{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "source": [
    "from datasets import load_dataset, Features\n",
    "from transformers import AutoTokenizer, AutoModelForSequenceClassification, Trainer, TrainingArguments\n",
    "import os\n",
    "from os.path import join\n",
    "import pandas as pd\n",
    "from datasets import ClassLabel, Value\n",
    "\n",
    "# https://huggingface.co/docs/datasets/loading_datasets.html\n",
    "\n",
    "DATASET_DIR = os.environ['DATASET_DIR']"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "source": [
    "dataset = load_dataset(\"holodata/sensai\", features=Features(\n",
    "                {\n",
    "                    \"body\": Value(\"string\"),\n",
    "                    \"toxic\": ClassLabel(num_classes=2, names=['0', '1'])\n",
    "                }\n",
    "            ))\n",
    "dataset = dataset['train']"
   ],
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "version_major": 2,
       "version_minor": 0,
       "model_id": "704789c10f1e44ddbb83262b8a826eec"
      },
      "text/plain": [
       "  0%|          | 0/1 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {}
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "Using custom data configuration sensai-4d9ed81389161083\n",
      "Reusing dataset parquet (/home/uetchy/.cache/huggingface/datasets/parquet/sensai-4d9ed81389161083/0.0.0/9296ce43568b20d72ff8ff8ecbc821a16b68e9b8b7058805ef11f06e035f911a)\n"
     ]
    },
    {
     "output_type": "display_data",
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "version_major": 2,
       "version_minor": 0,
       "model_id": "8acdde0c4caa4d4698f97ef29993195b"
      },
      "text/plain": [
       "  0%|          | 0/1 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {}
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "source": [
    "dataset.features"
   ],
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "{'body': Value(dtype='string', id=None),\n",
       " 'toxic': ClassLabel(num_classes=2, names=['0', '1'], names_file=None, id=None)}"
      ]
     },
     "metadata": {},
     "execution_count": 3
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "source": [
    "model = AutoModelForSequenceClassification.from_pretrained(\"bert-base-cased\", num_labels=2)\n",
    "tokenizer = AutoTokenizer.from_pretrained(\"bert-base-cased\")"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "Some weights of the model checkpoint at bert-base-cased were not used when initializing BertForSequenceClassification: ['cls.predictions.transform.LayerNorm.bias', 'cls.predictions.transform.dense.weight', 'cls.predictions.transform.LayerNorm.weight', 'cls.seq_relationship.weight', 'cls.seq_relationship.bias', 'cls.predictions.transform.dense.bias', 'cls.predictions.decoder.weight', 'cls.predictions.bias']\n",
      "- This IS expected if you are initializing BertForSequenceClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
      "- This IS NOT expected if you are initializing BertForSequenceClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-cased and are newly initialized: ['classifier.weight', 'classifier.bias']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "source": [
    "samples = dataset.shuffle().select(range(50000))\n",
    "\n",
    "def tokenize_function(examples):\n",
    "    return tokenizer(examples[\"body\"], padding=\"max_length\", truncation=True)\n",
    "\n",
    "tokenized_datasets = samples.map(tokenize_function, batched=True)\n",
    "tokenized_datasets.rename_column_(\"toxic\", \"label\")"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "Loading cached shuffled indices for dataset at /home/uetchy/.cache/huggingface/datasets/parquet/sensai-4d9ed81389161083/0.0.0/9296ce43568b20d72ff8ff8ecbc821a16b68e9b8b7058805ef11f06e035f911a/cache-24e3dd769ef2f1b7.arrow\n",
      "Loading cached processed dataset at /home/uetchy/.cache/huggingface/datasets/parquet/sensai-4d9ed81389161083/0.0.0/9296ce43568b20d72ff8ff8ecbc821a16b68e9b8b7058805ef11f06e035f911a/cache-8395f066c72e57d7.arrow\n",
      "/tmp/ipykernel_4082765/2982913603.py:7: FutureWarning: rename_column_ is deprecated and will be removed in the next major version of datasets. Use Dataset.rename_column instead.\n",
      "  tokenized_datasets.rename_column_(\"toxic\", \"label\")\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "source": [
    "splitset = tokenized_datasets.train_test_split(0.2)\n",
    "splitset"
   ],
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "DatasetDict({\n",
       "    train: Dataset({\n",
       "        features: ['attention_mask', 'body', 'input_ids', 'token_type_ids', 'label'],\n",
       "        num_rows: 40000\n",
       "    })\n",
       "    test: Dataset({\n",
       "        features: ['attention_mask', 'body', 'input_ids', 'token_type_ids', 'label'],\n",
       "        num_rows: 10000\n",
       "    })\n",
       "})"
      ]
     },
     "metadata": {},
     "execution_count": 6
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "source": [
    "training_args = TrainingArguments(\"test_trainer\")\n",
    "trainer = Trainer(\n",
