kernel_id
int64
24.2k
23.3M
prompt
stringlengths
8
1.85M
completetion
stringlengths
1
182k
comp_name
stringlengths
5
57
11,792,393
checkpoint = torch.load("/kaggle/input/deepfakes-inference-demo/resnext.pth", map_location=gpu) model = MyResNeXt().to(gpu) model.load_state_dict(checkpoint) _ = model.eval() del checkpoint<predict_on_test>
def flat_accuracy(preds, labels): pred_flat = np.argmax(preds, axis=1 ).flatten() labels_flat = labels.flatten() return np.sum(pred_flat == labels_flat)/ len(labels_flat )
Natural Language Processing with Disaster Tweets
11,792,393
def predict_on_video(video_path, batch_size): try: faces = face_extractor.process_video(video_path) face_extractor.keep_only_best_face(faces) if len(faces)> 0: x = np.zeros(( batch_size, input_size, input_size, 3), dtype=np.uint8) n = 0 for frame_data in faces: for face in frame_data["faces"]: resized_face = isotrop...
def format_time(elapsed): elapsed_rounded = int(round(( elapsed))) return str(datetime.timedelta(seconds=elapsed_rounded))
Natural Language Processing with Disaster Tweets
11,792,393
def predict_on_video_set(videos, num_workers): def process_file(i): filename = videos[i] y_pred = predict_on_video(os.path.join(test_dir, filename), batch_size=frames_per_video) return y_pred with ThreadPoolExecutor(max_workers=num_workers)as ex: predictions = ex.map(process_file, range(len(videos))) return list(pred...
seed_val = 42 random.seed(seed_val) np.random.seed(seed_val) torch.manual_seed(seed_val) torch.cuda.manual_seed_all(seed_val) loss_values = [] for epoch_i in range(0, epochs): print("") print('======== Epoch {:} / {:} ========'.format(epoch_i + 1, epochs)) print('Training...') t0 = time.time() total_loss = 0 mode...
Natural Language Processing with Disaster Tweets
11,792,393
predictions = predict_on_video_set(test_videos, num_workers=4 )<save_to_csv>
print("") print("Running Validation...") t0 = time.time() model.eval() preds=[] true=[] eval_loss, eval_accuracy = 0, 0 nb_eval_steps, nb_eval_examples = 0, 0 for batch in validation_dataloader: batch = tuple(t.to(device)for t in batch) b_input_ids, b_input_mask, b_labels = batch with torch.no_grad() : outputs = mod...
Natural Language Processing with Disaster Tweets
11,792,393
submission_df_resnext = pd.DataFrame({"filename": test_videos, "label": predictions}) submission_df_resnext.to_csv("submission_resnext.csv", index=False )<install_modules>
flat_predictions = [item for sublist in preds for item in sublist] flat_predictions = np.argmax(flat_predictions, axis=1 ).flatten() flat_true_labels = [item for sublist in true for item in sublist]
Natural Language Processing with Disaster Tweets
11,792,393
!pip install.. /input/deepfake-xception-trained-model/pytorchcv-0.0.55-py2.py3-none-any.whl --quiet<set_options>
print(classification_report(flat_predictions,flat_true_labels))
Natural Language Processing with Disaster Tweets
11,792,393
%matplotlib inline warnings.filterwarnings("ignore" )<define_variables>
comments1 = df_test.text.values indices1=tokenizer.batch_encode_plus(comments1,max_length=128,add_special_tokens=True, return_attention_mask=True,pad_to_max_length=True,truncation=True) input_ids1=indices1["input_ids"] attention_masks1=indices1["attention_mask"] prediction_inputs1= torch.tensor(input_ids1) prediction...
Natural Language Processing with Disaster Tweets
11,792,393
test_dir = "/kaggle/input/deepfake-detection-challenge/test_videos/" test_videos = sorted([x for x in os.listdir(test_dir)if x[-4:] == ".mp4"]) len(test_videos )<set_options>
print('Predicting labels for {:,} test sentences...'.format(len(prediction_inputs1))) model.eval() predictions = [] for batch in prediction_dataloader1: batch = tuple(t.to(device)for t in batch) b_input_ids1, b_input_mask1 = batch with torch.no_grad() : outputs1 = model(b_input_ids1, token_type_ids=None, attention_ma...
Natural Language Processing with Disaster Tweets
11,792,393
gpu = torch.device("cuda:0" if torch.cuda.is_available() else "cpu" )<load_pretrained>
sample_sub=pd.read_csv('.. /input/nlp-getting-started/sample_submission.csv') submit=pd.DataFrame({'id':sample_sub['id'].values.tolist() ,'target':flat_predictions})
Natural Language Processing with Disaster Tweets
11,792,393
<load_pretrained><EOS>
df_leak = pd.read_csv('/kaggle/input/disasters-on-social-media/socialmedia-disaster-tweets-DFE.csv', encoding ='ISO-8859-1')[['choose_one', 'text']] df_leak['target'] =(df_leak['choose_one'] == 'Relevant' ).astype(np.int8) df_leak['id'] = df_leak.index.astype(np.int16) df_leak.drop(columns=['choose_one', 'text'], inp...
Natural Language Processing with Disaster Tweets
11,449,263
<SOS> metric: meanfscore Kaggle data source: natural-language-processing-with-disaster-tweets<define_variables>
%matplotlib inline
Natural Language Processing with Disaster Tweets
11,449,263
input_size = 150<normalization>
train_df = pd.read_csv("/kaggle/input/nlp-getting-started/train.csv") test_df = pd.read_csv("/kaggle/input/nlp-getting-started/test.csv" )
Natural Language Processing with Disaster Tweets
11,449,263
mean = [0.485, 0.456, 0.406] std = [0.229, 0.224, 0.225] normalize_transform = Normalize(mean, std )<choose_model_class>
train_df['char'] = train_df['text'].str.len() train_df['words'] = train_df['text'].str.split().map(lambda x: len(x)) train_df.head(3 )
Natural Language Processing with Disaster Tweets
11,449,263
model = get_model("xception", pretrained=False) model = nn.Sequential(*list(model.children())[:-1]) class Pooling(nn.Module): def __init__(self): super(Pooling, self ).__init__() self.p1 = nn.AdaptiveAvgPool2d(( 1,1)) self.p2 = nn.AdaptiveMaxPool2d(( 1,1)) def forward(self, x): x1 = self.p1(x) x2 = self.p2(x) retur...
