dataset translation notebook

notebooks/DatasetTranslation.ipynb

@@ -0,0 +1,837 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "id": "f5e0b745",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from transformers import T5Tokenizer, T5ForConditionalGeneration\n",
+    "from transformers import AutoModel, AutoTokenizer\n",
+    "from transformers import MarianMTModel\n",
+    "from transformers import XLMRobertaForSequenceClassification\n",
+    "\n",
+    "\n",
+    "import os\n",
+    "import pandas as pd\n",
+    "from tqdm import tqdm\n",
+    "\n",
+    "tqdm.pandas()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "id": "6e228528",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def detect_language(model, tokenizer, text):\n",
+    "    \n",
+    "    token_dict = tokenizer(text, return_tensors=\"pt\").to(\"cuda\")\n",
+    "    outputs = model(token_dict.input_ids)\n",
+    "    decoded = outputs.logits.argmax(-1).item()\n",
+    "    lang = model.config.id2label[decoded]\n",
+    "    return lang\n",
+    "\n",
+    "def translate(model, tokenizer, text):\n",
+    "    input_ids = tokenizer(text, return_tensors=\"pt\").input_ids.to(\"cuda\")\n",
+    "    outputs = model.generate(input_ids)\n",
+    "    decoded = tokenizer.decode(outputs[0], skip_special_tokens=True)\n",
+    "    return decoded"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 24,
+   "id": "454ae3b6",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<div>\n",
+       "<style scoped>\n",
+       "    .dataframe tbody tr th:only-of-type {\n",
+       "        vertical-align: middle;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe tbody tr th {\n",
+       "        vertical-align: top;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe thead th {\n",
+       "        text-align: right;\n",
+       "    }\n",
+       "</style>\n",
+       "<table border=\"1\" class=\"dataframe\">\n",
+       "  <thead>\n",
+       "    <tr style=\"text-align: right;\">\n",
+       "      <th></th>\n",
+       "      <th>text</th>\n",
+       "      <th>label</th>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>INDEX</th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "    <tr>\n",
+       "      <th>0</th>\n",
+       "      <td>Maybe the UN could talk to those asian and afr...</td>\n",
+       "      <td>1</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1</th>\n",
+       "      <td>‘Juice (Jews) are worse that nukes.’ Caption t...</td>\n",
+       "      <td>1</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>2</th>\n",
+       "      <td>Foreigners on UK benefits leaps 41% in 5 years...</td>\n",
+       "      <td>1</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>3</th>\n",
+       "      <td>Jews cannot be patriots, since their allegianc...</td>\n",
+       "      <td>1</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>4</th>\n",
+       "      <td>Israel, Zionism and Jewish Supremacist control...</td>\n",
+       "      <td>1</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>...</th>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>4998</th>\n",
+       "      <td>Woman are basically childlike.</td>\n",
+       "      <td>1</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>4999</th>\n",
+       "      <td>women should be submissive to men.</td>\n",
+       "      <td>1</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>5000</th>\n",
+       "      <td>Women are essentially childlike, unable to und...</td>\n",
+       "      <td>1</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>5001</th>\n",
+       "      <td>Women should be submissive to men.</td>\n",
+       "      <td>1</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>5002</th>\n",
+       "      <td>Women should not be allowed to vote, this is a...</td>\n",
+       "      <td>1</td>\n",
+       "    </tr>\n",
+       "  </tbody>\n",
+       "</table>\n",
+       "<p>5003 rows × 2 columns</p>\n",
+       "</div>"
+      ],
+      "text/plain": [
+       "                                                    text  label\n",
+       "INDEX                                                          \n",
+       "0      Maybe the UN could talk to those asian and afr...      1\n",
+       "1      ‘Juice (Jews) are worse that nukes.’ Caption t...      1\n",
+       "2      Foreigners on UK benefits leaps 41% in 5 years...      1\n",
+       "3      Jews cannot be patriots, since their allegianc...      1\n",
+       "4      Israel, Zionism and Jewish Supremacist control...      1\n",
+       "...                                                  ...    ...\n",
+       "4998                      Woman are basically childlike.      1\n",
+       "4999                  women should be submissive to men.      1\n",
+       "5000   Women are essentially childlike, unable to und...      1\n",
+       "5001                  Women should be submissive to men.      1\n",
+       "5002   Women should not be allowed to vote, this is a...      1\n",
+       "\n",
+       "[5003 rows x 2 columns]"
+      ]
+     },
+     "execution_count": 24,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "DATASET_PATH = \"../data/processed/\"\n",
