text-generator-using-lstm .ipynb

{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.10.13","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"gpu","dataSources":[{"sourceId":31436,"sourceType":"datasetVersion","datasetId":19447}],"dockerImageVersionId":30699,"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":true}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"# **PROJECT: TEXT GENERATION USING LSTM**","metadata":{}},{"cell_type":"markdown","source":"## 0. Setup","metadata":{}},{"cell_type":"code","source":"!pip install --upgrade tensorflow\n!pip install --upgrade streamlit\n","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import tensorflow as tf\nimport streamlit as st\n\nprint()\nprint(\"TensorFlow version:\", tf.__version__)\nprint(\"Streamlit version:\", st.__version__)\n\n","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## 1. Import All Dependencies","metadata":{}},{"cell_type":"code","source":"# module for building LSTM\nimport tensorflow as tf\nfrom tensorflow.keras.preprocessing.sequence import pad_sequences\nfrom tensorflow.keras.layers import Embedding, LSTM, Dense, Dropout\nfrom tensorflow.keras.preprocessing.text import Tokenizer\nfrom tensorflow.keras.models import Sequential\nfrom tensorflow.keras.callbacks import EarlyStopping\nfrom tensorflow.keras.preprocessing.text import tokenizer_from_json\nimport tensorflow.keras.utils as ku\n\n# set the seed (initial starting point for PRNG(pseudorandom number generator) to maintaing the consistency during generating the random numbers)\nfrom numpy.random import seed\ntf.random.set_seed(2)\nseed(1)\n\n# computaional module\nimport pandas as pd\nimport numpy as np\nimport string\nimport os\n\nimport json\n\n# to ignore the warnings\nimport warnings\nwarnings.filterwarnings('ignore')\nwarnings.simplefilter(action = 'ignore', category = FutureWarning)","metadata":{"execution":{"iopub.status.busy":"2024-04-21T16:27:26.787351Z","iopub.execute_input":"2024-04-21T16:27:26.787731Z","iopub.status.idle":"2024-04-21T16:27:30.427747Z","shell.execute_reply.started":"2024-04-21T16:27:26.787699Z","shell.execute_reply":"2024-04-21T16:27:30.426836Z"},"collapsed":true,"jupyter":{"outputs_hidden":true},"trusted":true},"execution_count":1,"outputs":[{"name":"stderr","text":"2024-04-21 16:27:27.150007: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:9261] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered\n2024-04-21 16:27:27.150061: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:607] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered\n2024-04-21 16:27:27.151591: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1515] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered\n","output_type":"stream"}]},{"cell_type":"code","source":"# specifying the directory path\nBASE_DIR = '/kaggle/input/nyt-comments/'\nWORKING_DIR = '/kaggle/working/'\n","metadata":{"execution":{"iopub.status.busy":"2024-04-21T16:27:30.429398Z","iopub.execute_input":"2024-04-21T16:27:30.430020Z","iopub.status.idle":"2024-04-21T16:27:30.434655Z","shell.execute_reply.started":"2024-04-21T16:27:30.429979Z","shell.execute_reply":"2024-04-21T16:27:30.433608Z"},"trusted":true},"execution_count":2,"outputs":[]},{"cell_type":"markdown","source":"## 2. Load the dataset\n#### we are using the NewYork Times Comments and