print('There are {} tweets in total and {} tweets with keyword'.format(train_df['keyword'].shape[0], train_df['keyword'].notna().sum()))
Natural Language Processing with Disaster Tweets
11,449,263
def predict_on_video(video_path, batch_size): try: faces = face_extractor.process_video(video_path) face_extractor.keep_only_best_face(faces) if len(faces)> 0: x = np.zeros(( batch_size, input_size, input_size, 3), dtype=np.uint8) n = 0 for frame_data in faces: for face in frame_data["faces"]: resized_face = isotrop...
train_df['text'] = train_df['text'].str.replace("http\S+", " ") test_df['text'] = test_df['text'].str.replace("http\S+", " " )
Natural Language Processing with Disaster Tweets
11,449,263
def predict_on_video_set(videos, num_workers): def process_file(i): filename = videos[i] y_pred = predict_on_video(os.path.join(test_dir, filename), batch_size=frames_per_video) return y_pred with ThreadPoolExecutor(max_workers=num_workers)as ex: predictions = ex.map(process_file, range(len(videos))) return list(pred...
train_df = train_df[['text', 'target']] tweets = train_df.groupby('text' ).mean().reset_index() print('There are {} tweets with different label in duplicates.'.format(tweets[(tweets['target']!=1)&(tweets['target']!=0)].shape[0])) df_diff = tweets[(tweets['target']!=1)&(tweets['target']!=0)].reset_index(drop=True) twee...
Natural Language Processing with Disaster Tweets
11,449,263
%%time model.eval() predictions = predict_on_video_set(test_videos, num_workers=4 )<save_to_csv>
for i in range(64): print(train_df[train_df['text']==df_diff.loc[i]['text']] )
Natural Language Processing with Disaster Tweets
11,449,263
submission_df_xception = pd.DataFrame({"filename": test_videos, "label": predictions}) submission_df_xception.to_csv("submission_xception.csv", index=False )<create_dataframe>
nlp = spacy.load('en_core_web_lg' )
Natural Language Processing with Disaster Tweets
11,449,263
submission_df = pd.DataFrame({"filename": test_videos} )<feature_engineering>
X_train, X_valid, y_train, y_valid = train_test_split(vectors, tweets.target, test_size=0.2, random_state=52, stratify = tweets.target) model = LinearSVC(random_state=1, dual=False) model.fit(X_train, y_train) print(f'Model test accuracy: {model.score(X_valid, y_valid)*100:.3f}%' )
Natural Language Processing with Disaster Tweets
11,449,263
submission_df["label"] = 0.70*submission_df_resnext["label"] + 0.30*submission_df_xception["label"]<save_to_csv>
second_model = LogisticRegression(solver='saga') second_model.fit(X_train, y_train) print(f'Model test accuracy: {second_model.score(X_valid, y_valid)*100:.3f}%' )
Natural Language Processing with Disaster Tweets
11,449,263
submission_df.to_csv("submission.csv", index=False )<define_variables>
third_model = linear_model.RidgeClassifier() third_model.fit(X_train, y_train) print(f'Model test accuracy: {third_model.score(X_valid, y_valid)*100:.3f}%' )
Natural Language Processing with Disaster Tweets
11,449,263
TEST_DIR = "/kaggle/input/deepfake-detection-challenge/test_videos/" CHECKPOINT = '/kaggle/input/kha-deepfake-dataset/checkpoint_mobilev3_alldata_1903_withfaceforensics_3epochs_.pth' CHECKPOINT2 = '/kaggle/input/kha-deepfake-dataset/cpt_mbn_sqrimg_2503)2epochs_.pth' CHECKPOINT3 = '/kaggle/input/kha-deepfake-dataset/che...
lgb_train = lgb.Dataset(X_train, y_train) lgb_eval = lgb.Dataset(X_valid, y_valid, reference=lgb_train) params = { 'task' : 'train', 'boosting_type' : 'gbdt', 'objective' : 'binary', 'metric' : {'binary_logloss'}, 'num_leaves' : 51, 'learning_rate' : 0.01, 'max_bin': 397, 'feature_fraction' : 0.9, 'bagging_fraction' ...
Natural Language Processing with Disaster Tweets
11,449,263
package_path = '.. /input/kha-efficientnet/EfficientNet-PyTorch/' sys.path.append(package_path) <install_modules>
print('SVM f1 score {}'.format(f1_score(y_valid, np.round(model.predict(X_valid), 0 ).astype(int)))) print('Logistic Regression f1 score {}'.format(f1_score(y_valid, np.round(second_model.predict(X_valid), 0 ).astype(int)))) print('Ridge Classifier f1 score {}'.format(f1_score(y_valid, np.round(third_model.predict(X_va...
Natural Language Processing with Disaster Tweets
11,449,263
%%capture !pip install /kaggle/input/khafacenet/facenet_pytorch-2.2.7-py3-none-any.whl !pip install /kaggle/input/imutils/imutils-0.5.3<install_modules>
X_train = pd.DataFrame(vectors) y_train = tweets['target'] X_test = pd.DataFrame(vectors_pred) y_preds = [] models = [] oof_train = np.zeros(( len(X_train),)) cv = KFold(n_splits=5, shuffle=True, random_state=100892) params = { 'task' : 'train', 'boosting_type' : 'gbdt', 'objective' : 'binary', 'metric' : {'binary_l...