+    "\n",
+    "df = pd.read_csv(DATASET_PATH, index_col=[0])\n",
+    "df"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "f2db3a80",
+   "metadata": {
+    "scrolled": true
+   },
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "0         2.0\n",
+       "1         2.0\n",
+       "2         4.0\n",
+       "3         2.0\n",
+       "4         4.0\n",
+       "         ... \n",
+       "135551    2.0\n",
+       "135552    2.0\n",
+       "135553    1.0\n",
+       "135554    2.0\n",
+       "135555    2.0\n",
+       "Name: status, Length: 135556, dtype: float64"
+      ]
+     },
+     "execution_count": 9,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "# PREPROCESSING\n",
+    "# Need more preprocessing :\n",
+    "#  - stopwords : Name, URL\n",
+    "\n",
+    "# Worst 10%\n",
+    "# df = df.loc[df.hate_speech_score > df.hate_speech_score.quantile(0.90)]\n",
+    "\n",
+    "# Get negative examples\n",
+    "# n = 12785 # = number of gathered worst examples\n",
+    "# df = df.loc[df.hate_speech_score < df.hate_speech_score.quantile(0.1)].iloc[:n]\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "5ca78b94",
+   "metadata": {},
+   "source": [
+    "#### Language detection"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 21,
+   "id": "69fc7f94",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Language detection model\n",
+    "tokenizer = AutoTokenizer.from_pretrained(\"papluca/xlm-roberta-base-language-detection\")\n",
+    "model = XLMRobertaForSequenceClassification.from_pretrained(\"papluca/xlm-roberta-base-language-detection\").to(\"cuda\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 59,
+   "id": "f05bbff2",
+   "metadata": {
+    "scrolled": false
+   },
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "100%|████████████████████████████████████████████████████████████████████████████████████████████| 135556/135556 [18:54<00:00, 119.45it/s]\n"
+     ]
+    }
+   ],
+   "source": [
+    "# Add language prediction for later filtering before applying language-dependent translation model\n",
+    "df[\"lang\"] = df.text.progress_apply(lambda text: detect_language(model, tokenizer, text))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "05d07356",
+   "metadata": {},
+   "source": [
+    "#### Text translation\n",
+    "\n",
+    "- Process requires long processing time\n",
+    "- Process should be run multiple times to complete"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 19,
+   "id": "a82a5eb3",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "zsh:1: no matches found: temp/*\r\n"
+     ]
+    }
+   ],
+   "source": [
+    "! rm -rf temp/*"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 25,
+   "id": "3f3f1bfb",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def write_file(path, text):\n",
+    "    with open(path, \"w\") as f:\n",
+    "        f.write(text)\n",
+    "\n",
+    "LANGUAGE = \"en\"\n",
+    "OUTPUT_DIR = \"../data/translated\"\n",
+    "TEMP_DIR = \"../data/temp\"\n",
+    "IN_MEMORY = True\n",
+    "\n",
+    "output_dir = os.path.join(OUTPUT_DIR, LANGUAGE)\n",
+    "if not os.path.exists(OUTPUT_DIR):\n",
+    "    os.makedirs(output_dir)\n",
+    "#     os.mkdir(OUTPUT_DIR)\n",
+    "if not os.path.exists(TEMP_DIR):\n",
+    "    os.mkdir(TEMP_DIR)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "id": "b8122256",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Translation model\n",
+    "tokenizer = AutoTokenizer.from_pretrained(\"Abelll/marian-finetuned-kde4-en-to-fr\")\n",
+    "model = MarianMTModel.from_pretrained(\"Abelll/marian-finetuned-kde4-en-to-fr\").to(\"cuda\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 26,
+   "id": "b9ef106d",
+   "metadata": {
+    "scrolled": true
+   },
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "  0%|                                                                                              | 0/5003 [00:00<?, ?it/s]/home/louis/miniconda3/envs/sexism_detection/lib/python3.9/site-packages/transformers/generation/utils.py:1387: UserWarning: Neither `max_length` nor `max_new_tokens` has been set, `max_length` will default to 512 (`self.config.max_length`). Controlling `max_length` via the config is deprecated and `max_length` will be removed from the config in v5 of Transformers -- we recommend using `max_new_tokens` to control the maximum length of the generation.\n",
+      "  warnings.warn(\n",
+      "100%|███████████████████████████████████████████████████████████████████████████████████| 5003/5003 [10:27<00:00,  7.97it/s]\n"
+     ]
+    }
+   ],
+   "source": [
+    "translated_texts = []\n",
+    "for i, row in tqdm(df.iterrows(), total=df.shape[0]):\n",
+    "    output_path = os.path.join(TEMP_DIR, str(i) + \".txt\")\n",
+    "    if not IN_MEMORY:\n",
+    "        if os.path.exists(output_path):\n",
+    "            print(\"Already translated !\")\n",
+    "            continue\n",
+    "    translated = translate(model, tokenizer, row.text)\n",
+    "    if IN_MEMORY:\n",
+    "        translated_texts.append((i, translated))\n",
