001b[43mmaximize\u001b[49m\u001b[43m=\u001b[49m\u001b[43mmaximize\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    964\u001b[39m \u001b[43m    \u001b[49m\u001b[43mcapturable\u001b[49m\u001b[43m=\u001b[49m\u001b[43mcapturable\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    965\u001b[39m \u001b[43m    \u001b[49m\u001b[43mdifferentiable\u001b[49m\u001b[43m=\u001b[49m\u001b[43mdifferentiable\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    966\u001b[39m \u001b[43m    \u001b[49m\u001b[43mgrad_scale\u001b[49m\u001b[43m=\u001b[49m\u001b[43mgrad_scale\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    967\u001b[39m \u001b[43m    \u001b[49m\u001b[43mfound_inf\u001b[49m\u001b[43m=\u001b[49m\u001b[43mfound_inf\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    968\u001b[39m \u001b[43m    \u001b[49m\u001b[43mdecoupled_weight_decay\u001b[49m\u001b[43m=\u001b[49m\u001b[43mdecoupled_weight_decay\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m    969\u001b[39m \u001b[43m\u001b[49m\u001b[43m)\u001b[49m\n",
      "\u001b[36mFile \u001b[39m\u001b[32mc:\\Users\\Chris\\miniconda3\\envs\\deep_learning\\Lib\\site-packages\\torch\\optim\\adam.py:755\u001b[39m, in \u001b[36m_multi_tensor_adam\u001b[39m\u001b[34m(params, grads, exp_avgs, exp_avg_sqs, max_exp_avg_sqs, state_steps, grad_scale, found_inf, amsgrad, has_complex, beta1, beta2, lr, weight_decay, eps, maximize, capturable, differentiable, decoupled_weight_decay)\u001b[39m\n\u001b[32m    752\u001b[39m     torch._foreach_addcdiv_(device_params, device_exp_avgs, exp_avg_sq_sqrt)\n\u001b[32m    753\u001b[39m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[32m    754\u001b[39m     bias_correction1 = [\n\u001b[32m--> \u001b[39m\u001b[32m755\u001b[39m         \u001b[32m1\u001b[39m - beta1 ** \u001b[43m_get_value\u001b[49m\u001b[43m(\u001b[49m\u001b[43mstep\u001b[49m\u001b[43m)\u001b[49m \u001b[38;5;28;01mfor\u001b[39;00m step \u001b[38;5;129;01min\u001b[39;00m device_state_steps\n\u001b[32m    756\u001b[39m     ]\n\u001b[32m    757\u001b[39m     bias_correction2 = [\n\u001b[32m    758\u001b[39m         \u001b[32m1\u001b[39m - beta2 ** _get_value(step) \u001b[38;5;28;01mfor\u001b[39;00m step \u001b[38;5;129;01min\u001b[39;00m device_state_steps\n\u001b[32m    759\u001b[39m     ]\n\u001b[32m    761\u001b[39m     step_size = _stack_if_compiling([(lr / bc) * -\u001b[32m1\u001b[39m \u001b[38;5;28;01mfor\u001b[39;00m bc \u001b[38;5;129;01min\u001b[39;00m bias_correction1])\n",
      "\u001b[36mFile \u001b[39m\u001b[32mc:\\Users\\Chris\\miniconda3\\envs\\deep_learning\\Lib\\site-packages\\torch\\optim\\optimizer.py:96\u001b[39m, in \u001b[36m_get_value\u001b[39m\u001b[34m(x)\u001b[39m\n\u001b[32m     94\u001b[39m     \u001b[38;5;28;01mreturn\u001b[39;00m x\n\u001b[32m     95\u001b[39m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[32m---> \u001b[39m\u001b[32m96\u001b[39m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mx\u001b[49m\u001b[43m.\u001b[49m\u001b[43mitem\u001b[49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28misinstance\u001b[39m(x, torch.Tensor) \u001b[38;5;28;01melse\u001b[39;00m x\n",