Natural Language Processing with Disaster Tweets
11,449,263
<load_from_zip>
print('===CV scores===') print(scores) print(score )
Natural Language Processing with Disaster Tweets
11,449,263
<set_options>
pred = pd.DataFrame(np.round(y_preds, 0)).astype('int8' ).T
Natural Language Processing with Disaster Tweets
11,449,263
%matplotlib inline device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu") <categorify>
np.round(pred.mean(axis=1),0 ).astype(int )
Natural Language Processing with Disaster Tweets
11,449,263
def conv_bn(inp, oup, stride, conv_layer=nn.Conv2d, norm_layer=nn.BatchNorm2d, nlin_layer=nn.ReLU): return nn.Sequential( conv_layer(inp, oup, 3, stride, 1, bias=False), norm_layer(oup), nlin_layer(inplace=True) ) def conv_1x1_bn(inp, oup, conv_layer=nn.Conv2d, norm_layer=nn.BatchNorm2d, nlin_layer=nn.ReLU): return n...
Natural Language Processing with Disaster Tweets
11,449,263
net = mobilenetv3(mode='small', pretrained=False) net.classifier[1] = torch.nn.Linear(in_features=1280, out_features=1) net = net.to(device) state_dict = torch.load(CHECKPOINT) net.load_state_dict(state_dict) net.cuda() net.eval()<find_best_params>
sample_submission = pd.read_csv("/kaggle/input/nlp-getting-started/sample_submission.csv") sample_submission
Natural Language Processing with Disaster Tweets
11,449,263
net2 = mobilenetv3(mode='small', pretrained=False) net2.classifier[1] = torch.nn.Linear(in_features=1280, out_features=1) net2 = net2.to(device) state_dict = torch.load(CHECKPOINT2) net2.load_state_dict(state_dict) net2.cuda() net2.eval()<set_options>
sample_submission["target"] = np.round(pred.mean(axis=1),0 ).astype(int) sample_submission["target"] = sample_submission["target"].astype('int8' )
Natural Language Processing with Disaster Tweets
11,449,263
<set_options><EOS>
sample_submission.to_csv("submission.csv", index=False )
Natural Language Processing with Disaster Tweets
11,373,963
<SOS> metric: meanfscore Kaggle data source: natural-language-processing-with-disaster-tweets<define_search_model>
!pip install -q keras-bert keras-rectified-adam !wget -q https://storage.googleapis.com/bert_models/2018_10_18/uncased_L-12_H-768_A-12.zip !unzip -o uncased_L-12_H-768_A-12.zip
Natural Language Processing with Disaster Tweets
11,373,963
class SeparableConv2d(nn.Module): def __init__(self,in_channels,out_channels,kernel_size=1,stride=1,padding=0,dilation=1,bias=False): super(SeparableConv2d,self ).__init__() self.conv1 = nn.Conv2d(in_channels,in_channels,kernel_size,stride,padding,dilation,groups=in_channels,bias=bias) self.pointwise = nn.Conv2d(in_ch...
SEQ_LEN = 128 BATCH_SIZE = 1024 EPOCHS = 15 LR = 1e-4
Natural Language Processing with Disaster Tweets
11,373,963
class MaxPoolPad(nn.Module): def __init__(self): super(MaxPoolPad, self ).__init__() self.pad = nn.ZeroPad2d(( 1, 0, 1, 0)) self.pool = nn.MaxPool2d(3, stride=2, padding=1) def forward(self, x): x = self.pad(x) x = self.pool(x) x = x[:, :, 1:, 1:].contiguous() return x class AvgPoolPad(nn.Module): def __init__(self,...
pretrained_path = 'uncased_L-12_H-768_A-12' config_path = os.path.join(pretrained_path, 'bert_config.json') checkpoint_path = os.path.join(pretrained_path, 'bert_model.ckpt') vocab_path = os.path.join(pretrained_path, 'vocab.txt') os.environ['TF_KERAS'] = '1'
Natural Language Processing with Disaster Tweets
11,373,963
class CFG: seq_len=10 lstm_in = 16 lstm_out = 16 class LSTM_Model(nn.Module): def __init__(self): super(LSTM_Model, self ).__init__() self.cnn_net = mobilenetv3(mode='small', pretrained=False) self.cnn_net.classifier[1] = nn.Linear(in_features=1280, out_features=1) self.cnn_net.classifier[1] = nn.Linear(in_features=1...
tpu = tf.distribute.cluster_resolver.TPUClusterResolver() tf.config.experimental_connect_to_cluster(tpu) tf.tpu.experimental.initialize_tpu_system(tpu) tpu_strategy = tf.distribute.experimental.TPUStrategy(tpu )
Natural Language Processing with Disaster Tweets
11,373,963
<find_best_params>
token_dict = {} with codecs.open(vocab_path, 'r', 'utf8')as reader: for line in reader: token = line.strip() token_dict[token] = len(token_dict) with tpu_strategy.scope() : model = load_trained_model_from_checkpoint( config_path, checkpoint_path, training=True, trainable=True, seq_len=SEQ_LEN, )
Natural Language Processing with Disaster Tweets
11,373,963
net11 = torchvision.models.resnet18(pretrained=False) net11.fc = nn.Linear(in_features=512, out_features=1, bias=True) net11.load_state_dict(torch.load(CHECKPOINT11)) net11 = net11.to(device) net11.cuda() net11.eval()<import_modules>
train_df = pd.read_csv('.. /input/nlp-getting-started/train.csv', index_col='id') test_df = pd.read_csv('.. /input/nlp-getting-started/test.csv', index_col='id' )
Natural Language Processing with Disaster Tweets
11,373,963
import albumentations from albumentations.augmentations.transforms import ShiftScaleRotate, HorizontalFlip, RandomBrightnessContrast, MotionBlur, Blur, GaussNoise, JpegCompression <define_variables>
def strip_html(text): soup = BeautifulSoup(text, "html.parser") return soup.get_text() def remove_between_square_brackets(text): return re.sub('\[[^]]*\]', '', text) def remove_url(text): return re.sub(r'http\S+', '', text) def add_space(text): return re.sub('%20', ' ', text) def remove_hashtags(text): return re.su...