+    "    else:\n",
+    "        write_file(output_path, translated)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 29,
+   "id": "0ecc3000",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import glob\n",
+    "\n",
+    "output_filename = DATASET_PATH\n",
+    "\n",
+    "if not IN_MEMORY:\n",
+    "    filelist = glob.glob(TEMP_DIR + \"/*\")\n",
+    "    print(f\"Found {len(filelist)}\")\n",
+    "    filelist = sorted(filelist, key=lambda x: int(x.split(\"/\")[-1].split(\".\")[0]))\n",
+    "    translated_texts = []\n",
+    "    for text_file in tqdm(filelist):\n",
+    "        with open(text_file) as f:\n",
+    "            translated_text = f.read().splitlines()\n",
+    "        translated_texts.append(translated_text)\n",
+    "    translated_texts = pd.DataFrame(translated_texts)\n",
+    "    df[\"translated\"] = translated_texts\n",
+    "\n",
+    "df[\"translated\"] = translated_texts\n",
+    "df.translated = df.translated.apply(lambda x: x[1])\n",
+    "df.dropna().to_csv(output_filename)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 30,
+   "id": "88c69fd0",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<div>\n",
+       "<style scoped>\n",
+       "    .dataframe tbody tr th:only-of-type {\n",
+       "        vertical-align: middle;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe tbody tr th {\n",
+       "        vertical-align: top;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe thead th {\n",
+       "        text-align: right;\n",
+       "    }\n",
+       "</style>\n",
+       "<table border=\"1\" class=\"dataframe\">\n",
+       "  <thead>\n",
+       "    <tr style=\"text-align: right;\">\n",
+       "      <th></th>\n",
+       "      <th>text</th>\n",
+       "      <th>label</th>\n",
+       "      <th>translated</th>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>INDEX</th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "      <th></th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "    <tr>\n",
+       "      <th>0</th>\n",
+       "      <td>Maybe the UN could talk to those asian and afr...</td>\n",
+       "      <td>1</td>\n",
+       "      <td>Peut-être que l'Onu pourrait parler aux nation...</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1</th>\n",
+       "      <td>‘Juice (Jews) are worse that nukes.’ Caption t...</td>\n",
+       "      <td>1</td>\n",
+       "      <td>« Juice (Juifs) sont pire que les nuées. « &amp; #...</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>2</th>\n",
+       "      <td>Foreigners on UK benefits leaps 41% in 5 years...</td>\n",
+       "      <td>1</td>\n",
+       "      <td>Les étrangers au Royaume-Uni bénéficient d'un ...</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>3</th>\n",
+       "      <td>Jews cannot be patriots, since their allegianc...</td>\n",
+       "      <td>1</td>\n",
+       "      <td>Les Juifs ne peuvent pas être patriotes, car l...</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>4</th>\n",
+       "      <td>Israel, Zionism and Jewish Supremacist control...</td>\n",
+       "      <td>1</td>\n",
+       "      <td>Israël, le sionisme et le contrôle suprémacist...</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>...</th>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>4998</th>\n",
+       "      <td>Woman are basically childlike.</td>\n",
+       "      <td>1</td>\n",
+       "      <td>La femme ressemble à un enfant.</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>4999</th>\n",
+       "      <td>women should be submissive to men.</td>\n",
+       "      <td>1</td>\n",
+       "      <td>Les filles doivent être soumises aux hommes.</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>5000</th>\n",
+       "      <td>Women are essentially childlike, unable to und...</td>\n",
+       "      <td>1</td>\n",
+       "      <td>Les filles sont pour l'essentiel des filles, i...</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>5001</th>\n",
+       "      <td>Women should be submissive to men.</td>\n",
+       "      <td>1</td>\n",
+       "      <td>Les filles doivent être soumises aux hommes.</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>5002</th>\n",
+       "      <td>Women should not be allowed to vote, this is a...</td>\n",
+       "      <td>1</td>\n",
+       "      <td>Les femmes ne devraient pas être autorisées à ...</td>\n",
+       "    </tr>\n",
+       "  </tbody>\n",
+       "</table>\n",
+       "<p>5003 rows × 3 columns</p>\n",
+       "</div>"
+      ],
+      "text/plain": [
+       "                                                    text  label  \\\n",
+       "INDEX                                                             \n",
+       "0      Maybe the UN could talk to those asian and afr...      1   \n",
+       "1      ‘Juice (Jews) are worse that nukes.’ Caption t...      1   \n",
+       "2      Foreigners on UK benefits leaps 41% in 5 years...      1   \n",
+       "3      Jews cannot be patriots, since their allegianc...      1   \n",
+       "4      Israel, Zionism and Jewish Supremacist control...      1   \n",
