feat(notebook): ajout du notebook d'entraînement CamemBERT (fine-tuning)

notebooks/02_train_camenbert.ipynb

@@ -0,0 +1,586 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "id": "e63f42f0-e971-4017-8550-21fdcfc2de11",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Requirement already satisfied: scikit-learn in /usr/local/lib/python3.10/dist-packages (1.7.0)\n",
+      "Requirement already satisfied: numpy>=1.22.0 in /usr/local/lib/python3.10/dist-packages (from scikit-learn) (1.24.1)\n",
+      "Requirement already satisfied: scipy>=1.8.0 in /usr/local/lib/python3.10/dist-packages (from scikit-learn) (1.15.3)\n",
+      "Requirement already satisfied: joblib>=1.2.0 in /usr/local/lib/python3.10/dist-packages (from scikit-learn) (1.5.1)\n",
+      "Requirement already satisfied: threadpoolctl>=3.1.0 in /usr/local/lib/python3.10/dist-packages (from scikit-learn) (3.6.0)\n",
+      "\u001b[33mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv\u001b[0m\u001b[33m\n",
+      "\u001b[0m\n",
+      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m A new release of pip is available: \u001b[0m\u001b[31;49m23.3.1\u001b[0m\u001b[39;49m -> \u001b[0m\u001b[32;49m25.1.1\u001b[0m\n",
+      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m To update, run: \u001b[0m\u001b[32;49mpython -m pip install --upgrade pip\u001b[0m\n",
+      "Requirement already satisfied: pandas in /usr/local/lib/python3.10/dist-packages (2.3.0)\n",
+      "Requirement already satisfied: numpy>=1.22.4 in /usr/local/lib/python3.10/dist-packages (from pandas) (1.24.1)\n",
+      "Requirement already satisfied: python-dateutil>=2.8.2 in /usr/local/lib/python3.10/dist-packages (from pandas) (2.8.2)\n",
+      "Requirement already satisfied: pytz>=2020.1 in /usr/local/lib/python3.10/dist-packages (from pandas) (2025.2)\n",
+      "Requirement already satisfied: tzdata>=2022.7 in /usr/local/lib/python3.10/dist-packages (from pandas) (2025.2)\n",
+      "Requirement already satisfied: six>=1.5 in /usr/lib/python3/dist-packages (from python-dateutil>=2.8.2->pandas) (1.16.0)\n",
+      "\u001b[33mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv\u001b[0m\u001b[33m\n",
+      "\u001b[0m\n",
+      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m A new release of pip is available: \u001b[0m\u001b[31;49m23.3.1\u001b[0m\u001b[39;49m -> \u001b[0m\u001b[32;49m25.1.1\u001b[0m\n",
+      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m To update, run: \u001b[0m\u001b[32;49mpython -m pip install --upgrade pip\u001b[0m\n",
+      "Requirement already satisfied: tqdm in /usr/local/lib/python3.10/dist-packages (4.67.1)\n",
+      "\u001b[33mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv\u001b[0m\u001b[33m\n",
+      "\u001b[0m\n",
+      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m A new release of pip is available: \u001b[0m\u001b[31;49m23.3.1\u001b[0m\u001b[39;49m -> \u001b[0m\u001b[32;49m25.1.1\u001b[0m\n",
+      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m To update, run: \u001b[0m\u001b[32;49mpython -m pip install --upgrade pip\u001b[0m\n",
+      "Requirement already satisfied: sentencepiece in /usr/local/lib/python3.10/dist-packages (0.2.0)\n",
+      "\u001b[33mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv\u001b[0m\u001b[33m\n",
+      "\u001b[0m\n",
+      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m A new release of pip is available: \u001b[0m\u001b[31;49m23.3.1\u001b[0m\u001b[39;49m -> \u001b[0m\u001b[32;49m25.1.1\u001b[0m\n",
