{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "xFqoZo2jgBuP"
   },
   "source": [
    "# Finetuner un modèle avec l'API Trainer"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "odo72vovgBuR"
   },
   "source": [
    "Installez les bibliothèques 🤗 Transformers et 🤗 Datasets pour exécuter ce notebook."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "KL0srL9lgBuS"
   },
   "outputs": [],
   "source": [
    "!pip install datasets transformers[sentencepiece]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "lxEQZhrFgBuU"
   },
   "outputs": [],
   "source": [
    "from datasets import load_dataset\n",
    "from transformers import AutoTokenizer, DataCollatorWithPadding\n",
    "\n",
    "raw_datasets = load_dataset(\"paws-x\", \"fr\")\n",
    "checkpoint = \"camembert-base\"\n",
    "tokenizer = AutoTokenizer.from_pretrained(checkpoint)\n",
    "\n",
    "\n",
    "def tokenize_function(example):\n",
    "    return tokenizer(example[\"sentence1\"], example[\"sentence2\"], truncation=True)\n",
    "\n",
    "\n",
    "tokenized_datasets = raw_datasets.map(tokenize_function, batched=True)\n",
    "data_collator = DataCollatorWithPadding(tokenizer=tokenizer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "Dp-s6a1rgBuV"
   },
   "outputs": [],
   "source": [
    "from transformers import TrainingArguments\n",
    "\n",
    "training_args = TrainingArguments(\"test-trainer\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "1r1XVVN9gBuV"
   },
   "outputs": [],
   "source": [
    "from transformers import AutoModelForSequenceClassification\n",
    "\n",
    "model = AutoModelForSequenceClassification.from_pretrained(checkpoint, num_labels=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "d9eakOZOgBuW"
   },
   "outputs": [],
   "source": [
    "from transformers import Trainer\n",
    "\n",
    "trainer = Trainer(\n",
    "    model,\n",
    "    training_args,\n",
    "    train_dataset=tokenized_datasets[\"train\"],\n",
    "    eval_dataset=tokenized_datasets[\"validation\"],\n",
    "    data_collator=data_collator,\n",
    "    tokenizer=tokenizer,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "2hF9UHdXgBuX"
   },
   "outputs": [],
   "source": [
    "trainer.train() # Attention, une epoch prend 12h !"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "eWao7LvzgBuY"
   },
   "outputs": [],
   "source": [
    "predictions = trainer.predict(tokenized_datasets[\"validation\"])\n",
    "print(predictions.predictions.shape, predictions.label_ids.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "2hen_ecUgBuZ"
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "\n",
    "preds = np.argmax(predictions.predictions, axis=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "JS6HytHngBua"
   },
   "outputs": [],
   "source": [
    "from datasets import load_metric\n",
    "\n",
    "metric = load_metric(\"glue\", \"mrpc\")\n",
    "metric.compute(predictions=preds, references=predictions.label_ids)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "POWKrVmkgBub"
   },
   "outputs": [],
   "source": [
    "def compute_metrics(eval_preds):\n",
    "    metric = load_metric(\"glue\", \"mrpc\")\n",
    "    logits, labels = eval_preds\n",
    "    predictions = np.argmax(logits, axis=-1)\n",
    "    return metric.compute(predictions=predictions, references=labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "T03glg22gBuc"
   },
   "outputs": [],
   "source": [
    "training_args = TrainingArguments(\"test-trainer\", evaluation_strategy=\"epoch\")\n",
    "model = AutoModelForSequenceClassification.from_pretrained(checkpoint, num_labels=2)\n",
    "\n",
    "trainer = Trainer(\n",
    "    model,\n",
    "    training_args,\n",
    "    train_dataset=tokenized_datasets[\"train\"],\n",
    "    eval_dataset=tokenized_datasets[\"validation\"],\n",
    "    data_collator=data_collator,\n",
    "    tokenizer=tokenizer,\n",
    "    compute_metrics=compute_metrics,\n",
    ")"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "provenance": []
  },
  "gpuClass": "standard",
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}