+       "...                                                  ...    ...   \n",
+       "4998                      Woman are basically childlike.      1   \n",
+       "4999                  women should be submissive to men.      1   \n",
+       "5000   Women are essentially childlike, unable to und...      1   \n",
+       "5001                  Women should be submissive to men.      1   \n",
+       "5002   Women should not be allowed to vote, this is a...      1   \n",
+       "\n",
+       "                                              translated  \n",
+       "INDEX                                                     \n",
+       "0      Peut-être que l'Onu pourrait parler aux nation...  \n",
+       "1      « Juice (Juifs) sont pire que les nuées. « & #...  \n",
+       "2      Les étrangers au Royaume-Uni bénéficient d'un ...  \n",
+       "3      Les Juifs ne peuvent pas être patriotes, car l...  \n",
+       "4      Israël, le sionisme et le contrôle suprémacist...  \n",
+       "...                                                  ...  \n",
+       "4998                     La femme ressemble à un enfant.  \n",
+       "4999        Les filles doivent être soumises aux hommes.  \n",
+       "5000   Les filles sont pour l'essentiel des filles, i...  \n",
+       "5001        Les filles doivent être soumises aux hommes.  \n",
+       "5002   Les femmes ne devraient pas être autorisées à ...  \n",
+       "\n",
+       "[5003 rows x 3 columns]"
+      ]
+     },
+     "execution_count": 30,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "df"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "id": "885ebb18",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "df.to_csv(\"../data/ucberkeley-dlab_measuring-hate-speech_translated_positive.csv\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 65,
+   "id": "576a76f1",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "translate English to French: for starters bend over the one in pink and kick that ass and pussy to get a taste until she's begging for a dick inside her.\n"
+     ]
+    },
+    {
+     "data": {
+      "text/plain": [
+       "\"j'ai lu cette vidéo et j'ai\""
+      ]
+     },
+     "execution_count": 65,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "def translate(text):\n",
+    "    input_text = f\"translate {BASE_LANGUAGE} to French: \" + text.lower()\n",
+    "    print(input_text)\n",
+    "    input_ids = tokenizer(input_text, return_tensors=\"pt\").input_ids\n",
+    "    outputs = model.generate(input_ids)\n",
+    "    decoded = tokenizer.decode(outputs[0], skip_special_tokens=True)\n",
+    "    return decoded\n",
+    "\n",
+    "translate(df.text[4])"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "id": "1bbc6dab",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "MarianMTModel(\n",
+       "  (model): MarianModel(\n",
+       "    (shared): Embedding(59514, 512, padding_idx=59513)\n",
+       "    (encoder): MarianEncoder(\n",
+       "      (embed_tokens): Embedding(59514, 512, padding_idx=59513)\n",
+       "      (embed_positions): MarianSinusoidalPositionalEmbedding(512, 512)\n",
+       "      (layers): ModuleList(\n",
+       "        (0): MarianEncoderLayer(\n",
+       "          (self_attn): MarianAttention(\n",
+       "            (k_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (v_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (q_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (out_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "          )\n",
+       "          (self_attn_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "          (activation_fn): SiLUActivation()\n",
+       "          (fc1): Linear(in_features=512, out_features=2048, bias=True)\n",
+       "          (fc2): Linear(in_features=2048, out_features=512, bias=True)\n",
+       "          (final_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "        )\n",
+       "        (1): MarianEncoderLayer(\n",
+       "          (self_attn): MarianAttention(\n",
+       "            (k_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (v_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (q_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (out_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "          )\n",
+       "          (self_attn_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "          (activation_fn): SiLUActivation()\n",
+       "          (fc1): Linear(in_features=512, out_features=2048, bias=True)\n",
+       "          (fc2): Linear(in_features=2048, out_features=512, bias=True)\n",
+       "          (final_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "        )\n",
+       "        (2): MarianEncoderLayer(\n",
+       "          (self_attn): MarianAttention(\n",
+       "            (k_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (v_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (q_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (out_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "          )\n",
+       "          (self_attn_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "          (activation_fn): SiLUActivation()\n",