Natural Language Processing with Disaster Tweets
11,373,963
def predict_on_video(model, model2, model3, model4, model5, model6, model7, model8, model9, model10, model11, video_path): try: x, x_sqr, x_299_sqr, x_lstm, frame_skip = extract_frames(video_path) if x is None or x_sqr is None: return 0.5 else: with torch.no_grad() : y_pred = model(x.to(device)) y_pred = torch.sigmoid...
def preprocess_df(df): df = df.fillna("") df['text'] = df['keyword'] + " " + df['text'] del df['keyword'] df['location'] = df['location'].astype('category') df['location'] = df['location'].cat.codes df['text'] = df['text'].apply(denoise_text) return df train_df = preprocess_df(train_df) test_df = preprocess_df(test...
Natural Language Processing with Disaster Tweets
11,373,963
class FastMTCNN(object): def __init__(self, resize=1, *args, **kwargs): self.resize = resize self.mtcnn = MTCNN(*args, **kwargs) def __call__(self, frames): if self.resize != 1: frames = [f.resize([int(d * self.resize)for d in f.size])for f in frames] boxes, probs = self.mtcnn.detect(frames) boxes = [b.astype(int ).t...
np.random.seed(1) msk = np.random.rand(len(train_df)) < 0.8 train_df, dev_df = train_df[msk], train_df[~msk]
Natural Language Processing with Disaster Tweets
11,373,963
test_videos = sorted([x for x in os.listdir(TEST_DIR)if x[-4:] == ".mp4"]) len(test_videos )<predict_on_test>
tokenizer = Tokenizer(token_dict) def tokenize(df): X, y = [], [] for i, index in enumerate(tqdm(df.index.values)) : ids, segments = tokenizer.encode(df.text.values[i], max_len=SEQ_LEN) X.append(ids) try: label = df.target.values[i] y.append(label) except: y.append(0) items = list(zip(X, y)) np.random.shuffle(item...
Natural Language Processing with Disaster Tweets
11,373,963
def predict_on_video_set(model, model2, model3, model4, model5, model6, model7, model8, model9, model10, model11, videos, num_workers): def process_file(i): filename = videos[i] y_pred = predict_on_video(model, model2, model3, model4, model5, model6, model7, model8, model9, model10, model11, os.path.join(TEST_DIR, file...
X_test = [] for i, index in enumerate(tqdm(test_df.index.values)) : ids, segments = tokenizer.encode(test_df.text.values[i], max_len=SEQ_LEN) X_test.append(ids) X_test = [np.array(X_test), np.zeros_like(X_test)]
Natural Language Processing with Disaster Tweets
11,373,963
predictions = np.clip(predictions, 0.005, 0.995) submission_df = pd.DataFrame({"filename": test_videos, "label": predictions}) submission_df.to_csv("submission.csv", index=False )<import_modules>
with tpu_strategy.scope() : inputs = model.inputs[:2] dense = model.get_layer('NSP-Dense' ).output outputs = keras.layers.Dense(units=2, activation='softmax' )(dense) model = keras.models.Model(inputs, outputs) model.compile( RAdam(lr=LR), loss='sparse_categorical_crossentropy', metrics=['sparse_categorical_accuracy...
Natural Language Processing with Disaster Tweets
11,373,963
import random import re from copy import deepcopy from typing import Union, List, Tuple, Optional, Callable from collections import OrderedDict, defaultdict import math import cv2 import torch import torch.nn as nn from torch.utils.data import Dataset,DataLoader from torch.utils.data.sampler import SequentialSampler, R...
learning_rate_reduction = keras.callbacks.ReduceLROnPlateau(monitor='val_sparse_categorical_accuracy', patience=2, verbose=1,factor=0.5, min_lr=1e-5 )
Natural Language Processing with Disaster Tweets
11,373,963
TARGET_H, TARGET_W = 224, 224 FRAMES_PER_VIDEO = 30 TEST_VIDEOS_PATH = '.. /input/deepfake-detection-challenge/test_videos' NN_MODEL_PATHS = [ '.. /input/kdold-deepfake-effb2/fold0-effb2-000epoch.pt', '.. /input/kdold-deepfake-effb2/fold0-effb2-001epoch.pt', '.. /input/kdold-deepfake-effb2/fold0-effb2-002epoch.pt', '.....
hist = model.fit(X_train, y_train, validation_data=(X_dev, y_dev), epochs=EPOCHS, batch_size=BATCH_SIZE, callbacks=[learning_rate_reduction] )
Natural Language Processing with Disaster Tweets
11,373,963
SEED = 42 def seed_everything(seed): random.seed(seed) os.environ['PYTHONHASHSEED'] = str(seed) np.random.seed(seed) torch.manual_seed(seed) torch.cuda.manual_seed(seed) torch.backends.cudnn.deterministic = False torch.backends.cudnn.benchmark = True seed_everything(SEED )<choose_model_class>
print("Accuracy of the model on Training Data is - {} %".format(model.evaluate(X_train,y_train)[1]*100)) print("Accuracy of the model on Dev Data is - {} %".format(model.evaluate(X_dev,y_dev)[1]*100))
Natural Language Processing with Disaster Tweets
11,373,963
!pip install.. /input/pytorchefficientnet/EfficientNet-PyTorch-master > /dev/null def get_net() : net = EfficientNet.from_name('efficientnet-b2') net._fc = nn.Linear(in_features=net._fc.in_features, out_features=2, bias=True) return net<feature_engineering>
classes = model.predict(X_test)[:, 0]
Natural Language Processing with Disaster Tweets
11,373,963
class DatasetRetriever(Dataset): def __init__(self, df): self.video_paths = df['video_path'] self.filenames = df.index self.face_dr = FaceDetector(frames_per_video=FRAMES_PER_VIDEO) mean = [0.485, 0.456, 0.406] std = [0.229, 0.224, 0.225] self.normalize_transform = Normalize(mean, std) self.video_reader = VideoReader...
submission = pd.DataFrame( {'id': list(test_df.index.values), 'target': list(( classes < 0.5 ).astype(int)) , } ).set_index('id' )
Natural Language Processing with Disaster Tweets
11,373,963
class DeepFakePredictor: def __init__(self): self.models = [self.prepare_model(get_net() , path)for path in NN_MODEL_PATHS] self.models_count = len(self.models) def predict(self, dataset): result = [] with torch.no_grad() : for filename, video in dataset: video = video.to(self.device, dtype=torch.float32) try: label ...
