{"cells":[{"metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true},"cell_type":"code","source":"# This Python 3 environment comes with many helpful analytics libraries installed\n# It is defined by the kaggle/python Docker image: https://github.com/kaggle/docker-python\n# For example, here's several helpful packages to load\n\nimport numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\nfrom sklearn.model_selection import train_test_split\n\nfrom transformers import XLMRobertaTokenizer\nfrom transformers import TFXLMRobertaForSequenceClassification\n\nfrom keras.models import Sequential\nfrom keras.layers.recurrent import SimpleRNN,LSTM,GRU\nfrom keras.layers.core import Dense\nfrom keras.layers.embeddings import Embedding\nfrom keras.layers import SpatialDropout1D\nfrom keras.callbacks import EarlyStopping\nfrom keras.preprocessing import sequence, text\nfrom tqdm import tqdm\nimport numpy as np\nimport tensorflow as tf\n# Input data files are available in the read-only \"../input/\" directory\n# For example, running this (by clicking run or pressing Shift+Enter) will list all files under the input directory\n\nimport os\nfor dirname, _, filenames in os.walk('/kaggle/input'):\n    for filename in filenames:\n        print(os.path.join(dirname, filename))\n\n# You can write up to 5GB to the current directory (/kaggle/working/) that gets preserved as output when you create a version using \"Save & Run All\" \n# You can also write temporary files to /kaggle/temp/, but they won't be saved outside of the current session","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# Detect hardware, return appropriate distribution strategy\ntry:\n    # TPU detection. No parameters necessary if TPU_NAME environment variable is\n    # set: this is always the case on Kaggle.\n    tpu = tf.distribute.cluster_resolver.TPUClusterResolver()\n    print('Running on TPU ', tpu.master())\nexcept ValueError:\n    tpu = None\n\nif tpu:\n    tf.config.experimental_connect_to_cluster(tpu)\n    tf.tpu.experimental.initialize_tpu_system(tpu)\n    strategy = tf.distribute.experimental.TPUStrategy(tpu)\nelse:\n    # Default distribution strategy in Tensorflow. Works on CPU and single GPU.\n    strategy = tf.distribute.get_strategy()\n\nprint(\"REPLICAS: \", strategy.num_replicas_in_sync)","execution_count":null,"outputs":[]},{"metadata":{"_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","trusted":true},"cell_type":"code","source":"train = pd.read_csv('../input/jigsaw-multilingual-toxic-comment-classification/jigsaw-toxic-comment-train.csv')\nvalidation = pd.read_csv('../input/jigsaw-multilingual-toxic-comment-classification/validation.csv')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train.drop(['severe_toxic','obscene','threat','insult','identity_hate'],axis=1,inplace=True)\ntrain.head()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train.toxic.value_counts()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"validation.head()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"validation.toxic.value_counts()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"X1 = train[train.toxic == 0].iloc[:20000]['comment_text'].to_list()\nY1 =[0] * 20000\nX2 = train[train.toxic == 1].iloc[:20000]['comment_text'].to_list()\nY2 =[1] * 20000\nx_train = X1 + X2\ny_train = Y1 + Y2\nidx_remove_validation = validation[validation.toxic == 0].iloc[1230:].index.to_list()\nvalidation.drop(idx_remove_validation,inplace=True)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"validation.toxic.value_counts()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"token = text.Tokenizer(num_words=None)\nmax_len = 500\n\ntoken.fit_on_texts(train.comment_text.to_list() + validation.comment_text.to_list() +test.content.to_list())\nxtrain_seq = token.texts_to_sequences(x_train)\nxvalid_seq = token.texts_to_sequences(validation.comment_text.to_list())\n\n#zero pad the sequences\nxtrain_pad = sequence.pad_sequences(xtrain_seq, maxlen=max_len)\nxvalid_pad = sequence.pad_sequences(xvalid_seq, maxlen=max_len)\n\nword_index = token.word_index","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"len([x for x in x_train if (len(x.split()) > 250)])","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"max_seq_len = 250\nx_train_tk = xlm_roberta_tokenizer.batch_encode_plus(x_train, max_length=max_seq_len, pad_to_max_length=True, \n                             