+      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m To update, run: \u001b[0m\u001b[32;49mpython -m pip install --upgrade pip\u001b[0m\n",
+      "Looking in indexes: https://download.pytorch.org/whl/cu118\n",
+      "Collecting torch==2.0.1\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torch-2.0.1%2Bcu118-cp310-cp310-linux_x86_64.whl (2267.3 MB)\n",
+      "Requirement already satisfied: torchvision in /usr/local/lib/python3.10/dist-packages (0.16.0+cu118)\n",
+      "Requirement already satisfied: torchaudio in /usr/local/lib/python3.10/dist-packages (2.1.0+cu118)\n",
+      "Requirement already satisfied: filelock in /usr/local/lib/python3.10/dist-packages (from torch==2.0.1) (3.9.0)\n",
+      "Requirement already satisfied: typing-extensions in /usr/local/lib/python3.10/dist-packages (from torch==2.0.1) (4.4.0)\n",
+      "Requirement already satisfied: sympy in /usr/local/lib/python3.10/dist-packages (from torch==2.0.1) (1.12)\n",
+      "Requirement already satisfied: networkx in /usr/local/lib/python3.10/dist-packages (from torch==2.0.1) (3.0)\n",
+      "Requirement already satisfied: jinja2 in /usr/local/lib/python3.10/dist-packages (from torch==2.0.1) (3.1.2)\n",
+      "Requirement already satisfied: triton==2.0.0 in /usr/local/lib/python3.10/dist-packages (from torch==2.0.1) (2.0.0)\n",
+      "Requirement already satisfied: cmake in /usr/local/lib/python3.10/dist-packages (from triton==2.0.0->torch==2.0.1) (4.0.3)\n",
+      "Requirement already satisfied: lit in /usr/local/lib/python3.10/dist-packages (from triton==2.0.0->torch==2.0.1) (18.1.8)\n",
+      "Requirement already satisfied: numpy in /usr/local/lib/python3.10/dist-packages (from torchvision) (1.24.1)\n",
+      "Requirement already satisfied: requests in /usr/local/lib/python3.10/dist-packages (from torchvision) (2.31.0)\n",
+      "INFO: pip is looking at multiple versions of torchvision to determine which version is compatible with other requirements. This could take a while.\n",
+      "Collecting torchvision\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchvision-0.22.1%2Bcu118-cp310-cp310-manylinux_2_28_x86_64.whl.metadata (6.1 kB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchvision-0.22.0%2Bcu118-cp310-cp310-manylinux_2_28_x86_64.whl.metadata (6.1 kB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchvision-0.21.0%2Bcu118-cp310-cp310-linux_x86_64.whl.metadata (6.1 kB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchvision-0.20.1%2Bcu118-cp310-cp310-linux_x86_64.whl (6.5 MB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchvision-0.20.0%2Bcu118-cp310-cp310-linux_x86_64.whl (6.5 MB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchvision-0.19.1%2Bcu118-cp310-cp310-linux_x86_64.whl (6.3 MB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchvision-0.19.0%2Bcu118-cp310-cp310-linux_x86_64.whl (6.3 MB)\n",
+      "INFO: pip is still looking at multiple versions of torchvision to determine which version is compatible with other requirements. This could take a while.\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchvision-0.18.1%2Bcu118-cp310-cp310-linux_x86_64.whl (6.3 MB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchvision-0.18.0%2Bcu118-cp310-cp310-linux_x86_64.whl (6.3 MB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchvision-0.17.2%2Bcu118-cp310-cp310-linux_x86_64.whl (6.2 MB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchvision-0.17.1%2Bcu118-cp310-cp310-linux_x86_64.whl (6.2 MB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchvision-0.17.0%2Bcu118-cp310-cp310-linux_x86_64.whl (6.2 MB)\n",