submission.to_csv('submission.csv' )
Natural Language Processing with Disaster Tweets
12,182,973
deep_fake_predictor = DeepFakePredictor()<predict_on_test>
!pip install -q tf-models-official==2.3.0
Natural Language Processing with Disaster Tweets
12,182,973
def process_dfs(df, num_workers=2): def process_df(sub_df): dataset = DatasetRetriever(sub_df) result = deep_fake_predictor.predict(dataset) return result with ThreadPoolExecutor(max_workers=num_workers)as ex: results = ex.map(process_df, np.split(df, num_workers)) return results<save_to_csv>
train = pd.read_csv('.. /input/nlp-getting-started/train.csv') test=pd.read_csv('.. /input/nlp-getting-started/test.csv') submission = pd.read_csv(".. /input/nlp-getting-started/sample_submission.csv" )
Natural Language Processing with Disaster Tweets
12,182,973
result.to_csv('submission.csv' )<install_modules>
train['words'] = train['text'].str.split() train['word_len'] = train['words'].map(lambda x: len(x)) train['word_len'].max()
Natural Language Processing with Disaster Tweets
12,182,973
!pip install.. /input/pytorchcv/pytorchcv-0.0.55-py2.py3-none-any.whl --quiet<import_modules>
test['words'] = test['text'].str.split() test['word_len'] = test['words'].map(lambda x: len(x)) test['word_len'].max()
Natural Language Processing with Disaster Tweets
12,182,973
device = 'cuda' if torch.cuda.is_available() else 'cpu'<normalization>
ids_with_target_error = [328,443,513,2619,3640,3900,4342,5781,6552,6554,6570,6701,6702,6729,6861,7226] train.loc[train['id'].isin(ids_with_target_error),'target'] = 0 train[train['id'].isin(ids_with_target_error)]
Natural Language Processing with Disaster Tweets
12,182,973
def gem(x, p=3, eps=1e-6): return F.avg_pool2d(x.clamp(min=eps ).pow(p),(x.size(-2), x.size(-1)) ).pow(1./p) class GeM(nn.Module): def __init__(self, p=3, eps=1e-6): super(GeM,self ).__init__() self.p = Parameter(torch.ones(1)*p) self.eps = eps def forward(self, x): return gem(x, p=self.p, eps=self.eps) def __repr__...
df_all = pd.concat([train,test]) df_all.shape
Natural Language Processing with Disaster Tweets
12,182,973
<normalization>
df_all['text'] = df_all['text'].str.lower() df_all['text'].head(2 )
Natural Language Processing with Disaster Tweets
12,182,973
mean = [0.485, 0.456, 0.406] std = [0.229, 0.224, 0.225] normalize_transform = Normalize(mean, std )<set_options>
def remove_breaklines(text): return re.sub(' ','',text) df_all['text'] = df_all['text'].apply(lambda x: remove_breaklines(x))
Natural Language Processing with Disaster Tweets
12,182,973
detection_graph = tf.Graph() with detection_graph.as_default() : od_graph_def = tf.compat.v1.GraphDef() with tf.io.gfile.GFile('.. /input/mobilenet-face/frozen_inference_graph_face.pb', 'rb')as fid: serialized_graph = fid.read() od_graph_def.ParseFromString(serialized_graph) tf.import_graph_def(od_graph_def, name='') ...
def remove_numbers(text): return re.sub('\w*\d\w*', '', text) df_all['text'] = df_all['text'].apply(lambda x: remove_numbers(x))
Natural Language Processing with Disaster Tweets
12,182,973
probs = np.asarray(probs) probs[probs!=probs] = 0.5 plt.hist(probs, 40) filenames = [os.path.basename(f)for f in filenames] submission = pd.DataFrame({'filename': filenames, 'label': probs}) submission.to_csv('submission.csv', index=False) submission<import_modules>
def remove_URL(text): url = re.compile(r'https?://\S+|www\.\S+') return url.sub(r'',text) df_all['text'] = df_all['text'].apply(lambda x: remove_URL(x))
Natural Language Processing with Disaster Tweets
12,182,973
from fastai.vision import *<load_from_disk>
def remove_html(text): html=re.compile(r'<.*?>') return html.sub(r'',text) df_all['text']=df_all['text'].apply(lambda x : remove_html(x))
Natural Language Processing with Disaster Tweets
12,182,973
train_sample_metadata = pd.read_json('.. /input/deepfake-detection-challenge/train_sample_videos/metadata.json' ).T.reset_index() train_sample_metadata.columns = ['fname','label','split','original'] train_sample_metadata.head()<filter>
def remove_emoji(text): emoji_pattern = re.compile("[" u"\U0001F600-\U0001F64F" u"\U0001F300-\U0001F5FF" u"\U0001F680-\U0001F6FF" u"\U0001F1E0-\U0001F1FF" u"\U00002702-\U000027B0" u"\U000024C2-\U0001F251" "]+", flags=re.UNICODE) return emoji_pattern.sub(r'', text) df_all['text']=df_all['text'].apply(lambda x: remove_...
Natural Language Processing with Disaster Tweets
12,182,973
fake_sample_df = train_sample_metadata[train_sample_metadata.label == 'FAKE'] real_sample_df = train_sample_metadata[train_sample_metadata.label == 'REAL']<define_variables>
def remove_punct(text): table=str.maketrans('','',string.punctuation) return text.translate(table) df_all['text']=df_all['text'].apply(lambda x : remove_punct(x))
Natural Language Processing with Disaster Tweets
12,182,973
train_dir = Path('/kaggle/input/deepfake-detection-challenge/train_sample_videos/') test_dir = Path('/kaggle/input/deepfake-detection-challenge/test_videos/') train_video_files = get_files(train_dir, extensions=['.mp4']) test_video_files = get_files(test_dir, extensions=['.mp4'] )<define_variables>
abbreviations = { "$" : " dollar ", "€" : " euro ", "4ao" : "for adults only", "a.m" : "before midday", "a3" : "anytime anywhere anyplace", "aamof" : "as a matter of fact", "acct" : "account", "adih" : "another day in hell", "afaic" : "as far as i am concerned", "afaict" : "as far as i can tell", "afaik" : "as far as i...
