play.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "272d6dce890f45668ecb1c4fcc04a21f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "log.txt:   0%|          | 0.00/1.22M [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "064103651c544574b4caadde6aa77841",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "model_56000.pt:   0%|          | 0.00/397M [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'./pretrained/model_56000.pt'"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from huggingface_hub import hf_hub_download\n",
    "\n",
    "hf_hub_download(repo_id=\"eastwind/gpt2-audio-tiny-sherlock-100k-overfit\", filename=\"log.txt\", local_dir='./pretrained')\n",
    "hf_hub_download(repo_id=\"eastwind/gpt2-audio-tiny-sherlock-100k-overfit\", filename=\"model_56000.pt\", local_dir='./pretrained')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "val = 'validation'\n",
    "train = 'train'\n",
    "\n",
    "with open('./pretrained/log.txt') as f:\n",
    "    log = [[eval(y) for y in x.strip().split(' ')] for x in f.readlines()]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "val_x = [x[0] for x in log if x[1] == val]\n",
    "val_y = [x[-1] for x in log if x[1] == val]\n",
    "\n",
    "train_x = [x[0] for x in log if x[1] == train]\n",
    "train_y = [x[-1] for x in log if x[1] == train]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "\n",
    "# Data from the provided image for validation and training sets\n",
    "val_data = {\n",
    "    \"step\": val_x,\n",
    "    \"loss\": val_y\n",
    "}\n",
    "\n",
    "train_data = {\n",
    "    \"step\": train_x,\n",
    "    \"loss\": train_y\n",
    "}\n",
    "\n",
    "# Create DataFrames\n",
    "val_df = pd.DataFrame(val_data)\n",
    "train_df = pd.DataFrame(train_data)\n",
    "\n",
    "# Plotting step vs loss for both validation and training sets\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.plot(train_df['step'], train_df['loss'], label='Training Loss', color='orange', linewidth=1)\n",
    "plt.plot(val_df['step'], val_df['loss'], label='Validation Loss', color='green', linewidth=1)\n",
    "plt.xlabel('Step')\n",
    "plt.ylabel('Loss')\n",
    "plt.title('Loss over Time')\n",
    "# plt.xlim([0, 2000])\n",
    "plt.legend()\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "step    450.0000\n",
       "loss      5.9357\n",
       "Name: 18, dtype: float64"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "val_df.iloc[val_y.index(min(val_y))]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from gpt2_infer import GPT, GPTConfig\n",
    "import torch, torchaudio\n",
    "from collections import OrderedDict\n",
    "import numpy as np\n",
    "from tqdm import tqdm\n",
    "from torch.nn import functional as F\n",
    "from speech_tokenizer import SpeechTokenizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "device = \"cpu\"\n",
    "if torch.cuda.is_available():\n",
    "    device = \"cuda\"\n",
    "elif hasattr(torch.backends, \"mps\") and torch.backends.mps.is_available():\n",
    "    device = \"mps\"\n",
    "print(f\"using device: {device}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = GPT(GPTConfig())\n",
    "\n",
    "original_state_dict = torch.load('./pretrained/model_05000.pt', map_location=torch.device('cpu'))\n",
    "\n",
