detect_gpt_trial.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "55c46f33-ff3f-4f05-aed0-02358d314665",
   "metadata": {},
   "outputs": [],
   "source": [
    "import time\n",
    "import re\n",
    "\n",
    "import numpy as np\n",
    "import transformers\n",
    "import torch\n",
    "\n",
    "DEVICE = 'mps'\n",
    "BASE_MODEL_NAME = 'gpt2-medium'\n",
    "MASK_FILLING_MODEL_NAME = 't5-large'\n",
    "PROMPT_TOKENS = 30"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "48acf3d0-4121-4ddf-b185-5f0952262100",
   "metadata": {},
   "outputs": [],
   "source": [
    "real_text = \\\n",
    "''''Maj Richard Scott, 40, is accused of driving at speeds of up to 95mph (153km/h) in bad weather before the smash \n",
    "on a B-road in Wiltshire. Gareth Hicks, 24, suffered fatal injuries when the van he was asleep in was hit by Mr Scott\\'s \n",
    "Audi A6. Maj Scott denies a charge of causing death by careless driving. Prosecutor Charles Gabb alleged the defendant, \n",
    "from Green Lane in Shepperton, Surrey, had crossed the carriageway of the 60mph-limit B390 in Shrewton near Amesbury. \n",
    "The weather was \"awful\" and there was strong wind and rain, he told jurors. He said Mr Scott\\'s car was described as \n",
    "\"twitching\" and \"may have been aquaplaning\" before striking the first vehicle; a BMW driven by Craig Reed. Mr Scott\\'s \n",
    "Audi then returned to his side of the road but crossed the carriageway again before colliding head-on with a Ford Transit \n",
    "van in which Mr Hicks was a passenger, the court was told. \"There is no doubt that when the Audi smashed into the panel \n",
    "van he was on completely the wrong side of the road,\" Mr Gabb said. Mr Hicks, from Bath in Somerset, was asleep in the \n",
    "van being driven to a construction site in Salisbury by fellow DR Groundworks colleague, Patrick Gilleece. The jury was \n",
    "told the Maj Scott suffered \"substantial injuries\" and could not recall the crash, which happened shortly after 07:00 GMT \n",
    "on 6 October, 2014. He does not accept the charge and suggests it was in fact Mr Reed who had crossed the carriageway, \n",
    "causing the collision, Mr Gabb told the court. The trial continues.'''"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "55dacc34-cf5b-4937-bbe8-f1c09b7399c8",
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_base_model():\n",
    "    print('MOVING BASE MODEL TO GPU...', end='', flush=True)\n",
    "    start = time.time()\n",
    "    mask_model.cpu()\n",
    "    base_model.to(DEVICE)\n",
    "    print(f'DONE ({time.time() - start:.2f}s)')\n",
    "\n",
    "def load_mask_model():\n",
    "    print('MOVING MASK MODEL TO GPU...', end='', flush=True)\n",
    "    start = time.time()\n",
    "    mask_model.to(DEVICE)\n",
    "    base_model.cpu()\n",
    "    print(f'DONE ({time.time() - start:.2f}s)')\n",
    "\n",
    "def trim_to_shorter_length(texta, textb):\n",
    "    # truncate to shorter of o and s\n",
    "    shorter_length = min(len(texta.split(' ')), len(textb.split(' ')))\n",
    "    texta = ' '.join(texta.split(' ')[:shorter_length])\n",
    "    textb = ' '.join(textb.split(' ')[:shorter_length])\n",
    "    return texta, textb\n",
    "\n",
    "def get_ll(text):\n",
    "    with torch.no_grad():\n",
    "        tokenized = base_tokenizer(text, return_tensors=\"pt\").to(DEVICE)\n",
    "        labels = tokenized.input_ids\n",
    "        return -base_model(**tokenized, labels=labels).loss.item()\n",
    "\n",
    "\n",
    "def tokenize_and_mask(text, span_length, pct, ceil_pct=False):\n",
    "    tokens = text.split(' ')\n",
    "    mask_string = '<<<mask>>>'\n",
    "\n",
    "    n_spans = pct * len(tokens) / (span_length + 1 * 2)\n",
    "    if ceil_pct:\n",
    "        n_spans = np.ceil(n_spans)\n",
    "    n_spans = int(n_spans)\n",
    "\n",
    "    n_masks = 0\n",
    "    while n_masks < n_spans:\n",
    "        start = np.random.randint(0, len(tokens) - span_length)\n",
    "        end = start + span_length\n",
    "        search_start = max(0, start - 1)\n",
    "        search_end = min(len(tokens), end + 1)\n",
    "        if mask_string not in tokens[search_start:search_end]:\n",
    "            tokens[start:end] = [mask_string]\n",
    "            n_masks += 1\n",
    "    \n",
    "    # replace each occurrence of mask_string with <extra_id_NUM>, where NUM increments\n",
    "    num_filled = 0\n",
    "    for idx, token in enumerate(tokens):\n",
    "        if token == mask_string:\n",
    "            tokens[idx] = f'<extra_id_{num_filled}>'\n",