+       "          (fc1): Linear(in_features=512, out_features=2048, bias=True)\n",
+       "          (fc2): Linear(in_features=2048, out_features=512, bias=True)\n",
+       "          (final_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "        )\n",
+       "        (3): MarianEncoderLayer(\n",
+       "          (self_attn): MarianAttention(\n",
+       "            (k_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (v_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (q_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (out_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "          )\n",
+       "          (self_attn_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "          (activation_fn): SiLUActivation()\n",
+       "          (fc1): Linear(in_features=512, out_features=2048, bias=True)\n",
+       "          (fc2): Linear(in_features=2048, out_features=512, bias=True)\n",
+       "          (final_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "        )\n",
+       "        (4): MarianEncoderLayer(\n",
+       "          (self_attn): MarianAttention(\n",
+       "            (k_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (v_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (q_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (out_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "          )\n",
+       "          (self_attn_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "          (activation_fn): SiLUActivation()\n",
+       "          (fc1): Linear(in_features=512, out_features=2048, bias=True)\n",
+       "          (fc2): Linear(in_features=2048, out_features=512, bias=True)\n",
+       "          (final_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "        )\n",
+       "        (5): MarianEncoderLayer(\n",
+       "          (self_attn): MarianAttention(\n",
+       "            (k_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (v_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (q_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (out_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "          )\n",
+       "          (self_attn_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "          (activation_fn): SiLUActivation()\n",
+       "          (fc1): Linear(in_features=512, out_features=2048, bias=True)\n",
+       "          (fc2): Linear(in_features=2048, out_features=512, bias=True)\n",
+       "          (final_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "        )\n",
+       "      )\n",
+       "    )\n",
+       "    (decoder): MarianDecoder(\n",
+       "      (embed_tokens): Embedding(59514, 512, padding_idx=59513)\n",
+       "      (embed_positions): MarianSinusoidalPositionalEmbedding(512, 512)\n",
+       "      (layers): ModuleList(\n",
+       "        (0): MarianDecoderLayer(\n",
+       "          (self_attn): MarianAttention(\n",
+       "            (k_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (v_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (q_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (out_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "          )\n",
+       "          (activation_fn): SiLUActivation()\n",
+       "          (self_attn_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "          (encoder_attn): MarianAttention(\n",
+       "            (k_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (v_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (q_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (out_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "          )\n",
+       "          (encoder_attn_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "          (fc1): Linear(in_features=512, out_features=2048, bias=True)\n",
+       "          (fc2): Linear(in_features=2048, out_features=512, bias=True)\n",
+       "          (final_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "        )\n",
+       "        (1): MarianDecoderLayer(\n",
+       "          (self_attn): MarianAttention(\n",
+       "            (k_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (v_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (q_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (out_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "          )\n",
+       "          (activation_fn): SiLUActivation()\n",
+       "          (self_attn_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "          (encoder_attn): MarianAttention(\n",
+       "            (k_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (v_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (q_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (out_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "          )\n",
+       "          (encoder_attn_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "          (fc1): Linear(in_features=512, out_features=2048, bias=True)\n",
+       "          (fc2): Linear(in_features=2048, out_features=512, bias=True)\n",
+       "          (final_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "        )\n",