+      "INFO: This is taking longer than usual. You might need to provide the dependency resolver with stricter constraints to reduce runtime. See https://pip.pypa.io/warnings/backtracking for guidance. If you want to abort this run, press Ctrl + C.\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchvision-0.16.2%2Bcu118-cp310-cp310-linux_x86_64.whl (6.1 MB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchvision-0.16.1%2Bcu118-cp310-cp310-linux_x86_64.whl (6.1 MB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchvision-0.15.2%2Bcu118-cp310-cp310-linux_x86_64.whl (6.1 MB)\n",
+      "Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in /usr/local/lib/python3.10/dist-packages (from torchvision) (9.3.0)\n",
+      "INFO: pip is looking at multiple versions of torchaudio to determine which version is compatible with other requirements. This could take a while.\n",
+      "Collecting torchaudio\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchaudio-2.7.1%2Bcu118-cp310-cp310-manylinux_2_28_x86_64.whl.metadata (6.6 kB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchaudio-2.7.0%2Bcu118-cp310-cp310-manylinux_2_28_x86_64.whl.metadata (6.6 kB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchaudio-2.6.0%2Bcu118-cp310-cp310-linux_x86_64.whl.metadata (6.6 kB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchaudio-2.5.1%2Bcu118-cp310-cp310-linux_x86_64.whl (3.3 MB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchaudio-2.5.0%2Bcu118-cp310-cp310-linux_x86_64.whl (3.3 MB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchaudio-2.4.1%2Bcu118-cp310-cp310-linux_x86_64.whl (3.3 MB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchaudio-2.4.0%2Bcu118-cp310-cp310-linux_x86_64.whl (3.3 MB)\n",
+      "INFO: pip is still looking at multiple versions of torchaudio to determine which version is compatible with other requirements. This could take a while.\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchaudio-2.3.1%2Bcu118-cp310-cp310-linux_x86_64.whl (3.3 MB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchaudio-2.3.0%2Bcu118-cp310-cp310-linux_x86_64.whl (3.3 MB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchaudio-2.2.2%2Bcu118-cp310-cp310-linux_x86_64.whl (3.3 MB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchaudio-2.2.1%2Bcu118-cp310-cp310-linux_x86_64.whl (3.3 MB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchaudio-2.2.0%2Bcu118-cp310-cp310-linux_x86_64.whl (3.3 MB)\n",
+      "INFO: This is taking longer than usual. You might need to provide the dependency resolver with stricter constraints to reduce runtime. See https://pip.pypa.io/warnings/backtracking for guidance. If you want to abort this run, press Ctrl + C.\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchaudio-2.1.2%2Bcu118-cp310-cp310-linux_x86_64.whl (3.2 MB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchaudio-2.1.1%2Bcu118-cp310-cp310-linux_x86_64.whl (3.2 MB)\n",
+      "  Using cached https://download.pytorch.org/whl/cu118/torchaudio-2.0.2%2Bcu118-cp310-cp310-linux_x86_64.whl (4.4 MB)\n",
+      "Requirement already satisfied: MarkupSafe>=2.0 in /usr/local/lib/python3.10/dist-packages (from jinja2->torch==2.0.1) (2.1.2)\n",
+      "Requirement already satisfied: charset-normalizer<4,>=2 in /usr/local/lib/python3.10/dist-packages (from requests->torchvision) (2.1.1)\n",
+      "Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.10/dist-packages (from requests->torchvision) (3.4)\n",
+      "Requirement already satisfied: urllib3<3,>=1.21.1 in /usr/local/lib/python3.10/dist-packages (from requests->torchvision) (1.26.13)\n",