      "\u001b[36mFile \u001b[39m\u001b[32mc:\\Users\\Chris\\miniconda3\\envs\\deep_learning\\Lib\\site-packages\\torch\\utils\\_device.py:103\u001b[39m, in \u001b[36mDeviceContext.__torch_function__\u001b[39m\u001b[34m(self, func, types, args, kwargs)\u001b[39m\n\u001b[32m    101\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m func \u001b[38;5;129;01min\u001b[39;00m _device_constructors() \u001b[38;5;129;01mand\u001b[39;00m kwargs.get(\u001b[33m\"\u001b[39m\u001b[33mdevice\u001b[39m\u001b[33m\"\u001b[39m) \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m:\n\u001b[32m    102\u001b[39m     kwargs[\u001b[33m\"\u001b[39m\u001b[33mdevice\u001b[39m\u001b[33m\"\u001b[39m] = \u001b[38;5;28mself\u001b[39m.device\n\u001b[32m--> \u001b[39m\u001b[32m103\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mfunc\u001b[49m\u001b[43m(\u001b[49m\u001b[43m*\u001b[49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43m*\u001b[49m\u001b[43m*\u001b[49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n",
      "\u001b[31mKeyboardInterrupt\u001b[39m: "
     ]
    }
   ],
   "source": [
    "import random\n",
    "import torch\n",
    "import pandas as pd\n",
    "from pathlib import Path\n",
    "import Project_Model.Libs.Embedder as Embedder\n",
    "import Project_Model.Libs.BPE as BPE\n",
    "import Project_Model.Libs.Transformer as Transformer\n",
    "import Project_Model.Libs.TorchShims as torch_shims\n",
    "\n",
    "# set a fixed seed\n",
    "torch.manual_seed(0)\n",
    "random.seed(0)\n",
    "DEVICE = torch_shims.get_default_device()\n",
    "torch.set_default_device(DEVICE)\n",
    "\n",
    "# set a default device\n",
    "\n",
    "# BPE Init\n",
    "VOCABULARY_PATH = Path(\"Assets/Model/toy_10/toy_dictionary.json\")\n",
    "SPECIAL_VOC = BPE.default_special_tokens()\n",
    "\n",
    "VOCABULARY = BPE.load_nanos_vocabulary(VOCABULARY_PATH)\n",
    "TOKENANO = BPE.TokeNanoCore(VOCABULARY, SPECIAL_VOC)\n",
    "\n",
    "\n",
    "# Constants\n",
    "TOKEN_SPACE_SIZE = TOKENANO.vocabulary_size + 1\n",
    "EMBEDDED_SIZE = 256\n",
    "FEED_FORWARD_DIM = EMBEDDED_SIZE * 4\n",
    "ATTENTION_HEADS = 4\n",
    "SENTENCE_LENGTH = 256\n",
    "NUMBER_OF_BLOCKS = 2\n",
    "MAX_EPOCHS = int(1e3)\n",
    "\n",
    "\n",
    "# Model Init\n",
    "ENCODER_EMBEDDER = Embedder.NanoSocratesEmbedder(TOKEN_SPACE_SIZE, EMBEDDED_SIZE)\n",
    "DECODER_EMBEDDER = Embedder.NanoSocratesEmbedder(TOKEN_SPACE_SIZE, EMBEDDED_SIZE)\n",
    "\n",
    "ENCODERS = [\n",
    "    Transformer.Encoder(EMBEDDED_SIZE, FEED_FORWARD_DIM, ATTENTION_HEADS)\n",
    "] * NUMBER_OF_BLOCKS\n",
    "ENCODER = torch.nn.Sequential(*ENCODERS)\n",
    "\n",
    "DECODERS = [\n",
    "    Transformer.Decoder(EMBEDDED_SIZE, FEED_FORWARD_DIM, ATTENTION_HEADS)\n",
    "] * NUMBER_OF_BLOCKS\n",
    "DECODER = torch.nn.Sequential(*DECODERS)\n",
    "\n",
    "DETOKENER = Transformer.DeToken(EMBEDDED_SIZE, TOKEN_SPACE_SIZE)\n",
    "\n",
    "PAD_TOKEN = TOKENANO.encode(\"<PAD>\")[0]\n",
    "END_TOKEN = TOKENANO.encode(\"<END>\")[0]\n",
    "\n",
    "\n",
    "# Load CSV\n",
    "TOY_DATASET_PATH = Path(\"Assets/Dataset/1-hop/toy/rdf_text.csv\")\n",
    "\n",
    "TOY_DATASET = pd.read_csv(TOY_DATASET_PATH)\n",
    "\n",
    "TOY_BATCH_INPUT_LIST: list[list[int]] = []\n",
    "TOY_BATCH_PADDING_LIST: list[list[bool]] = []\n",
    "TOY_BATCH_TARGET_LIST: list[list[int]] = []\n",
    "TOY_BATCH_DECODER_DEFAULT: list[list[int]]= []\n",
    "\n",
    "\n",