+       "        (2): MarianDecoderLayer(\n",
+       "          (self_attn): MarianAttention(\n",
+       "            (k_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (v_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (q_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (out_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "          )\n",
+       "          (activation_fn): SiLUActivation()\n",
+       "          (self_attn_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "          (encoder_attn): MarianAttention(\n",
+       "            (k_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (v_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (q_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (out_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "          )\n",
+       "          (encoder_attn_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "          (fc1): Linear(in_features=512, out_features=2048, bias=True)\n",
+       "          (fc2): Linear(in_features=2048, out_features=512, bias=True)\n",
+       "          (final_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "        )\n",
+       "        (3): MarianDecoderLayer(\n",
+       "          (self_attn): MarianAttention(\n",
+       "            (k_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (v_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (q_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (out_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "          )\n",
+       "          (activation_fn): SiLUActivation()\n",
+       "          (self_attn_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "          (encoder_attn): MarianAttention(\n",
+       "            (k_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (v_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (q_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (out_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "          )\n",
+       "          (encoder_attn_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "          (fc1): Linear(in_features=512, out_features=2048, bias=True)\n",
+       "          (fc2): Linear(in_features=2048, out_features=512, bias=True)\n",
+       "          (final_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "        )\n",
+       "        (4): MarianDecoderLayer(\n",
+       "          (self_attn): MarianAttention(\n",
+       "            (k_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (v_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (q_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (out_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "          )\n",
+       "          (activation_fn): SiLUActivation()\n",
+       "          (self_attn_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "          (encoder_attn): MarianAttention(\n",
+       "            (k_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (v_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (q_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (out_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "          )\n",
+       "          (encoder_attn_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "          (fc1): Linear(in_features=512, out_features=2048, bias=True)\n",
+       "          (fc2): Linear(in_features=2048, out_features=512, bias=True)\n",
+       "          (final_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "        )\n",
+       "        (5): MarianDecoderLayer(\n",
+       "          (self_attn): MarianAttention(\n",
+       "            (k_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (v_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (q_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (out_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "          )\n",
+       "          (activation_fn): SiLUActivation()\n",
+       "          (self_attn_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "          (encoder_attn): MarianAttention(\n",
+       "            (k_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (v_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (q_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "            (out_proj): Linear(in_features=512, out_features=512, bias=True)\n",
+       "          )\n",
+       "          (encoder_attn_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "          (fc1): Linear(in_features=512, out_features=2048, bias=True)\n",
+       "          (fc2): Linear(in_features=2048, out_features=512, bias=True)\n",
+       "          (final_layer_norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
+       "        )\n",
+       "      )\n",
+       "    )\n",
+       "  )\n",
+       "  (lm_head): Linear(in_features=512, out_features=59514, bias=False)\n",
+       ")"
+      ]
+     },
+     "execution_count": 8,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "model"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "sexism_detection",
+   "language": "python",
+   "name": "sexism_detection"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.9.15"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}