Headlines","metadata":{}},{"cell_type":"code","source":"os.listdir(BASE_DIR)","metadata":{"execution":{"iopub.status.busy":"2024-04-21T16:27:30.435875Z","iopub.execute_input":"2024-04-21T16:27:30.437227Z","iopub.status.idle":"2024-04-21T16:27:30.463630Z","shell.execute_reply.started":"2024-04-21T16:27:30.437189Z","shell.execute_reply":"2024-04-21T16:27:30.462754Z"},"collapsed":true,"jupyter":{"outputs_hidden":true},"trusted":true},"execution_count":3,"outputs":[{"execution_count":3,"output_type":"execute_result","data":{"text/plain":"['CommentsFeb2018.csv',\n 'ArticlesFeb2017.csv',\n 'CommentsApril2018.csv',\n 'ArticlesJan2017.csv',\n 'ArticlesMay2017.csv',\n 'CommentsJan2017.csv',\n 'CommentsMarch2017.csv',\n 'CommentsMay2017.csv',\n 'CommentsMarch2018.csv',\n 'CommentsApril2017.csv',\n 'ArticlesMarch2017.csv',\n 'ArticlesApril2017.csv',\n 'CommentsFeb2017.csv',\n 'ArticlesJan2018.csv',\n 'ArticlesFeb2018.csv',\n 'ArticlesMarch2018.csv',\n 'CommentsJan2018.csv',\n 'ArticlesApril2018.csv']"},"metadata":{}}]},{"cell_type":"code","source":"# gather all the news headlines and stored it\nall_Headlines = []\n\nfor filename in os.listdir(BASE_DIR):\n    if 'Articles' in filename:\n        article_df = pd.read_csv(BASE_DIR + filename)\n        all_Headlines.extend(list(article_df.headline.values))\n        break\n        \n# removed the articles without headlines\nall_Headlines = [headline for headline in all_Headlines if headline != 'Unknown']\nlen(all_Headlines)","metadata":{"execution":{"iopub.status.busy":"2024-04-21T16:27:30.464545Z","iopub.execute_input":"2024-04-21T16:27:30.464778Z","iopub.status.idle":"2024-04-21T16:27:30.488841Z","shell.execute_reply.started":"2024-04-21T16:27:30.464757Z","shell.execute_reply":"2024-04-21T16:27:30.487988Z"},"trusted":true},"execution_count":4,"outputs":[{"execution_count":4,"output_type":"execute_result","data":{"text/plain":"829"},"metadata":{}}]},{"cell_type":"markdown","source":"## 3. Data Preparation","metadata":{}},{"cell_type":"markdown","source":"#### 3.1 Data Cleaning :\n##### Remove the punctuations and lower casting all the words","metadata":{}},{"cell_type":"code","source":"# define the function for data cleaning\ndef clean_data(text):\n    text = \"\".join(letter for letter in text if letter not in string.punctuation).lower()\n    text = text.encode(\"utf-8\").decode('ascii', 'ignore')\n    return text\n    \n# create corpus\ncorpus = [clean_data(headline) for headline in all_Headlines]\ncorpus[:10]\n    ","metadata":{"execution":{"iopub.status.busy":"2024-04-21T16:27:30.491955Z","iopub.execute_input":"2024-04-21T16:27:30.492226Z","iopub.status.idle":"2024-04-21T16:27:30.506494Z","shell.execute_reply.started":"2024-04-21T16:27:30.492202Z","shell.execute_reply":"2024-04-21T16:27:30.505578Z"},"collapsed":true,"jupyter":{"outputs_hidden":true},"trusted":true},"execution_count":5,"outputs":[{"execution_count":5,"output_type":"execute_result","data":{"text/plain":"['nfl vs politics has been battle all season long',\n 'voice vice veracity',\n 'a standups downward slide',\n 'new york today a groundhog has her day',\n 'a swimmers communion with the ocean',\n 'trail activity',\n 'super bowl',\n 'trumps mexican shakedown',\n 'pences presidential pet',\n 'fruit of a poison tree']"},"metadata":{}}]},{"cell_type":"markdown","source":"#### 3.2  Tokenization: \n##### Tokenized the sequences