    "            num_filled += 1\n",
    "    assert num_filled == n_masks, f\"num_filled {num_filled} != n_masks {n_masks}\"\n",
    "    text = ' '.join(tokens)\n",
    "    return text\n",
    "\n",
    "def count_masks(texts):\n",
    "    return [len([x for x in text.split() if x.startswith(\"<extra_id_\")]) for text in texts]\n",
    "\n",
    "def replace_masks(texts):\n",
    "    n_expected = count_masks(texts)\n",
    "    stop_id = mask_tokenizer.encode(f\"<extra_id_{max(n_expected)}>\")[0]\n",
    "    tokens = mask_tokenizer(texts, return_tensors=\"pt\", padding=True).to(DEVICE)\n",
    "    outputs = mask_model.generate(**tokens, max_length=150, do_sample=True, top_p=1, num_return_sequences=1, eos_token_id=stop_id)\n",
    "    return mask_tokenizer.batch_decode(outputs, skip_special_tokens=False)\n",
    "\n",
    "def extract_fills(texts, pattern=re.compile(r\"<extra_id_\\d+>\")):\n",
    "    # remove <pad> from beginning of each text\n",
    "    texts = [x.replace(\"<pad>\", \"\").replace(\"</s>\", \"\").strip() for x in texts]\n",
    "\n",
    "    # return the text in between each matched mask token\n",
    "    extracted_fills = [pattern.split(x)[1:-1] for x in texts]\n",
    "\n",
    "    # remove whitespace around each fill\n",
    "    extracted_fills = [[y.strip() for y in x] for x in extracted_fills]\n",
    "\n",
    "    return extracted_fills\n",
    "\n",
    "def apply_extracted_fills(masked_texts, extracted_fills):\n",
    "    # split masked text into tokens, only splitting on spaces (not newlines)\n",
    "    tokens = [x.split(' ') for x in masked_texts]\n",
    "\n",
    "    n_expected = count_masks(masked_texts)\n",
    "\n",
    "    # replace each mask token with the corresponding fill\n",
    "    for idx, (text, fills, n) in enumerate(zip(tokens, extracted_fills, n_expected)):\n",
    "        if len(fills) < n:\n",
    "            tokens[idx] = []\n",
    "        else:\n",
    "            for fill_idx in range(n):\n",
    "                text[text.index(f\"<extra_id_{fill_idx}>\")] = fills[fill_idx]\n",
    "\n",
    "    # join tokens back into text\n",
    "    texts = [\" \".join(x) for x in tokens]\n",
    "    return texts\n",
    "\n",
    "def generate_perturbed_texts(text, n):\n",
    "    perturbed_texts = []\n",
    "    for _ in range(n):\n",
    "        masked_sample = tokenize_and_mask(text, span_length=2, pct=0.3)\n",
    "        raw_fills = replace_masks([masked_sample])\n",
    "        extracted_fills = extract_fills(raw_fills)\n",
    "        perturbed_text = apply_extracted_fills([masked_sample], extracted_fills)\n",
    "        perturbed_texts.append(perturbed_text)\n",
    "        pass\n",
    "    return perturbed_texts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "abc8c3f0-facd-4040-840a-95420c362450",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "('gpt2-medium', 't5-large')"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "BASE_MODEL_NAME, MASK_FILLING_MODEL_NAME"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "d41eb774-2e6d-4213-a00e-a27a666bdde6",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/dhavalpotdar/.pyenv/versions/3.10.0/envs/ids703-nlp-detectgpt/lib/python3.10/site-packages/transformers/models/t5/tokenization_t5_fast.py:160: FutureWarning: This tokenizer was incorrectly instantiated with a model max length of 512 which will be corrected in Transformers v5.\n",
      "For now, this behavior is kept to avoid breaking backwards compatibility when padding/encoding with `truncation is True`.\n",
      "- Be aware that you SHOULD NOT rely on t5-large automatically truncating your input to 512 when padding/encoding.\n",
      "- If you want to encode/pad to sequences longer than 512 you can either instantiate this tokenizer with `model_max_length` or pass `max_length` when encoding/padding.\n",
      "- To avoid this warning, please instantiate this tokenizer with `model_max_length` set to your preferred value.\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "base_model = transformers.AutoModelForCausalLM.from_pretrained(BASE_MODEL_NAME)\n",
    "base_tokenizer = transformers.AutoTokenizer.from_pretrained(BASE_MODEL_NAME)\n",
    "base_tokenizer.pad_token = base_tokenizer.eos_token # throws error if not set\n",
    "\n",
    "mask_model = transformers.AutoModelForSeq2SeqLM.from_pretrained(MASK_FILLING_MODEL_NAME)\n",
    "mask_tokenizer = transformers.AutoTokenizer.from_pretrained(MASK_FILLING_MODEL_NAME)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "8c518c4f-42cf-48ec-aac6-3709747ccbe9",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MOVING BASE MODEL TO GPU...DONE (0.32s)\n"