Natural Language Processing with Disaster Tweets
12,182,973
dummy_video_file = train_video_files[0]<set_options>
def convert_abbrev_in_text(text): tokens = word_tokenize(text) tokens = [convert_abbrev(word)for word in tokens] text = ' '.join(tokens) return text df_all['text']=df_all['text'].apply(lambda x : convert_abbrev_in_text(x))
Natural Language Processing with Disaster Tweets
12,182,973
sys.path.insert(0,'/kaggle/working/reader/python') set_bridge('torch') device = torch.device("cuda" )<categorify>
df_all['text']=df_all['text'].apply(lambda x : word_tokenize(x))
Natural Language Processing with Disaster Tweets
12,182,973
retinaface_stats = tensor([123,117,104] ).to(device) def decord_cpu_video_reader(path, freq=None): video = VideoReader(str(path), ctx=cpu()) len_video = len(video) if freq: t = video.get_batch(range(0, len(video), freq)).permute(0,3,1,2) else: t = video.get_batch(range(len_video)) return t, len_video def get_decord...
stop = set(stopwords.words('english')) def remove_stopwords(text): words = [w for w in text if w not in stop] return ' '.join(words) df_all['text']=df_all['text'].apply(lambda x : remove_stopwords(x))
Natural Language Processing with Disaster Tweets
12,182,973
sys.path.insert(0,"/kaggle/input/retina-face-2/Pytorch_Retinaface_2/" )<import_modules>
cl_ch_len = df_all['text'].apply(lambda x: len(x)) cl_wd_len = df_all['text'].str.split().map(lambda x: len(x)) print('Max words length for cleaned tweets: {}'.format(max(cl_wd_len))) print('Max characters length for cleaned tweets: {}'.format(max(cl_ch_len))) MAX_LEN = max(cl_wd_len )
Natural Language Processing with Disaster Tweets
12,182,973
import os import torch import torch.backends.cudnn as cudnn import numpy as np from data import cfg_mnet, cfg_re50 from layers.functions.prior_box import PriorBox from utils.nms.py_cpu_nms import py_cpu_nms import cv2 from models.retinaface import RetinaFace from utils.box_utils import decode, decode_landm import time<...
def create_corpus(df): corpus=[] for tweet in tqdm(df_all['text']): words=[word.lower() for word in word_tokenize(tweet)if(( word.isalpha() ==1)&(word not in stop)) ] corpus.append(words) return corpus corpus = create_corpus(df_all) corpus[0]
Natural Language Processing with Disaster Tweets
12,182,973
def check_keys(model, pretrained_state_dict): ckpt_keys = set(pretrained_state_dict.keys()) model_keys = set(model.state_dict().keys()) used_pretrained_keys = model_keys & ckpt_keys unused_pretrained_keys = ckpt_keys - model_keys missing_keys = model_keys - ckpt_keys print('Missing keys:{}'.format(len(missing_keys)))...
glove_embedding_dict={} with open('.. /input/glove-global-vectors-for-word-representation/glove.6B.200d.txt','r')as f: for line in tqdm(f): values=line.split() word=values[0] vectors=np.asarray(values[1:],'float32') glove_embedding_dict[word]=vectors
Natural Language Processing with Disaster Tweets
12,182,973
cudnn.benchmark = True<define_variables>
W_E_DIM = 200
Natural Language Processing with Disaster Tweets
12,182,973
def get_model(modelname="mobilenet"): torch.set_grad_enabled(False) cfg = None cfg_mnet['pretrain'] = False cfg_re50['pretrain'] = False if modelname == "mobilenet": pretrained_path = ".. /input/retina-face-2/Pytorch_Retinaface_2/weights/mobilenet0.25_Final.pth" cfg = cfg_mnet if modelname == "resnet50": pretrained_pa...
tokenizer_obj=Tokenizer() tokenizer_obj.fit_on_texts(corpus) sequences=tokenizer_obj.texts_to_sequences(corpus) tweet_pad=pad_sequences(sequences,maxlen=MAX_LEN,truncating='post',padding='post' )
Natural Language Processing with Disaster Tweets
12,182,973
def predict(model, t, sz, cfg, confidence_threshold = 0.5, top_k = 5, nms_threshold = 0.5, keep_top_k = 5): "get prediction for a batch t by model with image sz" resize = 1 scale_rate = 1 im_height, im_width = sz, sz scale = torch.Tensor([sz, sz, sz, sz]) scale = scale.to(device) locs, confs, landmss = torch.Tensor([...
word_index=tokenizer_obj.word_index print('Number of unique words:',len(word_index))
Natural Language Processing with Disaster Tweets
12,182,973
%%time model, cfg = get_model("mobilenet" )<categorify>
num_words=len(word_index)+1 embedding_matrix=np.zeros(( num_words,W_E_DIM)) for word,i in tqdm(word_index.items()): if i < num_words: emb_vec=glove_embedding_dict.get(word) if emb_vec is not None: embedding_matrix[i]=emb_vec
Natural Language Processing with Disaster Tweets
12,182,973
def bboxes_to_original_scale(bboxes, H, W, sz): res = [] for bb in bboxes: h_scale, w_scale = H/sz, W/sz orig_bboxes =(bb*array([w_scale, h_scale, w_scale, h_scale])[None,...] ).astype(int) res.append(orig_bboxes) return res<categorify>
train_text = tweet_pad[:train.shape[0]] test_text = tweet_pad[train.shape[0]:] X_train,X_dev,Y_train,Y_dev=train_test_split(train_text,train['target'].values,test_size=0.2) print('Shape of train',X_train.shape) print("Shape of Validation ",X_dev.shape )
Natural Language Processing with Disaster Tweets
12,182,973
def landmarks_to_original_scale(landmarks, H, W, sz): res = [] for landms in landmarks: h_scale, w_scale = H/sz, W/sz orig_landms =(landms*array([w_scale, h_scale]*5)[None,...] ).astype(int) res.append(orig_landms) return res<import_modules>
model=Sequential() embedding=Embedding(num_words,W_E_DIM,embeddings_initializer=Constant(embedding_matrix), input_length=MAX_LEN,trainable=False) model.add(embedding) model.add(SpatialDropout1D(0.2)) model.add(LSTM(100, dropout=0.2, recurrent_dropout=0.2)) model.add(Dense(1, activation='sigmoid')) optimzer=Adam(learn...