into tokens","metadata":{}},{"cell_type":"code","source":"# create instance of tokenizer \ntokenizer = Tokenizer()\n\n# define the funtion to perform the tokenization on our data\ndef tokenized_sequence(corpus):\n    tokenizer.fit_on_texts(corpus)\n    vocab_size = len(tokenizer.word_index) + 1\n    \n    # convert data to sequence of token ( N-Grams sequences)\n    input_sequences = []\n    for line in corpus:\n        list_of_token = tokenizer.texts_to_sequences([line])[0]\n        for i in range(1, len(list_of_token)):\n            n_gram_seq = list_of_token[:i+1]\n            input_sequences.append(n_gram_seq)\n    \n    return input_sequences, vocab_size\n    \n\n# implement this function on our data\ninput_sequences, vocab_size = tokenized_sequence(corpus)\ninput_sequences[:10]","metadata":{"execution":{"iopub.status.busy":"2024-04-21T16:27:30.507534Z","iopub.execute_input":"2024-04-21T16:27:30.508027Z","iopub.status.idle":"2024-04-21T16:27:30.549390Z","shell.execute_reply.started":"2024-04-21T16:27:30.508002Z","shell.execute_reply":"2024-04-21T16:27:30.548549Z"},"collapsed":true,"jupyter":{"outputs_hidden":true},"trusted":true},"execution_count":6,"outputs":[{"execution_count":6,"output_type":"execute_result","data":{"text/plain":"[[660, 117],\n [660, 117, 72],\n [660, 117, 72, 73],\n [660, 117, 72, 73, 661],\n [660, 117, 72, 73, 661, 662],\n [660, 117, 72, 73, 661, 662, 63],\n [660, 117, 72, 73, 661, 662, 63, 29],\n [660, 117, 72, 73, 661, 662, 63, 29, 210],\n [211, 663],\n [211, 663, 664]]"},"metadata":{}}]},{"cell_type":"markdown","source":"### Save the tokenizer","metadata":{}},{"cell_type":"code","source":"# Save the tokenizer to a JSON file\ntokenizer_json = tokenizer.to_json()\nwith open(\"tokenizer.json\", \"w\", encoding=\"utf-8\") as f:\n    f.write(json.dumps(tokenizer_json))","metadata":{"execution":{"iopub.status.busy":"2024-04-21T16:27:30.550410Z","iopub.execute_input":"2024-04-21T16:27:30.550691Z","iopub.status.idle":"2024-04-21T16:27:30.562974Z","shell.execute_reply.started":"2024-04-21T16:27:30.550668Z","shell.execute_reply":"2024-04-21T16:27:30.562123Z"},"trusted":true},"execution_count":7,"outputs":[]},{"cell_type":"markdown","source":"#### 3.3 Padding the Sequences and split the data into training and target data\n##### padding sequences is crucial as it ensure that all the vectors are of same size\n###### padding = 'pre' --> add zeros at begining of each sequence","metadata":{}},{"cell_type":"code","source":"# define the function to padd the sequences on our data\ndef padded_sequences(input_sequences):\n    max_seq_len = max([len(x) for x in input_sequences])\n    input_seq = np.array(pad_sequences(input_sequences, maxlen = max_seq_len, padding = 'pre'))\n    \n    # split sequence into the input data and label data\n    input_data, label_data = input_seq[:,:-1], input_seq[:,-1]\n    label_data = ku.to_categorical(label_data, num_classes = vocab_size)\n    \n    return input_data, label_data, max_seq_len\n    \n#  padded the sequences\ninput_data, label_data, max_seq_len = padded_sequences(input_sequences)","metadata":{"execution":{"iopub.status.busy":"2024-04-21T16:27:30.564063Z","iopub.execute_input":"2024-04-21T16:27:30.564561Z","iopub.status.idle":"2024-04-21T16:27:30.600363Z","shell.execute_reply.started":"2024-04-21T16:27:30.564535Z","shell.execute_reply":"2024-04-21T16:27:30.599670Z"},"trusted":true},"execution_count":8,"outputs":[]},{"cell_type":"markdown","source":"## 4. Text Generation using LSTM","metadata":{}},{"cell_type":"markdown","source":"#### 4.1 Build Model\n##### Model Architecture: Three major layer (Embedding Layer, LSTM, Dense) added into the model.