    "# Corrected state dictionary\n",
    "state_dict = {\n",
    "    'model': OrderedDict([\n",
    "        (key.replace('_orig_mod.', ''), value) for key, value in original_state_dict['model'].items()\n",
    "    ]),\n",
    "    'config': original_state_dict['config'],\n",
    "    'step': original_state_dict['step'],\n",
    "    'val_loss': original_state_dict['val_loss']\n",
    "}\n",
    "\n",
    "model.load_state_dict(state_dict['model'])\n",
    "model.eval()\n",
    "model.to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "unseen = [4097, 547, 426, 2825, 1441, 2209, 1300, 161, 4097, 1646]\n",
    "seperator = 4097\n",
    "\n",
    "num_return_sequences = 4\n",
    "max_length = 1024\n",
    "\n",
    "tokens = [seperator]\n",
    "# tokens = unseen\n",
    "tokens = torch.tensor(tokens, dtype=torch.long)\n",
    "tokens = tokens.unsqueeze(0).repeat(num_return_sequences, 1)\n",
    "xgen = tokens.to(device)\n",
    "sample_rng = torch.Generator(device=device)\n",
    "sample_rng.manual_seed(42)\n",
    "\n",
    "output_tokens = []\n",
    "\n",
    "# Initialize the progress bar\n",
    "with tqdm(total=max_length) as pbar:\n",
    "    # Set initial progress\n",
    "    pbar.update(xgen.size(1))\n",
    "    \n",
    "    while xgen.size(1) < max_length:\n",
    "        # forward the model to get the logits\n",
    "        with torch.no_grad():\n",
    "            logits, loss = model(xgen) # (B, T, vocab_size)\n",
    "            # take the logits at the last position\n",
    "            logits = logits[:, -1, :] # (B, vocab_size)\n",
    "            # get the probabilities\n",
    "            probs = F.softmax(logits, dim=-1)\n",
    "            # do top-k sampling of 50 (huggingface pipeline default)\n",
    "            # topk_probs here becomes (5, 50), topk_indices is (5, 50)\n",
    "            topk_probs, topk_indices = torch.topk(probs, 50, dim=-1)\n",
    "            # select a token from the top-k probabilities\n",
    "            # note: multinomial does not demand the input to sum to 1\n",
    "            ix = torch.multinomial(topk_probs, 1, generator=sample_rng) # (B, 1)\n",
    "            # gather the corresponding indices\n",
    "            xcol = torch.gather(topk_indices, -1, ix) # (B, 1)\n",
    "            # append to the sequence\n",
    "            xgen = torch.cat((xgen, xcol), dim=1)\n",
    "\n",
    "            pbar.update(1)  # Update by the number of new tokens added\n",
    "    # print the generated text\n",
    "    for i in range(num_return_sequences):\n",
    "        tokens = xgen[i, :max_length].tolist()\n",
    "        output_tokens.append(tokens)\n",
    "\n",
    "with torch.no_grad():\n",
    "    del xgen\n",
    "    torch.cuda.empty_cache()\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "tokenizer = SpeechTokenizer(device=device)\n",
    "\n",
    "def find_last_instance_of_seperator(lst, element=4097):\n",
    "    reversed_list = lst[::-1]\n",
    "    try:\n",
    "        reversed_index = reversed_list.index(element)\n",
    "        return len(lst) - 1 - reversed_index\n",
    "    except ValueError:\n",