    "for index, row in TOY_DATASET.iterrows():\n",
    "\n",
    "    RDFs: str = row[\"RDFs\"]\n",
    "    Abstract: str = row[\"Abstract\"]\n",
    "\n",
    "    input_tokens = TOKENANO.encode(RDFs)\n",
    "    output_tokens = TOKENANO.encode(Abstract)\n",
    "    decoder_default_tokens = TOKENANO.encode(\"<SOS>\")\n",
    "\n",
    "    input_tokens, padding = Transformer.normalize_sequence(\n",
    "        input_tokens, SENTENCE_LENGTH, PAD_TOKEN, END_TOKEN\n",
    "    )\n",
    "    output_tokens, _ = Transformer.normalize_sequence(\n",
    "        output_tokens, SENTENCE_LENGTH, PAD_TOKEN, END_TOKEN\n",
    "    )\n",
    "    decoder_default_tokens, _ = Transformer.normalize_sequence(\n",
    "        decoder_default_tokens, SENTENCE_LENGTH, PAD_TOKEN, END_TOKEN\n",
    "    )\n",
    "\n",
    "    TOY_BATCH_INPUT_LIST.append(input_tokens)\n",
    "    TOY_BATCH_PADDING_LIST.append(padding)\n",
    "    TOY_BATCH_TARGET_LIST.append(output_tokens)\n",
    "    TOY_BATCH_DECODER_DEFAULT.append(decoder_default_tokens)\n",
    "\n",
    "# Training loop\n",
    "LOSS_HISTORY = []\n",
    "NANOSOCRATES = torch.nn.ModuleList([\n",
    "    ENCODER_EMBEDDER,\n",
    "    ENCODER,\n",
    "    DECODER_EMBEDDER,\n",
    "    DECODER,\n",
    "    DETOKENER\n",
    "])\n",
    "cross_entropy = torch.nn.CrossEntropyLoss()\n",
    "optimizer = torch.optim.AdamW(NANOSOCRATES.parameters())\n",
    "# scheduler = torch.optim.lr_scheduler.LRScheduler(optimizer)\n",
    "last_loss = 0\n",
    "current_epoch = 0\n",
    "\n",
    "while current_epoch < MAX_EPOCHS:\n",
    "\n",
    "    optimizer.zero_grad()\n",
    "\n",
    "    INPUT_LIST = TOY_BATCH_INPUT_LIST[:]\n",
    "    TARGET_LIST = TOY_BATCH_TARGET_LIST[:]\n",
    "    # Transform target into logits\n",
    "    target_logits = torch.tensor(TOY_BATCH_TARGET_LIST[:])\n",
    "    DECODER_DEFAULT_LIST = TOY_BATCH_DECODER_DEFAULT[:]\n",
    "    PADDINGS = torch.tensor(TOY_BATCH_PADDING_LIST, dtype=torch.bool)\n",
    "    ENCODER_INPUTS = ENCODER_EMBEDDER(INPUT_LIST)\n",
    "    DECODER_INPUTS = DECODER_EMBEDDER(DECODER_DEFAULT_LIST)\n",
    "\n",
    "    for _ in range(0, SENTENCE_LENGTH):\n",
    "\n",
    "        optimizer.zero_grad()\n",
    "\n",
    "        \n",
    "\n",
    "        encoder_output, _ = ENCODER((ENCODER_INPUTS, PADDINGS))\n",
    "\n",
    "        decoder_output, _, _, _ = DECODER(\n",
    "            (DECODER_INPUTS, encoder_output, encoder_output, None)\n",
    "        )\n",
    "\n",
    "        logits: torch.Tensor = DETOKENER(decoder_output)\n",
    "        logits = logits.permute(0, 2, 1)\n",
    "\n",
    "\n",
    "        loss: torch.Tensor= cross_entropy(logits, target_logits)\n",
    "        last_loss = loss\n",
    "        LOSS_HISTORY.append(loss)\n",
    "\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "        # scheduler.step()\n",
    "\n",
    "        most_probable_tokens = torch.argmax(logits, 2)\n",
    "\n",
    "    if current_epoch % 10 == 0:\n",
    "\n",
    "        print(f\"EPOCH {current_epoch}\\n\\tCurrent Loss: {last_loss}\")\n",
    "\n",
    "    current_epoch += 1\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "deep_learning",
   "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": "3.13.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}