+      "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.10/dist-packages (from requests->torchvision) (2022.12.7)\n",
+      "Requirement already satisfied: mpmath>=0.19 in /usr/local/lib/python3.10/dist-packages (from sympy->torch==2.0.1) (1.3.0)\n",
+      "Installing collected packages: torch, torchvision, torchaudio\n",
+      "  Attempting uninstall: torch\n",
+      "    Found existing installation: torch 2.1.0+cu118\n",
+      "    Uninstalling torch-2.1.0+cu118:\n",
+      "      Successfully uninstalled torch-2.1.0+cu118\n",
+      "  Rolling back uninstall of torch\n",
+      "  Moving to /usr/local/bin/convert-caffe2-to-onnx\n",
+      "   from /tmp/pip-uninstall-ior9qvf0/convert-caffe2-to-onnx\n",
+      "  Moving to /usr/local/bin/convert-onnx-to-caffe2\n",
+      "   from /tmp/pip-uninstall-ior9qvf0/convert-onnx-to-caffe2\n",
+      "  Moving to /usr/local/bin/torchrun\n",
+      "   from /tmp/pip-uninstall-ior9qvf0/torchrun\n",
+      "  Moving to /usr/local/lib/python3.10/dist-packages/functorch/\n",
+      "   from /usr/local/lib/python3.10/dist-packages/~unctorch\n",
+      "  Moving to /usr/local/lib/python3.10/dist-packages/nvfuser/\n",
+      "   from /usr/local/lib/python3.10/dist-packages/~vfuser\n",
+      "  Moving to /usr/local/lib/python3.10/dist-packages/torch-2.1.0+cu118.dist-info/\n",
+      "   from /usr/local/lib/python3.10/dist-packages/~orch-2.1.0+cu118.dist-info\n",
+      "  Moving to /usr/local/lib/python3.10/dist-packages/torch/\n",
+      "   from /usr/local/lib/python3.10/dist-packages/~orch\n",
+      "  Moving to /usr/local/lib/python3.10/dist-packages/torchgen/\n",
+      "   from /usr/local/lib/python3.10/dist-packages/~orchgen\n",
+      "\u001b[31mERROR: Could not install packages due to an OSError: [Errno 28] No space left on device\n",
+      "\u001b[0m\u001b[31m\n",
+      "\u001b[0m\n",
+      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m A new release of pip is available: \u001b[0m\u001b[31;49m23.3.1\u001b[0m\u001b[39;49m -> \u001b[0m\u001b[32;49m25.1.1\u001b[0m\n",
+      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m To update, run: \u001b[0m\u001b[32;49mpython -m pip install --upgrade pip\u001b[0m\n",
+      "Requirement already satisfied: transformers in /usr/local/lib/python3.10/dist-packages (4.52.4)\n",
+      "Requirement already satisfied: filelock in /usr/local/lib/python3.10/dist-packages (from transformers) (3.9.0)\n",
+      "Requirement already satisfied: huggingface-hub<1.0,>=0.30.0 in /usr/local/lib/python3.10/dist-packages (from transformers) (0.33.0)\n",
+      "Requirement already satisfied: numpy>=1.17 in /usr/local/lib/python3.10/dist-packages (from transformers) (1.24.1)\n",
+      "Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.10/dist-packages (from transformers) (23.2)\n",
+      "Requirement already satisfied: pyyaml>=5.1 in /usr/local/lib/python3.10/dist-packages (from transformers) (6.0.1)\n",
+      "Requirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.10/dist-packages (from transformers) (2024.11.6)\n",
+      "Requirement already satisfied: requests in /usr/local/lib/python3.10/dist-packages (from transformers) (2.31.0)\n",
+      "Requirement already satisfied: tokenizers<0.22,>=0.21 in /usr/local/lib/python3.10/dist-packages (from transformers) (0.21.1)\n",
+      "Requirement already satisfied: safetensors>=0.4.3 in /usr/local/lib/python3.10/dist-packages (from transformers) (0.5.3)\n",
+      "Requirement already satisfied: tqdm>=4.27 in /usr/local/lib/python3.10/dist-packages (from transformers) (4.67.1)\n",