\n","metadata":{}},{"cell_type":"code","source":"def build_model(max_seq_len, vocab_size):\n    input_len = max_seq_len - 1\n    model = Sequential()\n    \n    # Add Input layer <-- Embedding Layer\n    model.add(Embedding(vocab_size, 10))\n    \n    # Add hidden layer <-- LSTM and Dropout layers\n    model.add(LSTM(256))\n    model.add(Dropout(0.5))\n    \n    # Add output layer\n    model.add(Dense(vocab_size, activation = 'softmax'))\n    \n    # compile model\n    model.compile(loss = 'categorical_crossentropy', optimizer = 'adam')\n    \n    return model\n","metadata":{"execution":{"iopub.status.busy":"2024-04-21T16:27:30.601382Z","iopub.execute_input":"2024-04-21T16:27:30.601675Z","iopub.status.idle":"2024-04-21T16:27:30.609655Z","shell.execute_reply.started":"2024-04-21T16:27:30.601650Z","shell.execute_reply":"2024-04-21T16:27:30.608672Z"},"trusted":true},"execution_count":9,"outputs":[]},{"cell_type":"code","source":"model = build_model(max_seq_len, vocab_size)\n\n# fit the model\nhistory = model.fit(input_data, label_data, epochs = 500, verbose = 5)","metadata":{"execution":{"iopub.status.busy":"2024-04-21T16:27:30.611077Z","iopub.execute_input":"2024-04-21T16:27:30.611655Z","iopub.status.idle":"2024-04-21T16:32:15.051835Z","shell.execute_reply.started":"2024-04-21T16:27:30.611620Z","shell.execute_reply":"2024-04-21T16:32:15.050955Z"},"collapsed":true,"jupyter":{"outputs_hidden":true},"trusted":true},"execution_count":10,"outputs":[{"name":"stdout","text":"Epoch 1/500\nEpoch 2/500\nEpoch 3/500\nEpoch 4/500\nEpoch 5/500\nEpoch 6/500\nEpoch 7/500\nEpoch 8/500\nEpoch 9/500\nEpoch 10/500\nEpoch 11/500\nEpoch 12/500\nEpoch 13/500\nEpoch 14/500\nEpoch 15/500\nEpoch 16/500\nEpoch 17/500\nEpoch 18/500\nEpoch 19/500\nEpoch 20/500\nEpoch 21/500\nEpoch 22/500\nEpoch 23/500\nEpoch 24/500\nEpoch 25/500\nEpoch 26/500\nEpoch 27/500\nEpoch 28/500\nEpoch 29/500\nEpoch 30/500\nEpoch 31/500\nEpoch 32/500\nEpoch 33/500\nEpoch 34/500\nEpoch 35/500\nEpoch 36/500\nEpoch 37/500\nEpoch 38/500\nEpoch 39/500\nEpoch 40/500\nEpoch 41/500\nEpoch 42/500\nEpoch 43/500\nEpoch 44/500\nEpoch 45/500\nEpoch 46/500\nEpoch 47/500\nEpoch 48/500\nEpoch 49/500\nEpoch 50/500\nEpoch 51/500\nEpoch 52/500\nEpoch 53/500\nEpoch 54/500\nEpoch 55/500\nEpoch 56/500\nEpoch 57/500\nEpoch 58/500\nEpoch 59/500\nEpoch 60/500\nEpoch 61/500\nEpoch 62/500\nEpoch 63/500\nEpoch 64/500\nEpoch 65/500\nEpoch 66/500\nEpoch 67/500\nEpoch 68/500\nEpoch 69/500\nEpoch 70/500\nEpoch 71/500\nEpoch 72/500\nEpoch 73/500\nEpoch 74/500\nEpoch 75/500\nEpoch 76/500\nEpoch 77/500\nEpoch 78/500\nEpoch 79/500\nEpoch 80/500\nEpoch 81/500\nEpoch 82/500\nEpoch 83/500\nEpoch 84/500\nEpoch 85/500\nEpoch 86/500\nEpoch 87/500\nEpoch 88/500\nEpoch 89/500\nEpoch 90/500\nEpoch 91/500\nEpoch 92/500\nEpoch 93/500\nEpoch 94/500\nEpoch 95/500\nEpoch 96/500\nEpoch 97/500\nEpoch 98/500\nEpoch 99/500\nEpoch 100/500\nEpoch 101/500\nEpoch 102/500\nEpoch 103/500\nEpoch 104/500\nEpoch 105/500\nEpoch 106/500\nEpoch 107/500\nEpoch 108/500\nEpoch 109/500\nEpoch 110/500\nEpoch 111/500\nEpoch 