     ]
    }
   ],
   "source": [
    "load_base_model()\n",
    "\n",
    "# tokenize text and slice the first PROMPT_TOKENS\n",
    "all_encoded = base_tokenizer(real_text, return_tensors=\"pt\", padding=True).to(DEVICE)\n",
    "all_encoded = {key: value[:, :PROMPT_TOKENS] for key, value in all_encoded.items()}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "bd0d2113-0d51-4208-a39f-810d1b0b0f42",
   "metadata": {},
   "outputs": [],
   "source": [
    "sampling_kwargs = dict()\n",
    "sampling_kwargs['top_p'] = 0.96\n",
    "sampling_kwargs['top_k'] = 40\n",
    "min_length = 50\n",
    "max_length = 200\n",
    "\n",
    "outputs = base_model.generate(**all_encoded, \n",
    "                              min_length=min_length, \n",
    "                              max_length=max_length, \n",
    "                              do_sample=True, \n",
    "                              **sampling_kwargs, \n",
    "                              pad_token_id=base_tokenizer.eos_token_id, \n",
    "                              eos_token_id=base_tokenizer.eos_token_id)\n",
    "sampled_text = base_tokenizer.batch_decode(outputs, skip_special_tokens=True)\n",
    "sampled_text = sampled_text[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "1aa8c5e7-4de2-4457-bc12-09b91de1bdd1",
   "metadata": {},
   "outputs": [],
   "source": [
    "data = {}\n",
    "data['real'], data['sampled'] = trim_to_shorter_length(real_text, sampled_text)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "33dfc7a1-0f5a-4372-a00f-05ff1ceec373",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(-2.9955995082855225, -2.2111732959747314)"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# get likelihood of each text under base model\n",
    "ll_real = get_ll(data['real'])\n",
    "ll_sampled = get_ll(data['sampled'])\n",
    "\n",
    "(ll_real, ll_sampled)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "1f8c5d0e-be98-4335-8320-19e1229a418a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MOVING MASK MODEL TO GPU...DONE (0.69s)\n"
     ]
    }
   ],
   "source": [
    "load_mask_model()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "c7df63f9-d987-49e7-adb4-0e72c942d116",
   "metadata": {},
   "outputs": [],
   "source": [
    "perturbed_texts = generate_perturbed_texts(data['sampled'], 10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "9be78ffa-2704-4dec-9dd4-0f150bab24a5",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[['. Scott, 40, is accused of driving at speeds of up to 100mph in bad weather before the crash that killed eight people. He was arrested on charges of failing to stop and avoid an obstruction of traffic, speeding and dangerous driving and has been bailed until July 14.\\n\\nPolice said the driver of a vehicle hit by the motorway were unhurt but did not know the extent of his injuries and the condition of the driver of another vehicle suffered as the direct result of the collision.\\n\\nThe two vehicles collided at the A12 in the far north-west after a man drove an electric blue Volkswagen Touareg through a red light at an estimated speed of 90mph\\n\\nFour people, five cyclists and a pedestrian were killed and nearly 100 were injured when a Mercedes, a blue BMW and a white Renault hit a light rail line near North London on February 13, 2015. The cars were travelling at up to 130mph and police claimed the vehicles were using the wrong signals.\\n\\n'],\n",
       " [\"'Maj Richard Scott, 40, is accused of driving at speeds of up to 95mph (153km/h) in bad weather before the crash that killed seven.\\n\\nScott was arrested on charges of failing to stop and causing obstruction of traffic, speeding and disorderly driving and has been bailed until February 14.\\n\\nPolice said the driver of a second vehicle on the motorway had minor injuries and the condition of the road was unknown. A man was arrested on two further charges for damage to another vehicle suffered as a result of the collision.\\n\\nThe two vehicles collided at the same bridge in the far north-west after a man drove an electric blue Volkswagen Touareg through a red light at an estimated speed of 90mph\\n\\nFour pedestrians, two cyclists and one pedestrian were killed and nearly 100 were injured when a Mercedes, a blue BMW and a white Renault crashed through light -changes in central London on February 13, 2015. The vehicles were travelling at up to 175mph and police believe they were using the wrong signals.\\n\\n\"],\n",