+      "Requirement already satisfied: fsspec>=2023.5.0 in /usr/local/lib/python3.10/dist-packages (from huggingface-hub<1.0,>=0.30.0->transformers) (2025.5.1)\n",
+      "Requirement already satisfied: typing-extensions>=3.7.4.3 in /usr/local/lib/python3.10/dist-packages (from huggingface-hub<1.0,>=0.30.0->transformers) (4.4.0)\n",
+      "Requirement already satisfied: hf-xet<2.0.0,>=1.1.2 in /usr/local/lib/python3.10/dist-packages (from huggingface-hub<1.0,>=0.30.0->transformers) (1.1.3)\n",
+      "Requirement already satisfied: charset-normalizer<4,>=2 in /usr/local/lib/python3.10/dist-packages (from requests->transformers) (2.1.1)\n",
+      "Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.10/dist-packages (from requests->transformers) (3.4)\n",
+      "Requirement already satisfied: urllib3<3,>=1.21.1 in /usr/local/lib/python3.10/dist-packages (from requests->transformers) (1.26.13)\n",
+      "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.10/dist-packages (from requests->transformers) (2022.12.7)\n",
+      "\u001b[33mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv\u001b[0m\u001b[33m\n",
+      "\u001b[0m\n",
+      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m A new release of pip is available: \u001b[0m\u001b[31;49m23.3.1\u001b[0m\u001b[39;49m -> \u001b[0m\u001b[32;49m25.1.1\u001b[0m\n",
+      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m To update, run: \u001b[0m\u001b[32;49mpython -m pip install --upgrade pip\u001b[0m\n"
+     ]
+    }
+   ],
+   "source": [
+    "!pip install scikit-learn\n",
+    "!pip install pandas\n",
+    "!pip install tqdm\n",
+    "!pip install sentencepiece\n",
+    "!pip install torch==2.0.1 torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118\n",
+    "!pip install --upgrade transformers"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "id": "7db96f7b-0cd3-4710-93d8-391622e60c25",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "import json\n",
+    "import pandas as pd\n",
+    "import torch\n",
+    "from torch.optim import AdamW\n",
+    "from torch.utils.data import Dataset, DataLoader\n",
+    "from torch.nn import CrossEntropyLoss\n",
+    "from transformers import CamembertTokenizer, CamembertForSequenceClassification, get_scheduler\n",
+    "from sklearn.model_selection import train_test_split\n",
+    "from sklearn.metrics import classification_report\n",
+    "from tqdm import tqdm"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "id": "338ab5df-bc6d-4f2c-8a5e-bc9bb7b658f1",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# ─────────────────────────────────────────────\n",
+    "# ⚙️ Config\n",
+    "# ─────────────────────────────────────────────\n",
+    "DEBUG = False\n",
+    "BATCH_SIZE = 64\n",
+    "EPOCHS = 3 if not DEBUG else 1\n",
+    "MAX_LEN = 128\n",
+    "LR = 2e-5\n",
+    "PATIENCE = 2  # pour l'early stopping"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "id": "3ccb7df2-77fc-461b-ac1e-05f1d8be7ed0",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Classes : df_labels\n",
+      "0    189412\n",
+      "1     33982\n",
+      "Name: count, dtype: int64\n"
+     ]
+    }
+   ],
+   "source": [
+    "# ─────────────────────────────────────────────\n",
+    "# 📁 Chargement du dataset\n",
+    "# ─────────────────────────────────────────────\n",
+    "df = pd.read_csv(\"jigsaw-toxic-comment-train-google-fr-cleaned.csv\")\n",
+    "df['comment_text'] = df['comment_text'].astype(str)\n",
+    "df.rename(columns={'comment_text': 'texts'}, inplace=True)\n",