112/500\nEpoch 113/500\nEpoch 114/500\nEpoch 115/500\nEpoch 116/500\nEpoch 117/500\nEpoch 118/500\nEpoch 119/500\nEpoch 120/500\nEpoch 121/500\nEpoch 122/500\nEpoch 123/500\nEpoch 124/500\nEpoch 125/500\nEpoch 126/500\nEpoch 127/500\nEpoch 128/500\nEpoch 129/500\nEpoch 130/500\nEpoch 131/500\nEpoch 132/500\nEpoch 133/500\nEpoch 134/500\nEpoch 135/500\nEpoch 136/500\nEpoch 137/500\nEpoch 138/500\nEpoch 139/500\nEpoch 140/500\nEpoch 141/500\nEpoch 142/500\nEpoch 143/500\nEpoch 144/500\nEpoch 145/500\nEpoch 146/500\nEpoch 147/500\nEpoch 148/500\nEpoch 149/500\nEpoch 150/500\nEpoch 151/500\nEpoch 152/500\nEpoch 153/500\nEpoch 154/500\nEpoch 155/500\nEpoch 156/500\nEpoch 157/500\nEpoch 158/500\nEpoch 159/500\nEpoch 160/500\nEpoch 161/500\nEpoch 162/500\nEpoch 163/500\nEpoch 164/500\nEpoch 165/500\nEpoch 166/500\nEpoch 167/500\nEpoch 168/500\nEpoch 169/500\nEpoch 170/500\nEpoch 171/500\nEpoch 172/500\nEpoch 173/500\nEpoch 174/500\nEpoch 175/500\nEpoch 176/500\nEpoch 177/500\nEpoch 178/500\nEpoch 179/500\nEpoch 180/500\nEpoch 181/500\nEpoch 182/500\nEpoch 183/500\nEpoch 184/500\nEpoch 185/500\nEpoch 186/500\nEpoch 187/500\nEpoch 188/500\nEpoch 189/500\nEpoch 190/500\nEpoch 191/500\nEpoch 192/500\nEpoch 193/500\nEpoch 194/500\nEpoch 195/500\nEpoch 196/500\nEpoch 197/500\nEpoch 198/500\nEpoch 199/500\nEpoch 200/500\nEpoch 201/500\nEpoch 202/500\nEpoch 203/500\nEpoch 204/500\nEpoch 205/500\nEpoch 206/500\nEpoch 207/500\nEpoch 208/500\nEpoch 209/500\nEpoch 210/500\nEpoch 211/500\nEpoch 212/500\nEpoch 213/500\nEpoch 214/500\nEpoch 215/500\nEpoch 216/500\nEpoch 217/500\nEpoch 218/500\nEpoch 219/500\nEpoch 220/500\nEpoch 221/500\nEpoch 222/500\nEpoch 223/500\nEpoch 224/500\nEpoch 225/500\nEpoch 226/500\nEpoch 227/500\nEpoch 228/500\nEpoch 229/500\nEpoch 230/500\nEpoch 231/500\nEpoch 232/500\nEpoch 233/500\nEpoch 234/500\nEpoch 235/500\nEpoch 236/500\nEpoch 237/500\nEpoch 238/500\nEpoch 239/500\nEpoch 240/500\nEpoch 241/500\nEpoch 242/500\nEpoch 243/500\nEpoch 244/500\nEpoch 245/500\nEpoch 246/500\nEpoch 247/500\nEpoch 248/500\nEpoch 249/500\nEpoch 250/500\nEpoch 251/500\nEpoch 252/500\nEpoch 253/500\nEpoch 254/500\nEpoch 255/500\nEpoch 256/500\nEpoch 257/500\nEpoch 258/500\nEpoch 259/500\nEpoch 260/500\nEpoch 261/500\nEpoch 262/500\nEpoch 263/500\nEpoch 264/500\nEpoch 265/500\nEpoch 266/500\nEpoch 267/500\nEpoch 268/500\nEpoch 269/500\nEpoch 270/500\nEpoch 271/500\nEpoch 272/500\nEpoch 273/500\nEpoch 274/500\nEpoch 275/500\nEpoch 276/500\nEpoch 277/500\nEpoch 278/500\nEpoch 279/500\nEpoch 280/500\nEpoch 281/500\nEpoch 282/500\nEpoch 283/500\nEpoch 284/500\nEpoch 285/500\nEpoch 286/500\nEpoch 287/500\nEpoch 288/500\nEpoch 289/500\nEpoch 290/500\nEpoch 291/500\nEpoch 292/500\nEpoch 293/500\nEpoch 294/500\nEpoch 295/500\nEpoch 296/500\nEpoch 297/500\nEpoch 298/500\nEpoch 299/500\nEpoch 300/500\nEpoch 301/500\nEpoch 302/500\nEpoch 303/500\nEpoch 304/500\nEpoch 305/500\nEpoch 306/500\nEpoch 307/500\nEpoch 308/500\nEpoch 309/500\nEpoch 310/500\nEpoch 311/500\nEpoch 312/500\nEpoch 313/500\nEpoch 314/500\nEpoch 315/500\nEpoch 316/500\nEpoch 317/500\nEpoch 318/500\nEpoch 319/500\nEpoch 320/500\nEpoch 321/500\nEpoch 322/500\nEpoch 323/500\nEpoch 324/500\nEpoch 325/500\nEpoch 326/500\nEpoch 327/500\nEpoch 328/500\nEpoch 329/500\nEpoch 330/500\nEpoch 331/500\nEpoch 332/500\nEpoch 333/500\nEpoch 334/500\nEpoch 335/500\nEpoch 