       " [\"'Maj Richard Scott, 40, is accused of driving at speeds of up to 100mph (153km/h) in the hours before the crash , killing seven.\\n\\nScott is being held on charges of failing to stop and avoid an obstruction of traffic, speeding and careless driving and has been bailed until next Wednesday, May 14.\\n\\nPolice said the driver of a vehicle hit by the collision on the motorway had minor injuries and the condition of the driver of another vehicle suffered as a result of the collision.\\n\\nThe two cars collided into each other at the same bridge in the far north-west after a man drove an electric blue Volkswagen Touareg through a traffic light at an estimated speed of 90mph\\n\\nFour people, five cyclists and a pedestrian were killed and nearly 100 were injured when a Mercedes, a blue Audi and a white Renault hit a light rail line in central London on February 13, 2015. The cars were travelling at up to 120mph, and police believe they were using the wrong signals.\\n\\n\"],\n",
       " [\"'Maj Richard Scott, 40, is accused of driving at speeds of up to 95mph (153km/h) in bad weather before the crash that killed two people. He was arrested on suspicion of failing to stop and avoid an incident with traffic, speeding and careless driving and has been bailed until February 14.\\n\\nPolice said the driver of a Mercedes on the right had minor injuries and the condition of the driver of another vehicle suffered from a serious injury as a result of the collision.\\n\\nThe trucks also collided at the same bridge in the far north-west after a man drove an electric blue Volkswagen Touareg through a red light at an estimated speed of 90mph\\n\\nFour people, five cyclists and a pedestrian were killed and nearly 100 were injured when a Mercedes, a blue BMW and a white Renault hit a light rail line in east London on February 13, 2018. The cars were travelling at up to 65mph and police believe they were using the wrong signals.\\n\\n\"],\n",
       " [\"'Maj Richard Scott, 40, is charged with driving at speeds of up to 95mph (153km/h) in the moments before the crash that killed five people. He was arrested on charges of failure to stop , an obstruction of traffic, speeding and careless driving and has been bailed until February 14.\\n\\nPolice said the driver of a vehicle hit on the motorway had minor injuries and the condition of the driver of a vehicle involved in the wreck had been assessed, but that he had suffered as a result of the collision.\\n\\nThe two vehicles collided at the same bridge in the far north-west after a man drove an electric blue Volkswagen Touareg through a red light at an estimated speed of 90mph\\n\\nFour people, five cyclists and a pedestrian were killed and over 100 were injured before police arrived. A red Mercedes, red BMW and a white Renault hit a light rail line in central London on February 13, 2015. The cars were travelling up to 130mph and police believe they were using the wrong signals.\\n\\n\"],\n",
       " [\"'Maj Richard Scott, 40, is accused of driving at speeds of 75 to 95mph (153km/h) in bad weather before the crash that killed seven.\\n\\nScott was acquitted of charges of failing to stop and avoid obstruction of traffic, speeding and careless driving and has been released on bail until February 14.\\n\\nPolice said Tuesday the driver of a vehicle on the motorway had minor injuries and the condition of the driver of another vehicle suffered as a result of the collision.\\n\\nThe two vehicles collided at the same bridge in the north-west after a man drove an electric blue Volkswagen Touareg through a red light at an estimated speed of 90mph\\n\\nFour people, five cyclists and a pedestrian were killed and nearly 100 were injured when a Mercedes, a blue BMW and a white Renault collided while negotiating a red light along a motorway in central London on Saturday, March 13, 2015. The cars were travelling at speeds of up to 130mph and police believe they were using the wrong signals.\\n\\n\"],\n",
       " [\"'Maj Richard Scott, 31, was accused of driving at speeds of up to 95mph (153km/h) in bad weather while travelling east on the motorway in a crash that killed seven.\\n\\nScott was arrested on charges of failing to stop to avoid an obstruction of traffic, speeding and careless driving and has been in custody since February 14.\\n\\nPolice said the driver of a vehicle hit on the motorway had minor injuries and the condition of the driver of another vehicle suffered as a result of the collision.\\n\\nThe two vehicles collided at the same bridge in the north-west after a man drove an electric Volkswagen Touareg through a red light at an estimated speed of 90mph\\n\\nFour people, five cyclists and a pedestrian were killed and 35 others were injured when a Mercedes, a blue BMW and a white Audi crashed into each other on a light rail line in central London on February 13, 2015. The cars were travelling at up to 130mph when they collided. Police believe they were using the wrong signals.\\n\\n\"],\n",