+    "\n",
+    "label_cols = ['toxic', 'severe_toxic', 'obscene', 'threat', 'insult', 'identity_hate']\n",
+    "other_cols_to_drop = ['Unnamed: 0.1', 'Unnamed: 0', 'id']\n",
+    "cols_to_drop = label_cols + other_cols_to_drop\n",
+    "\n",
+    "df['df_labels'] = df[label_cols].max(axis=1)\n",
+    "df = df.drop(columns=cols_to_drop)\n",
+    "\n",
+    "# Debug : sous-échantillonnage équilibré\n",
+    "if DEBUG:\n",
+    "    df_0 = df[df[\"df_labels\"] == 0].sample(500, random_state=42)\n",
+    "    df_1 = df[df[\"df_labels\"] == 1].sample(500, random_state=42)\n",
+    "    df = pd.concat([df_0, df_1]).sample(frac=1, random_state=42)\n",
+    "\n",
+    "print(\"Classes :\", df['df_labels'].value_counts())"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "id": "83c4cf79-57d9-4d54-8d8f-0e4249b8a930",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# ─────────────────────────────────────────────\n",
+    "# 🔢 Dataset\n",
+    "# ─────────────────────────────────────────────\n",
+    "tokenizer = CamembertTokenizer.from_pretrained(\"camembert-base\")\n",
+    "\n",
+    "class CommentDataset(Dataset):\n",
+    "    def __init__(self, texts, labels, tokenizer, max_len):\n",
+    "        self.texts = texts\n",
+    "        self.labels = labels\n",
+    "        self.tokenizer = tokenizer\n",
+    "        self.max_len = max_len\n",
+    "\n",
+    "    def __len__(self):\n",
+    "        return len(self.texts)\n",
+    "\n",
+    "    def __getitem__(self, idx):\n",
+    "        encoding = self.tokenizer(\n",
+    "            self.texts[idx],\n",
+    "            padding=\"max_length\",\n",
+    "            truncation=True,\n",
+    "            max_length=self.max_len,\n",
+    "            return_tensors=\"pt\"\n",
+    "        )\n",
+    "        item = {key: val.squeeze() for key, val in encoding.items()}\n",
+    "        item['labels'] = torch.tensor(self.labels[idx], dtype=torch.long)\n",
+    "        return item\n",
+    "\n",
+    "# Split\n",
+    "X_train, X_val, y_train, y_val = train_test_split(df[\"texts\"].tolist(), df[\"df_labels\"].tolist(), test_size=0.2, random_state=42)\n",
+    "\n",
+    "train_dataset = CommentDataset(X_train, y_train, tokenizer, MAX_LEN)\n",
+    "val_dataset = CommentDataset(X_val, y_val, tokenizer, MAX_LEN)\n",
+    "\n",
+    "train_loader = DataLoader(train_dataset, batch_size=BATCH_SIZE, shuffle=True)\n",
+    "val_loader = DataLoader(val_dataset, batch_size=BATCH_SIZE)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "id": "05e8419e-dbeb-42ba-b9ed-ea099e96244a",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Some weights of CamembertForSequenceClassification were not initialized from the model checkpoint at camembert-base and are newly initialized: ['classifier.dense.bias', 'classifier.dense.weight', 'classifier.out_proj.bias', 'classifier.out_proj.weight']\n",
+      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Poids pour la loss : tensor([1.0000, 5.5739])\n"
+     ]
+    }
+   ],
+   "source": [
+    "# ─────────────────────────────────────────────\n",
+    "# 🧠 Modèle + loss pondérée\n",
+    "# ─────────────────────────────────────────────\n",
+    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
+    "model = CamembertForSequenceClassification.from_pretrained(\"camembert-base\", num_labels=2).to(device)\n",
+    "\n",
+    "# pondération dynamique\n",
+    "if DEBUG:\n",
+    "    class_weights = torch.tensor([1.0, 1.0], dtype=torch.float)\n",
+    "else:\n",
+    "    count_0 = df[df[\"df_labels\"] == 0].shape[0]\n",