336/500\nEpoch 337/500\nEpoch 338/500\nEpoch 339/500\nEpoch 340/500\nEpoch 341/500\nEpoch 342/500\nEpoch 343/500\nEpoch 344/500\nEpoch 345/500\nEpoch 346/500\nEpoch 347/500\nEpoch 348/500\nEpoch 349/500\nEpoch 350/500\nEpoch 351/500\nEpoch 352/500\nEpoch 353/500\nEpoch 354/500\nEpoch 355/500\nEpoch 356/500\nEpoch 357/500\nEpoch 358/500\nEpoch 359/500\nEpoch 360/500\nEpoch 361/500\nEpoch 362/500\nEpoch 363/500\nEpoch 364/500\nEpoch 365/500\nEpoch 366/500\nEpoch 367/500\nEpoch 368/500\nEpoch 369/500\nEpoch 370/500\nEpoch 371/500\nEpoch 372/500\nEpoch 373/500\nEpoch 374/500\nEpoch 375/500\nEpoch 376/500\nEpoch 377/500\nEpoch 378/500\nEpoch 379/500\nEpoch 380/500\nEpoch 381/500\nEpoch 382/500\nEpoch 383/500\nEpoch 384/500\nEpoch 385/500\nEpoch 386/500\nEpoch 387/500\nEpoch 388/500\nEpoch 389/500\nEpoch 390/500\nEpoch 391/500\nEpoch 392/500\nEpoch 393/500\nEpoch 394/500\nEpoch 395/500\nEpoch 396/500\nEpoch 397/500\nEpoch 398/500\nEpoch 399/500\nEpoch 400/500\nEpoch 401/500\nEpoch 402/500\nEpoch 403/500\nEpoch 404/500\nEpoch 405/500\nEpoch 406/500\nEpoch 407/500\nEpoch 408/500\nEpoch 409/500\nEpoch 410/500\nEpoch 411/500\nEpoch 412/500\nEpoch 413/500\nEpoch 414/500\nEpoch 415/500\nEpoch 416/500\nEpoch 417/500\nEpoch 418/500\nEpoch 419/500\nEpoch 420/500\nEpoch 421/500\nEpoch 422/500\nEpoch 423/500\nEpoch 424/500\nEpoch 425/500\nEpoch 426/500\nEpoch 427/500\nEpoch 428/500\nEpoch 429/500\nEpoch 430/500\nEpoch 431/500\nEpoch 432/500\nEpoch 433/500\nEpoch 434/500\nEpoch 435/500\nEpoch 436/500\nEpoch 437/500\nEpoch 438/500\nEpoch 439/500\nEpoch 440/500\nEpoch 441/500\nEpoch 442/500\nEpoch 443/500\nEpoch 444/500\nEpoch 445/500\nEpoch 446/500\nEpoch 447/500\nEpoch 448/500\nEpoch 449/500\nEpoch 450/500\nEpoch 451/500\nEpoch 452/500\nEpoch 453/500\nEpoch 454/500\nEpoch 455/500\nEpoch 456/500\nEpoch 457/500\nEpoch 458/500\nEpoch 459/500\nEpoch 460/500\nEpoch 461/500\nEpoch 462/500\nEpoch 463/500\nEpoch 464/500\nEpoch 465/500\nEpoch 466/500\nEpoch 467/500\nEpoch 468/500\nEpoch 469/500\nEpoch 470/500\nEpoch 471/500\nEpoch 472/500\nEpoch 473/500\nEpoch 474/500\nEpoch 475/500\nEpoch 476/500\nEpoch 477/500\nEpoch 478/500\nEpoch 479/500\nEpoch 480/500\nEpoch 481/500\nEpoch 482/500\nEpoch 483/500\nEpoch 484/500\nEpoch 485/500\nEpoch 486/500\nEpoch 487/500\nEpoch 488/500\nEpoch 489/500\nEpoch 490/500\nEpoch 491/500\nEpoch 492/500\nEpoch 493/500\nEpoch 494/500\nEpoch 495/500\nEpoch 496/500\nEpoch 497/500\nEpoch 498/500\nEpoch 499/500\nEpoch 500/500\n","output_type":"stream"}]},{"cell_type":"markdown","source":"### Save the model","metadata":{}},{"cell_type":"code","source":"model.save('model.h5')","metadata":{"execution":{"iopub.status.busy":"2024-04-21T16:32:15.053131Z","iopub.execute_input":"2024-04-21T16:32:15.053446Z","iopub.status.idle":"2024-04-21T16:32:15.110264Z","shell.execute_reply.started":"2024-04-21T16:32:15.053399Z","shell.execute_reply":"2024-04-21T16:32:15.109461Z"},"trusted":true},"execution_count":11,"outputs":[]},{"cell_type":"markdown","source":"### Plot model ","metadata":{}},{"cell_type":"code","source":"ku.plot_model(model, show_shapes = True)","metadata":{"execution":{"iopub.status.busy":"2024-04-21T16:32:15.111321Z","iopub.execute_input":"2024-04-21T16:32:15.111611Z","iopub.status.idle":"2024-04-21T16:32:15.198396Z","shell.execute_reply.started":"2024-04-21T16:32:15.111587Z","shell.execute_reply":"2024-04-21T16:32:15.197531Z"},"trusted":true},"execution_count":12,"outputs":[{"execution_count":12,"output_type":"execute_result","data":{"image/png":"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","text/plain":"<IPython.core.display.Image