    "    model=model, args=training_args, train_dataset=splitset['train'], eval_dataset=splitset['test']\n",
    ")"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "source": [
    "trainer.train(resume_from_checkpoint=True)"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "Loading model from test_trainer/checkpoint-12500).\n",
      "The following columns in the training set  don't have a corresponding argument in `BertForSequenceClassification.forward` and have been ignored: body.\n",
      "***** Running training *****\n",
      "  Num examples = 40000\n",
      "  Num Epochs = 3\n",
      "  Instantaneous batch size per device = 8\n",
      "  Total train batch size (w. parallel, distributed & accumulation) = 8\n",
      "  Gradient Accumulation steps = 1\n",
      "  Total optimization steps = 15000\n",
      "  Continuing training from checkpoint, will skip to saved global_step\n",
      "  Continuing training from epoch 2\n",
      "  Continuing training from global step 12500\n",
      "  Will skip the first 2 epochs then the first 2500 batches in the first epoch. If this takes a lot of time, you can add the `--ignore_data_skip` flag to your launch command, but you will resume the training on data already seen by your model.\n"
     ]
    },
    {
     "output_type": "display_data",
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "version_major": 2,
       "version_minor": 0,
       "model_id": "d749dea90f914485916ce82c65e73fe6"
      },
      "text/plain": [
       "  0%|          | 0/2500 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {}
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "Didn't find an RNG file, if you are resuming a training that was launched in a distributed fashion, reproducibility is not guaranteed.\n"
     ]
    },
    {
     "output_type": "display_data",
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "      \n",
       "      <progress value='15000' max='15000' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      [15000/15000 31:45, Epoch 3/3]\n",
       "    </div>\n",
       "    <table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>Step</th>\n",
       "      <th>Training Loss</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>13000</td>\n",
       "      <td>0.687500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13500</td>\n",
       "      <td>0.686300</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14000</td>\n",
       "      <td>0.637900</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14500</td>\n",
       "      <td>0.643200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15000</td>\n",
       "      <td>0.627700</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table><p>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {}
    },
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "Saving model checkpoint to test_trainer/checkpoint-13000\n",
      "Configuration saved in test_trainer/checkpoint-13000/config.json\n",
      "Model weights saved in test_trainer/checkpoint-13000/pytorch_model.bin\n",
      "Saving model checkpoint to test_trainer/checkpoint-13500\n",
      "Configuration saved in test_trainer/checkpoint-13500/config.json\n",
      "Model weights saved in test_trainer/checkpoint-13500/pytorch_model.bin\n",
      "Saving model checkpoint to test_trainer/checkpoint-14000\n",
      "Configuration saved in test_trainer/checkpoint-14000/config.json\n",
      "Model weights saved in test_trainer/checkpoint-14000/pytorch_model.bin\n",
      "Saving model checkpoint to test_trainer/checkpoint-14500\n",
      "Configuration saved in test_trainer/checkpoint-14500/config.json\n",
      "Model weights saved in test_trainer/checkpoint-14500/pytorch_model.bin\n",
      "Saving model checkpoint to test_trainer/checkpoint-15000\n",
      "Configuration saved in test_trainer/checkpoint-15000/config.json\n",
      "Model weights saved in test_trainer/checkpoint-15000/pytorch_model.bin\n",
      "\n",
      "\n",
      "Training completed. Do not forget to share your model on huggingface.co/models =)\n",
      "\n",
      "\n"
     ]
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "TrainOutput(global_step=15000, training_loss=0.10941998901367188, metrics={'train_runtime': 1918.0916, 'train_samples_per_second': 62.562, 'train_steps_per_second': 7.82, 'total_flos': 3.24994775580672e+16, 'train_loss': 0.10941998901367188, 'epoch': 3.0})"
      ]
     },
     "metadata": {},
     "execution_count": 9
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "source": [],
   "outputs": [],
   "metadata": {}
  }
 ],
 "metadata": {
  "orig_nbformat": 4,
  "language_info": {
   "name": "python",
   "version": "3.8.6",
   "mimetype": "text/x-python",
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "pygments_lexer": "ipython3",
   "nbconvert_exporter": "python",
   "file_extension": ".py"
  },
  "kernelspec": {
   "name": "python3",
   "display_name": "Python 3.8.6 64-bit"
  },
  "interpreter": {
   "hash": "c8daecebaf2d81430b8373e4b4af380b12df116248cd1bbadd3fc947f45a1f88"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}