+    "    count_1 = df[df[\"df_labels\"] == 1].shape[0]\n",
+    "    class_weights = torch.tensor([1.0, count_0 / count_1], dtype=torch.float)\n",
+    "\n",
+    "print(f\"Poids pour la loss : {class_weights}\")\n",
+    "loss_fn = CrossEntropyLoss(weight=class_weights.to(device))\n",
+    "\n",
+    "# Optimiseur et scheduler\n",
+    "optimizer = AdamW(model.parameters(), lr=LR)\n",
+    "scheduler = get_scheduler(\"linear\", optimizer=optimizer, num_warmup_steps=0, num_training_steps=len(train_loader) * EPOCHS)\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "id": "0b181b9f-64a6-479f-b585-221a598cded6",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "\n",
+      "🌟 Epoch 1/3\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Entraînement: 100%|██████████| 2793/2793 [18:23<00:00,  2.53it/s]\n"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "📉 Loss moyenne : 0.5043\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Évaluation: 100%|██████████| 699/699 [01:50<00:00,  6.32it/s]\n"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "🎯 F1-score (weighted) : 0.8826\n",
+      "✅ Nouveau meilleur modèle — sauvegarde manuelle...\n",
+      "\n",
+      "🌟 Epoch 2/3\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Entraînement: 100%|██████████| 2793/2793 [18:26<00:00,  2.53it/s]\n"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "📉 Loss moyenne : 0.4711\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Évaluation: 100%|██████████| 699/699 [01:49<00:00,  6.39it/s]\n"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "🎯 F1-score (weighted) : 0.8735\n",
+      "⏳ EarlyStopping patience : 1/2\n",
+      "\n",
+      "🌟 Epoch 3/3\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Entraînement: 100%|██████████| 2793/2793 [18:26<00:00,  2.52it/s]\n"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "📉 Loss moyenne : 0.4485\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Évaluation: 100%|██████████| 699/699 [01:50<00:00,  6.35it/s]\n"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "🎯 F1-score (weighted) : 0.8816\n",
+      "⏳ EarlyStopping patience : 2/2\n",
+      "🛑 Arrêt anticipé — pas d'amélioration\n"
+     ]
+    }
+   ],
+   "source": [
+    "best_f1 = 0\n",
+    "patience_counter = 0\n",
+    "os.makedirs(\"outputs/model\", exist_ok=True)\n",
+    "\n",
+    "for epoch in range(EPOCHS):\n",
+    "    print(f\"\\n🌟 Epoch {epoch + 1}/{EPOCHS}\")\n",
+    "    model.train()\n",
+    "    total_loss = 0\n",
+    "\n",
+    "    for batch in tqdm(train_loader, desc=\"Entraînement\"):\n",
+    "        batch = {k: v.to(device) for k, v in batch.items()}\n",
+    "        logits = model(**batch).logits\n",
+    "        loss = loss_fn(logits, batch[\"labels\"])\n",
+    "        loss.backward()\n",
+    "        optimizer.step()\n",
+    "        scheduler.step()\n",
+    "        optimizer.zero_grad()\n",
+    "        total_loss += loss.item()\n",
+    "\n",
+    "    avg_loss = total_loss / len(train_loader)\n",
+    "    print(f\"📉 Loss moyenne : {avg_loss:.4f}\")\n",
+    "\n",
+    "    # 🔍 Évaluation\n",
+    "    model.eval()\n",
+    "    y_true, y_pred = [], []\n",
+    "    with torch.no_grad():\n",
+    "        for batch in tqdm(val_loader, desc=\"Évaluation\"):\n",
+    "            batch = {k: v.to(device) for k, v in batch.items()}\n",
+    "            logits = model(**batch).logits\n",
+    "            preds = torch.argmax(logits, dim=1)\n",
+    "            y_true.extend(batch[\"labels\"].cpu().tolist())\n",
+    "            y_pred.extend(preds.cpu().tolist())\n",