    "        raise ValueError"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "audio_out = tokenizer.decode(np.array([output_tokens[0][:find_last_instance_of_seperator(output_tokens[0]) + 1]]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[4097, 3919, 426, 2825, 1391, 4081, 3741, 1755, 4097, 717, 690, 2486, 1667, 1029, 3651, 161, 4097, 3919, 4081, 2370, 161, 690, 429, 1811, 4097, 3718, 576, 1781, 2172, 1404, 161, 953, 4097, 3718, 576, 2370, 2448, 1404, 161, 3830, 4097, 3718, 576, 3525, 2014, 576, 3072, 1337, 4097, 98, 241, 1135, 2471, 3402, 442, 3606, 4097, 2417, 1046, 754, 427, 316, 765, 3971, 4097, 2622, 2431, 1711, 1117, 3638, 2874, 3195, 4097, 117, 3795, 2430, 1602, 3866, 3182, 1775, 4097, 71, 2332, 2217, 424, 1551, 891, 3147, 4097, 3499, 1233, 882, 745, 3491, 756, 2544, 4097, 3897, 1943, 442, 279, 2129, 1924, 3486, 4097, 1726, 3347, 1462, 1248, 2444, 490, 2514, 4097, 3026, 3785, 3509, 353, 248, 55, 802, 4097, 2793, 3686, 3934, 2167, 1094, 2167, 3450, 4097, 70, 1148, 690, 2902, 1024, 1804, 2622, 4097, 2670, 3641, 2903, 2692, 1897, 329, 3840, 4097, 818, 841, 3775, 1542, 1397, 4049, 3073, 4097, 3872, 4006, 1588, 3693, 449, 1063, 1289, 4097, 117, 3316, 2750, 442, 1862, 3463, 2666, 4097, 3872, 2439, 15, 838, 3016, 2614, 845, 4097, 1212, 2988, 2504, 1016, 286, 1094, 3253, 4097, 3562, 2371, 1679, 1302, 2506, 626, 1762, 4097, 1909, 3822, 1311, 1024, 3759, 4066, 3486, 4097, 2561, 3693, 3681, 4032, 2939, 2194, 1374, 4097, 3919, 3779, 1678, 426, 1868, 100, 100, 4097, 2669, 2226, 314, 3538, 1459, 4069, 661, 4097, 2622, 431, 583, 3685, 1723, 2328, 595, 4097, 2862, 3662, 1913, 3070, 1878, 921, 2461, 4097, 3941, 2015, 259, 2717, 1997, 603, 2011, 4097, 3718, 1943, 3920, 2008, 1983, 1520, 2394, 4097, 3611, 1900, 3741, 3525, 1003, 1527, 869, 4097, 1936, 3866, 3802, 934, 223, 2850, 2871, 4097, 2114, 1843, 3392, 2172, 1499, 3431, 1889, 4097, 602, 71, 579, 424, 2609, 1513, 1278, 4097, 3872, 2771, 2936, 839, 3924, 3649, 946, 4097, 1075, 354, 2556, 2068, 3814, 1636, 1636, 4097, 199, 3107, 1444, 2957, 230, 2458, 2331, 4097, 3112, 2620, 894, 2430, 220, 3007, 219, 4097, 2622, 3816, 3229, 879, 2882, 2929, 3175, 4097, 3586, 2887, 2430, 244, 805, 3724, 1413, 4097, 3933, 2310, 501, 3919, 1024, 3302, 1145, 4097, 865, 1148, 2179, 3264, 1170, 1562, 4032, 4097, 2290, 1114, 331, 2520, 3487, 1020, 2917, 4097, 2291, 1344, 1633, 591, 496, 325, 2945, 4097, 2033, 731, 1907, 1689, 1132, 2697, 3918, 4097, 3531, 3066, 2839, 2457, 3047, 1811, 847, 4097, 117, 639, 3785, 4004, 3364, 2718, 859, 4097, 2125, 3510, 3514, 4006, 667, 3893, 3521, 4097, 4062, 64, 1387, 2688, 3686, 146, 1656, 4097, 419, 3317, 3076, 678, 1450, 2586, 1590, 4097, 2480, 3360, 2846, 1665, 3998, 1005, 511, 4097, 856, 1134, 416, 2554, 3899, 2216, 2675, 4097, 731, 775, 2813, 854, 4050, 1838, 1322, 4097, 2114, 3123, 3028, 1278, 1404, 916, 3431, 4097, 2114, 2110, 3741, 883, 690, 1295, 3759, 4097, 2114, 1029, 1889, 1776, 1843, 1391, 1873, 4097, 692, 2072, 3960, 2275, 1815, 1477, 2458, 4097, 541, 1829, 916, 1158, 3107, 1811, 3155, 4097, 3103, 1803, 3803, 897, 1699, 914, 4072, 4097, 3263, 1805, 2816, 515, 3555, 3421, 1726, 4097, 3352, 3180, 3951, 1321, 1339, 1202, 1985, 4097, 3455, 2246, 1652, 1444, 2080, 1395, 950]\n"
     ]
    }
   ],
   "source": [
    "print(list(np.load('./data/sherlock_train_02.npy')[:512]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "assisted_search",
   "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": "3.11.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}