return_attention_mask=True)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"x_val_tk = xlm_roberta_tokenizer.batch_encode_plus(validation.comment_text.to_list(), max_length=max_seq_len, pad_to_max_length=True, \n                             return_attention_mask=True)\ny_val = validation.toxic.to_list()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"word_index = xlm_roberta_tokenizer.get_vocab()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"embeddings_index = {}\nf = open('/kaggle/input/mutli-vec/wiki.multi.en.vec','r',encoding='utf-8')\nfor line in tqdm(f):\n    values = line.split(' ')\n    word = values[0]\n    coefs = np.asarray([float(val) for val in values[1:]])\n    embeddings_index[word] = coefs\nf.close()\n\nprint('Found %s word vectors.' % len(embeddings_index))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"embedding_matrix = np.zeros((len(word_index) + 1, 300))\nfor word, i in tqdm(word_index.items()):\n    embedding_vector = embeddings_index.get(word)\n    if embedding_vector is not None:\n        embedding_matrix[i] = embedding_vector","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"%%time\n# A simpleRNN without any pretrained embeddings and one dense layer\nmodel = Sequential()\nmodel.add(Embedding(len(word_index) + 1,\n                    300,\n                    input_length=max_seq_len,\n                    weights=[embedding_matrix],\n                    trainable=False))\nmodel.add(SimpleRNN(100))\nmodel.add(Dense(1, activation='sigmoid'))\nmodel.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n    \nmodel.summary()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"%%time\n# A LSTM without any pretrained embeddings and one dense layer\nmodel = Sequential()\nmodel.add(Embedding(len(word_index) + 1,\n                    300,\n                    input_length=max_len))\nmodel.add(LSTM(100, dropout=0.4, recurrent_dropout=0.4))\nmodel.add(Dense(1, activation='sigmoid'))\nmodel.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n    \nmodel.summary()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"%%time\n# A GRU with pretrained embeddings and one dense layer\nmodel = Sequential()\nmodel.add(Embedding(len(word_index) + 1,\n                    300,\n                    input_length=max_seq_len,\n                    weights=[embedding_matrix],\n                    trainable=False))\nmodel.add(SpatialDropout1D(0.3))\nmodel.add(GRU(300))\nmodel.add(Dense(1, activation='sigmoid'))\nmodel.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n    \nmodel.summary()\n","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"callbacks = [EarlyStopping(patience=3)]\n\nhistory = model.fit(xtrain_pad, y_train, epochs=1000, verbose=1,\n          validation_data=(xvalid_pad,validation.toxic.to_list()),batch_size=1024, callbacks=callbacks)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"callbacks = [EarlyStopping(patience=3)]\n\nhistory = model.fit(np.array(x_train_tk['input_ids']), y_train, epochs=1000, verbose=1,\n          validation_data=(np.array(x_val_tk['input_ids']),y_val),batch_size=4096, callbacks=callbacks)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"test =pd.read_csv('../input/jigsaw-multilingual-toxic-comment-classification/test.csv')\nx_test_tk = xlm_roberta_tokenizer.batch_encode_plus(test.content.to_list(), max_length=max_seq_len, pad_to_max_length=True, \n                             return_attention_mask=True)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"from transformers import TFXLMRobertaForSequenceClassification,AutoTokenizer","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"max_seq_len =252\nxlm_roberta_tokenizer = AutoTokenizer.from_pretrained('jplu/tf-xlm-roberta-large')\nbase_model= TFXLMRobertaForSequenceClassification.from_pretrained('jplu/tf-xlm-roberta-large')\n\nx_train_tk = xlm_roberta_tokenizer.batch_encode_plus(x_train, max_length=max_seq_len,return_tensors='tf',\n                                                     pad_to_max_length=True)\nx_val_tk = xlm_roberta_tokenizer.batch_encode_plus(validation.comment_text.to_list(),return_tensors='tf',\n                                                   max_length=max_seq_len, pad_to_max_length=True)\ny_val = validation.toxic.to_list()\n\n# Prepare training: Compile tf.keras model with optimizer, loss and learning rate schedule\noptimizer = tf.keras.optimizers.Adam(learning_rate=3e-5, epsilon=1e-08, clipnorm=1.0)\nloss = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True)\nmetric = tf.keras.metrics.SparseCategoricalAccuracy('accuracy')\nbase_model.compile(optimizer=optimizer, loss=loss, metrics=[metric])\nbase_model.summary()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"callbacks = [tf.keras.callbacks.EarlyStopping(patience=3)]\n\nhistory = base_model.fit(x_train_tk['input_ids'], tf.convert_to_tensor(y_train), epochs=1000, verbose=1,\n          validation_data=(x_val_tk['input_ids'],tf.convert_to_tensor(y_val)),batch_size=20, callbacks=callbacks)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"x_train_tk.keys()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"test['toxic'] =np.round(model.predict(np.array(x_test_tk['input_ids']),batch_size=252),1)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"test.head()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"test.drop(['lang','content'],axis=1,inplace=True)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"test.to_csv('submission.csv',index=False)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"import pandas as pd\nimport numpy as np\n\nimport tensorflow as tf\nfrom tensorflow.keras.layers import Input,Dense\nfrom tensorflow.keras.optimizers import Adam\nfrom tensorflow.keras.callbacks import EarlyStopping\nfrom tensorflow.keras.models import Model\n\nfrom tokenizers import BertWordPieceTokenizer\nimport transformers\n\nfrom tqdm import tqdm","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# Detect hardware, return appropriate distribution strategy\ntry:\n    # TPU detection. No parameters necessary if TPU_NAME environment variable is\n    # set: this is always the case on Kaggle.\n    tpu = tf.distribute.cluster_resolver.TPUClusterResolver()\n    print('Running on TPU ', tpu.master())\nexcept ValueError:\n    tpu = None\n\nif tpu:\n    tf.config.experimental_connect_to_cluster(tpu)\n    tf.tpu.experimental.initialize_tpu_system(tpu)\n    strategy = tf.distribute.experimental.TPUStrategy(tpu)\nelse:\n    # Default distribution strategy in Tensorflow. Works on CPU and single GPU.\n    strategy = tf.distribute.get_strategy()\n\nprint(\"REPLICAS: \", strategy.num_replicas_in_sync)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"def build_model(transformer, max_len=512):\n    \"\"\"\n    function for training the BERT model\n    \"\"\"\n    input_word_ids = Input(shape=(max_len,), dtype=tf.int32, name=\"input_word_ids\")\n    sequence_output = transformer(input_word_ids)[0]\n    cls_token = sequence_output[:, 0, :]\n    out = Dense(1, activation='sigmoid')(cls_token)\n    \n    model = Model(inputs=input_word_ids, outputs=out)\n    model.compile(Adam(lr=1e-6), loss='binary_crossentropy', metrics=['accuracy'])\n    \n    return model","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"def regular_encode(texts, tokenizer, maxlen=512):\n    enc_di = tokenizer.batch_encode_plus(\n        texts, \n        return_attention_masks=False, \n        return_token_type_ids=False,\n        pad_to_max_length=True,\n        max_length=maxlen\n    )\n    \n    return np.array(enc_di['input_ids'])","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"def fast_encode(texts, tokenizer, chunk_size=256, maxlen=512):\n    \"\"\"\n    Encoder for encoding the text into sequence of integers for BERT Input\n    \"\"\"\n    tokenizer.enable_truncation(max_length=maxlen)\n    tokenizer.enable_padding(max_length=maxlen)\n    all_ids = []\n    \n    for i in tqdm(range(0, len(texts), chunk_size)):\n        text_chunk = texts[i:i+chunk_size].tolist()\n        encs = tokenizer.encode_batch(text_chunk)\n        all_ids.extend([enc.ids for enc in encs])\n    \n    return np.array(all_ids)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"test =pd.read_csv('../input/jigsaw-multilingual-toxic-comment-classification/test.csv')\ntrain = pd.read_csv('../input/jigsaw-multilingual-toxic-comment-classification/jigsaw-toxic-comment-train.csv')\nvalidation = pd.read_csv('../input/jigsaw-multilingual-toxic-comment-classification/validation.csv')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"test =pd.read_csv('../input/jigsaw-multilingual-toxic-comment-classification/test.csv')\ntrain1 = pd.read_csv('../input/jigsaw-multilingual-toxic-comment-classification/jigsaw-toxic-comment-train.csv')\ntrain2 = pd.read_csv('../input/jigsaw-multilingual-toxic-comment-classification/jigsaw-toxic-comment-train-processed-seqlen128.csv')\ntrain3 = pd.read_csv('../input/jigsaw-multilingual-toxic-comment-classification/jigsaw-unintended-bias-train-processed-seqlen128.csv')\ntrain4 = pd.read_csv('../input/jigsaw-multilingual-toxic-comment-classification/jigsaw-unintended-bias-train.csv')\nvalidation = pd.read_csv('../input/jigsaw-multilingual-toxic-comment-classification/validation.csv')\n\ntrain ","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"model_path = 'jplu/tf-xlm-roberta-base'\ntokenizer_path = 'jplu/tf-xlm-roberta-base'\ntransformer_model = transformers.TFAutoModelWithLMHead\ntokenizer = transformers.AutoTokenizer","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"tokenizer = tokenizer.from_pretrained(tokenizer_path)","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"**Fast encoding**","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"# First load the real tokenizer\ntokenizer = tokenizer.from_pretrained(tokenizer_path)\n# Save the loaded tokenizer locally\ntokenizer.save_pretrained('.')\n# Reload it with the huggingface tokenizers library\nfast_tokenizer = BertWordPieceTokenizer('vocab.txt', lowercase=False)\nfast_tokenizer","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"%%time\n\nmax_seq_len=196\nx_train = regular_encode(train.comment_text.astype(str), tokenizer, maxlen=max_seq_len)\nx_valid = regular_encode(validation.comment_text.astype(str), tokenizer, maxlen=max_seq_len)\nx_test = regular_encode(test.content.astype(str), tokenizer, maxlen=max_seq_len)\n\ny_train = train.toxic.values\ny_valid = validation.toxic.values","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"max_seq_len=196\nx_train = fast_encode(train.comment_text.astype(str), fast_tokenizer, maxlen=max_seq_len)\nx_valid = fast_encode(validation.comment_text.astype(str), fast_tokenizer, maxlen=max_seq_len)\nx_test = fast_encode(test.content.astype(str), fast_tokenizer, maxlen=max_seq_len)\n\ny_train = train.toxic.values\ny_valid = validation.toxic.values","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"%%time\n\nAUTO = tf.data.experimental.AUTOTUNE\nBATCH_SIZE = 8 * strategy.num_replicas_in_sync\n\ntrain_dataset = (\n    tf.data.Dataset\n    .from_tensor_slices((x_train, y_train))\n    .repeat()\n    .shuffle(2048)\n    .batch(BATCH_SIZE)\n    .prefetch(AUTO)\n)\n\nvalid_dataset = (\n    tf.data.Dataset\n    .from_tensor_slices((x_valid, y_valid))\n    .batch(BATCH_SIZE)\n    .cache()\n    .prefetch(AUTO)\n)\n\ntest_dataset = (\n    tf.data.Dataset\n    .from_tensor_slices(x_test)\n    .batch(BATCH_SIZE)\n)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"%%time\nwith strategy.scope():\n    transformer_layer = (transformer_model.from_pretrained(model_path))\n    model = build_model(transformer_layer,max_len=max_seq_len)\nmodel.summary()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"model.layers[1].trainable = False","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"Run for more epoch","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"callbacks = [tf.keras.callbacks.EarlyStopping(patience=3)]\nn_steps = len(x_train) // BATCH_SIZE\n\nhistory = model.fit(train_dataset, epochs=4, verbose=1,\n                    validation_data=valid_dataset,\n                    steps_per_epoch=n_steps,\n                    callbacks=callbacks)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"model.layers[1].trainable = True\nhistory = model.fit(train_dataset, epochs=1, verbose=1,\n                    validation_data=valid_dataset,\n                    steps_per_epoch=n_steps,\n                    callbacks=callbacks)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"model.layers[1].trainable = True\nn_steps = x_valid.shape[0] // BATCH_SIZE\ntrain_history_2 = model.fit(valid_dataset.repeat(),\n                            steps_per_epoch=n_steps,\n                            epochs=3)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"sub = pd.read_csv('../input/jigsaw-multilingual-toxic-comment-classification/sample_submission.csv')","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"sub['toxic'] = model.predict(test_dataset, verbose=1)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"sub.to_csv('submission.csv',index=False)","execution_count":null,"outputs":[]}],"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat":4,"nbformat_minor":4}