Natural Language Processing with Disaster Tweets
12,182,973
from tqdm import tqdm<init_hyperparams>
history=model.fit(X_train,Y_train,batch_size=4, epochs=10,validation_data=(X_dev,Y_dev),verbose=2)
Natural Language Processing with Disaster Tweets
12,182,973
freq = 5 model_args = dict(confidence_threshold = 0.5, top_k = 5, nms_threshold = 0.5, keep_top_k = 5) sz = cfg['image_size'] imgnet_stats = [tensor(o)for o in imagenet_stats] rescale_param = 1.3<install_modules>
pred = model.predict(test_text) pred = pred.round().astype('int' )
Natural Language Processing with Disaster Tweets
12,182,973
!pip install -q.. /input/efficientnetpytorchpip/efficientnet_pytorch-0.6.3/<import_modules>
df_sub = pd.read_csv('/kaggle/input/nlp-getting-started/test.csv') df_sub['target'] = pred df_sub = df_sub[['id', 'target']] df_sub.to_csv('lstm_submission.csv', index=False, header=True) df_sub.head(10 )
Natural Language Processing with Disaster Tweets
12,182,973
from fastai.vision.models.efficientnet import *<categorify>
%%time bert_url = 'https://tfhub.dev/tensorflow/bert_en_uncased_L-24_H-1024_A-16/2' bert_layer = hub.KerasLayer(bert_url, trainable=True, name='Bert_layer' )
Natural Language Processing with Disaster Tweets
12,182,973
class DummyDatabunch: c = 2 path = '.' device = defaults.device loss_func = None data = DummyDatabunch()<load_pretrained>
def bert_encode(texts, tokenizer, max_len=512): input_tokens = [] input_masks = [] input_segments = [] for text in texts: text = tokenizer.tokenize(text) text = text[:max_len-2] input_sequence = ["[CLS]"] + text + ["[SEP]"] pad_len = max_len - len(input_sequence) tokens = tokenizer.convert_tokens_to_ids(input_sequenc...
Natural Language Processing with Disaster Tweets
12,182,973
effnet_model = EfficientNet.from_name("efficientnet-b5", override_params={'num_classes': 2}) learner = Learner(data, effnet_model); learner.model_dir = '.' learner.load('.. /input/deepfakerandmergeaugmodels/single_frame_effnetb5_randmerge') effnetb5_inference_model = learner.model.eval()<load_pretrained>
def build_model(bert_layer, max_len=512): input_word_ids = Input(shape=(max_len,), dtype=tf.int32, name="input_word_ids") input_mask = Input(shape=(max_len,), dtype=tf.int32, name="input_mask") segment_ids = Input(shape=(max_len,), dtype=tf.int32, name="segment_ids") pooled_output, sequence_output = bert_layer([inpu...
Natural Language Processing with Disaster Tweets
12,182,973
effnet_model = EfficientNet.from_name("efficientnet-b7", override_params={'num_classes': 2}) learner = Learner(data, effnet_model); learner.model_dir = '.' learner.load('.. /input/deepfakerandmergeaugmodels/single_frame_effnetb7_randmerge_fp16') effnetb7_inference_model = learner.model.float().eval()<load_pretrained>
vocab_file = bert_layer.resolved_object.vocab_file.asset_path.numpy() do_lower_case = bert_layer.resolved_object.do_lower_case.numpy() tokenizer = tokenization.FullTokenizer(vocab_file, do_lower_case )
Natural Language Processing with Disaster Tweets
12,182,973
learner = cnn_learner(data, models.resnet34, pretrained=False); learner.model_dir = '.' learner.load('.. /input/deepfakerandmergeaugmodels/single_frame_resnet34_randmerge') resnet_inference_model = learner.model.eval()<define_variables>
train_text = df_all[:train.shape[0]].text test_text = df_all[train.shape[0]:].text train_input = bert_encode(train_text, tokenizer, max_len=160) test_input = bert_encode(test_text, tokenizer, max_len = 160) train_labels = train.target.values
Natural Language Processing with Disaster Tweets
12,182,973
predictions = [] video_fnames = []<define_variables>
checkpoint = ModelCheckpoint('model.h5', monitor='val_loss', save_best_only=True) train_history = model.fit( train_input, train_labels, validation_split=0.2, epochs=5, callbacks=[checkpoint], batch_size=16 )
Natural Language Processing with Disaster Tweets
12,182,973
fname2pred = dict(zip(video_fnames, predictions))<load_from_csv>
model.load_weights('model.h5') test_pred = model.predict(test_input )
Natural Language Processing with Disaster Tweets
12,182,973
submission_df = pd.read_csv("/kaggle/input/deepfake-detection-challenge/sample_submission.csv" )<categorify>
submission['target'] = test_pred.round().astype(int) submission.to_csv('bert_submission.csv', index=False )
Natural Language Processing with Disaster Tweets
12,165,368
submission_df.label = submission_df.filename.map(fname2pred )<feature_engineering>
test_set = pd.read_csv("/kaggle/input/nlp-getting-started/test.csv") train_set = pd.read_csv("/kaggle/input/nlp-getting-started/train.csv" )
Natural Language Processing with Disaster Tweets
12,165,368
submission_df['label'] = np.clip(submission_df['label'], 0.01, 0.99 )<save_to_csv>
import seaborn as sns from matplotlib import pyplot as plt
Natural Language Processing with Disaster Tweets
12,165,368
submission_df.to_csv("submission.csv",index=False )<install_modules>
print("train set") print(train_set.count()) print("test set") print(test_set.count() )
Natural Language Processing with Disaster Tweets
12,165,368
!pip install fastai==0.7.0 --no-deps !pip install torch==0.4.1 torchvision==0.2.1 !pip install torchtext==0.2.3<define_variables>
!pip install emoji
Natural Language Processing with Disaster Tweets
12,165,368
MODEL_NAME = 'Densenet201' TRAIN = '.. /input/train/' TEST = '.. /input/test/' LABELS = '.. /input/train.csv' SAMPLE_SUB = '.. /input/sample_submission.csv' arch = resnet50 num_workers = 8<feature_engineering>
def strip_emoji(text: str)-> str: return re.sub(emoji.get_emoji_regexp() , r"", text )
Natural Language Processing with Disaster Tweets
12,165,368
df = pd.read_csv(LABELS ).set_index('Image') new_whale_df = df[df.Id == "new_whale"] train_df = df[~(df.Id == "new_whale")] unique_labels = np.unique(train_df.Id.values) labels_dict = dict() labels_list = [] for i in range(len(unique_labels)) : labels_dict[unique_labels[i]] = i labels_list.append(unique_labels[i]) p...