+    "\n",
+    "    report = classification_report(y_true, y_pred, target_names=[\"Non toxique\", \"Toxique\"], output_dict=True)\n",
+    "    f1 = report[\"weighted avg\"][\"f1-score\"]\n",
+    "    print(f\"🎯 F1-score (weighted) : {f1:.4f}\")\n",
+    "\n",
+    "    if f1 > best_f1:\n",
+    "        best_f1 = f1\n",
+    "        patience_counter = 0\n",
+    "        print(\"✅ Nouveau meilleur modèle — sauvegarde manuelle...\")\n",
+    "\n",
+    "        import os\n",
+    "\n",
+    "        # 📂 Dossier de sauvegarde\n",
+    "        save_dir = \"outputs/model\"\n",
+    "        os.makedirs(save_dir, exist_ok=True)\n",
+    "\n",
+    "        # 💾 Sauvegarde manuelle des poids\n",
+    "        torch.save(model.state_dict(), os.path.join(save_dir, \"pytorch_model.bin\"))\n",
+    "\n",
+    "        # 💾 Sauvegarde de la configuration du modèle\n",
+    "        model.config.to_json_file(os.path.join(save_dir, \"config.json\"))\n",
+    "\n",
+    "        # 💾 Sauvegarde du tokenizer\n",
+    "        tokenizer.save_pretrained(save_dir)\n",
+    "\n",
+    "        # 💾 Sauvegarde des métriques\n",
+    "        with open(\"outputs/metrics.json\", \"w\") as f:\n",
+    "            json.dump(report, f, indent=4)\n",
+    "\n",
+    "    else:\n",
+    "        patience_counter += 1\n",
+    "        print(f\"⏳ EarlyStopping patience : {patience_counter}/{PATIENCE}\")\n",
+    "        if patience_counter >= PATIENCE:\n",
+    "            print(\"🛑 Arrêt anticipé — pas d'amélioration\")\n",
+    "            break"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "id": "ba6f2d7c-0daf-48db-96a2-33935dca1d9e",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "📊 Métriques sauvegardées :\n",
+      "\n",
+      "🗂 Classe : Non toxique\n",
+      "  🔸 Précision : 0.9294\n",
+      "  🔸 Rappel     : 0.9329\n",
+      "  🔸 F1-score   : 0.9312\n",
+      "\n",
+      "🗂 Classe : Toxique\n",
+      "  🔸 Précision : 0.6193\n",
+      "  🔸 Rappel     : 0.6065\n",
+      "  🔸 F1-score   : 0.6129\n",
+      "\n",
+      "🔄 Moyennes pondérées (weighted avg) :\n",
+      "  ✅ Précision : 0.8821\n",
+      "  ✅ Rappel    : 0.8831\n",
+      "  ✅ F1-score  : 0.8826\n"
+     ]
+    }
+   ],
+   "source": [
+    "import json\n",
+    "import os\n",
+    "\n",
+    "# 📁 Chemin du fichier de métriques\n",
+    "metrics_path = \"outputs/metrics.json\"\n",
+    "\n",
+    "# ✅ Vérifie l'existence du fichier\n",
+    "if os.path.exists(metrics_path):\n",
+    "    with open(metrics_path, \"r\") as f:\n",
+    "        metrics = json.load(f)\n",
+    "\n",
+    "    print(\"📊 Métriques sauvegardées :\\n\")\n",
+    "    for label in [\"Non toxique\", \"Toxique\"]:\n",
+    "        print(f\"🗂 Classe : {label}\")\n",
+    "        print(f\"  🔸 Précision : {metrics[label]['precision']:.4f}\")\n",
+    "        print(f\"  🔸 Rappel     : {metrics[label]['recall']:.4f}\")\n",
+    "        print(f\"  🔸 F1-score   : {metrics[label]['f1-score']:.4f}\\n\")\n",
+    "\n",
+    "    print(\"🔄 Moyennes pondérées (weighted avg) :\")\n",
+    "    print(f\"  ✅ Précision : {metrics['weighted avg']['precision']:.4f}\")\n",
+    "    print(f\"  ✅ Rappel    : {metrics['weighted avg']['recall']:.4f}\")\n",
+    "    print(f\"  ✅ F1-score  : {metrics['weighted avg']['f1-score']:.4f}\")\n",
+    "else:\n",
+    "    print(\"❌ Aucune métrique trouvée dans outputs/metrics.json\")"
+   ]
+  }
+ ],
+ "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": "3.10.12"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}