       " [\"'Maj Richard Scott, 40, is accused of driving at speeds of up to 100mph (153km/h) in the hours before the crash that killed seven.\\n\\nScott was arrested on charges of failing to stop to avoid an obstruction of traffic, speeding and reckless driving and has been bailed until February 14.\\n\\nPolice said the driver of a vehicle hit along the north-east motorway had died and the condition of the driver of another vehicle suffered as a result of the collision.\\n\\nThe two vehicles collided at the same bridge in the far north-west after a man drove an electric blue Volkswagen Touareg through a red light at an estimated speed of 90mph\\n\\nFour people, five cyclists and a pedestrian were killed and at least 40 were injured when a Mercedes, a blue BMW and a white Renault hit a light at Wimbledon Green in central London on January 13, 2015. The accident started when a Mercedes was travelling at up to 130mph and police believe it was using the wrong signals.\\n\\n\"],\n",
       " [\"'Maj Richard ' Scott is accused of driving through traffic lights at speeds of up to 95mph in bad weather before the crash that killed seven.\\n\\nScott was arrested on charges of failing to stop and cross the centerline, obstruction of traffic, speeding and careless driving and has been bailed until February 14.\\n\\nPolice said the driver of a vehicle hit on the M6 motorway had minor injuries and the condition of the driver of another vehicle suffered as a result of the collision.\\n\\nThe two vehicles collided at the same bridge in the far north-west of the city after a man drove an electric blue Volkswagen Touareg through a red light at a speed of 90mph\\n\\nFour people, five cyclists and a pedestrian were killed and nearly 100 were injured after a Mercedes, a blue BMW and a white Renault hit a light rail line in central London on February 13, 2018. The cars were travelling at speeds of up to 130mph and police believe they were using wrong signals.\\n\\n\"],\n",
       " [\"'Maj Richard Scott, 43, from Manchester is accused of driving at speeds of up to 67mph (153km/h) in bad weather before the crash that killed seven.\\n\\nScott was arrested on suspicion of failing to stop and avoid an obstruction of traffic, speeding and careless driving and has been bailed until February 14.\\n\\nPolice said the driver of the vehicle hit on the motorway had minor injuries and the condition of the driver of another vehicle suffered as a result of the collision.\\n\\nThe two vehicles were at the same bridge in the far north-west of Manchester. A man drove an electric blue Volkswagen Touareg through a red light at an estimated speed of 90mph\\n\\nFour people, five cyclists and a pedestrian were killed and nearly 100 were injured when a Mercedes, a blue Renault and a green BT hit a light rail line in central London on February 13, 2015. The cars were travelling at up to 130mph and police said the cars were using the wrong signals.\\n\\n\"]]"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "perturbed_texts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "849de7a4-4930-4f74-8960-0896668b239f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MOVING BASE MODEL TO GPU...DONE (0.76s)\n"
     ]
    }
   ],
   "source": [
    "load_base_model()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "ed754b23-e79a-45f0-bf2b-0515e5aaa9ba",
   "metadata": {},
   "outputs": [],
   "source": [
    "lls_perturbed = torch.tensor([get_ll(perturbed_text) for perturbed_text in perturbed_texts])\n",
    "\n",
    "MU = (1/10) * torch.sum(lls_perturbed)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "1ae44cc0-2e3e-4743-bd32-0cd27f427a6a",
   "metadata": {},
   "outputs": [],
   "source": [
    "variance_norm = torch.sqrt((1/(10 - 1)) * torch.square(torch.sum(lls_perturbed - MU)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "6f7a4562-6c6b-461e-afdc-6bafb20f506b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor(-1837881.7500)"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(ll_real - MU)/variance_norm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a2aa2eed-d306-4487-a947-1c8e4cedb7bc",
   "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": "3.10.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}