train_set["text"] = train_set.text.str.strip().str.replace(" ", "") train_set["text"] = train_set.text.str.strip().str.replace("\r", "") train_set["text"] = train_set.text.apply(strip_emoji) train_set["text"] = train_set.text.str.strip().str.replace(" train_set["text"] = train_set.text.str.lower()
Natural Language Processing with Disaster Tweets
12,165,368
train_df['image_name'] = train_df.index rs = np.random.RandomState(42) perm = rs.permutation(len(train_df)) tr_n = train_df['image_name'].values val_n = train_df['image_name'].values[perm][:1000] print('Train/val:', len(tr_n), len(val_n)) print('Train classes', len(train_df.loc[tr_n].Id.unique())) print('Val classes',...
test_set["text"] = test_set.text.str.strip().str.replace(" ", "") test_set["text"] = test_set.text.str.strip().str.replace("\r", "") test_set["text"] = test_set.text.apply(strip_emoji) test_set["text"] = test_set.text.str.strip().str.replace(" test_set["text"] = test_set.text.str.lower()
Natural Language Processing with Disaster Tweets
12,165,368
class HWIDataset(FilesDataset): def __init__(self, fnames, path, transform): self.train_df = train_df super().__init__(fnames, transform, path) def get_x(self, i): img = open_image(os.path.join(self.path, self.fnames[i])) img = cv2.resize(img,(self.sz, self.sz)) return img def get_y(self, i): if(self.path == TEST): re...
!pip install torch>=1.6.0 transformers==3.3.1
Natural Language Processing with Disaster Tweets
12,165,368
class RandomLighting(Transform): def __init__(self, b, c, tfm_y=TfmType.NO): super().__init__(tfm_y) self.b, self.c = b, c def set_state(self): self.store.b_rand = rand0(self.b) self.store.c_rand = rand0(self.c) def do_transform(self, x, is_y): if is_y and self.tfm_y != TfmType.PIXEL: return x b = self.store.b_rand ...
import random from typing import Sequence from transformers import BertForMaskedLM, BertTokenizer import torch
Natural Language Processing with Disaster Tweets
12,165,368
image_size = 224 batch_size = 48 md = get_data(image_size, batch_size) extra_fc_layers_size = [] learn = ConvLearner.pretrained(arch, md, xtra_fc=extra_fc_layers_size) learn.opt_fn = optim.Adam<init_hyperparams>
tokenizer = BertTokenizer.from_pretrained("bert-base-uncased") text = train_set.text[0] tokenized_text = tokenizer.tokenize(text) print(tokenized_text )
Natural Language Processing with Disaster Tweets
12,165,368
print('Number of layer groups:', len(learn.get_layer_groups()), '\t(first 2 groups is pretrained backbone)') print('This is our extra thin on top of the backbone Resnet50 architecture:') learn.get_layer_groups() [2]<train_model>
index = 4 tokenized_text.insert(index, "[MASK]") print("Text Tokens") print(tokenized_text) tokenized_text_ids = torch.LongTensor(tokenizer.encode(tokenized_text, max_length=512, truncation=True)) print("Text Tokens ids") print(tokenized_text_ids )
Natural Language Processing with Disaster Tweets
12,165,368
base_lr = 5e-4 fc_lr = 5e-3 lrs = [base_lr, base_lr, fc_lr] learn.fit(lrs=lrs, n_cycle=2, cycle_len=None) learn.unfreeze() learn.fit(lrs, n_cycle=3, cycle_len=1, cycle_mult=2) learn.save('weights' )<define_variables>
model = BertForMaskedLM.from_pretrained("bert-base-uncased") model.eval() with torch.no_grad() : predictions = model(tokenized_text_ids.unsqueeze(0)) [0] predicted_word = tokenizer.convert_ids_to_tokens([torch.argmax(predictions[0, index+1] ).item() ])[0] print(predicted_word )
Natural Language Processing with Disaster Tweets
12,165,368
image_size = 448 batch_size = 24 md = get_data(image_size, batch_size) learn.set_data(md )<train_model>
tokenized_text[index] = predicted_word augmented_text = " ".join(tokenized_text) print(augmented_text )
Natural Language Processing with Disaster Tweets
12,165,368
base_lr = 1e-5 fc_lr = 1e-3 lrs = [base_lr, base_lr, fc_lr] learn.fit(lrs, n_cycle=6, cycle_len=1) learn.save('weights_v2' )<prepare_output>
class WordAugmentation: _mask_token: str = "[MASK]" def __init__(self, model_path: str="bert-base-uncased", device: str="cuda"): self._tokenizer = BertTokenizer.from_pretrained(model_path) self._device = device self._model = BertForMaskedLM.from_pretrained(model_path ).to(self._device) def _predict(self, inputs_i...
Natural Language Processing with Disaster Tweets