object>"},"metadata":{}}]},{"cell_type":"markdown","source":"## 5. Evaluate The Model","metadata":{}},{"cell_type":"code","source":"def evaluate_model_accuracy(model, input_data, label_data):\n    # Predict classes for input data\n    predictions = model.predict(input_data)\n    \n    # Convert predicted probabilities to class labels\n    predicted_classes = np.argmax(predictions, axis=1)\n    \n    # Convert one-hot encoded labels to class indices\n    true_classes = np.argmax(label_data, axis=1)\n    \n    # Compare predicted classes with true classes\n    correct_predictions = np.sum(predicted_classes == true_classes)\n    total_samples = len(true_classes)\n    \n    # Calculate accuracy\n    accuracy = correct_predictions / total_samples\n    \n    return accuracy\n\n# Usage\naccuracy = evaluate_model_accuracy(model, input_data, label_data)\nprint(\"Accuracy:\", accuracy)\n","metadata":{"execution":{"iopub.status.busy":"2024-04-21T16:32:15.201073Z","iopub.execute_input":"2024-04-21T16:32:15.201366Z","iopub.status.idle":"2024-04-21T16:32:15.728552Z","shell.execute_reply.started":"2024-04-21T16:32:15.201334Z","shell.execute_reply":"2024-04-21T16:32:15.727522Z"},"trusted":true},"execution_count":13,"outputs":[{"name":"stdout","text":"\u001b[1m142/142\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 2ms/step\nAccuracy: 0.8391285211267606\n","output_type":"stream"}]},{"cell_type":"code","source":"def evaluate_model_loss(model, input_data, label_data):\n    # Evaluate model loss\n    loss = model.evaluate(input_data, label_data, verbose=0)\n    \n    return loss\n\n# Usage\nloss = evaluate_model_loss(model, input_data, label_data)\nprint(\"Loss:\", loss)\n","metadata":{"execution":{"iopub.status.busy":"2024-04-21T16:32:15.729585Z","iopub.execute_input":"2024-04-21T16:32:15.729855Z","iopub.status.idle":"2024-04-21T16:32:16.339235Z","shell.execute_reply.started":"2024-04-21T16:32:15.729831Z","shell.execute_reply":"2024-04-21T16:32:16.338296Z"},"trusted":true},"execution_count":14,"outputs":[{"name":"stdout","text":"Loss: 1.0006893873214722\n","output_type":"stream"}]},{"cell_type":"markdown","source":"## 6. TEST MODEL","metadata":{}},{"cell_type":"markdown","source":"#### 5.1 Generating the text ","metadata":{}},{"cell_type":"code","source":"def generate_text(seed_text, num_next_words, model, max_sequence_length):\n    for _ in range(num_next_words):\n        \n        # tokenized the input model\n        token_list = tokenizer.texts_to_sequences([seed_text])[0]\n        \n        # padd the seqeunces\n        token_list = pad_sequences([token_list], maxlen = max_sequence_length, padding = 'pre')\n        \n        # feed the preprocessed input data into the model to generate the next word\n        predict_probs = model.predict(token_list)[0]\n        predict_index = np.argmax(predict_probs)\n        \n        # decode the sequence into human readable format\n        output_word = ''\n        for word, index in tokenizer.word_index.items():\n            if index == predict_index:\n                output_word = word\n                break\n            \n        # update the seed text (input_taken + generate word --> to predict next word)\n        seed_text += \" \" + output_word\n        \n    return seed_text.title()\n                ","metadata":{"execution":{"iopub.status.busy":"2024-04-21T16:32:16.340251Z","iopub.execute_input":"2024-04-21T16:32:16.340563Z","iopub.status.idle":"2024-04-21T16:32:16.348168Z","shell.execute_reply.started":"2024-04-21T16:32:16.340537Z","shell.execute_reply":"2024-04-21T16:32:16.347257Z"},"trusted":true},"execution_count":15,"outputs":[]},{"cell_type":"markdown","source":"## 7. Results","metadata":{}},{"cell_type":"code","source":"print(max_seq_len)","metadata":{"execution":{"iopub.status.busy":"2024-04-21T16:32:16.349226Z","iopub.execute_input":"2024-04-21T16:32:16.349550Z","iopub.status.idle":"2024-04-21T16:32:16.363253Z","shell.execute_reply.started":"2024-04-21T16:32:16.349525Z","shell.execute_reply":"2024-04-21T16:32:16.362456Z"},"trusted":true},"execution_count":16,"outputs":[{"name":"stdout","text":"17\n","output_type":"stream"}]},{"cell_type":"code","source":"# test the model\n\nprint('------------- Generated Text -----------------------')\nprint(generate_text('i am', 5, model, max_seq_len))\nprint()\n\nprint('------------- Generated Text -----------------------')\nprint(generate_text('Dear', 5, model, max_seq_len))\nprint()\n\nprint('------------- Generated Text -----------------------')\nprint(generate_text('Who are', 5, model, max_seq_len))\nprint()\n\nprint('------------- Generated Text -----------------------')\nprint(generate_text('Nothings', 5, model, max_seq_len))\nprint()\n\nprint('------------- Generated Text -----------------------')\nprint(generate_text('whats', 5, model, max_seq_len))\nprint()\n\n","metadata":{"execution":{"iopub.status.busy":"2024-04-21T16:32:16.364387Z","iopub.execute_input":"2024-04-21T16:32:16.364683Z","iopub.status.idle":"2024-04-21T16:32:17.923879Z","shell.execute_reply.started":"2024-04-21T16:32:16.364660Z","shell.execute_reply":"2024-04-21T16:32:17.922917Z"},"trusted":true},"execution_count":17,"outputs":[{"name":"stdout","text":"------------- Generated Text -----------------------\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 128ms/step\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 16ms/step\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 19ms/step\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 17ms/step\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 16ms/step\nI Am Must Press A Twostate Path\n\n------------- Generated Text -----------------------\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 16ms/step\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 17ms/step\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 17ms/step\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 17ms/step\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 18ms/step\nDear Gains For Republicans On Bold\n\n------------- Generated Text -----------------------\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 17ms/step\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 17ms/step\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 17ms/step\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 16ms/step\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 17ms/step\nWho Are We A Coop Liable Concept\n\n------------- Generated Text -----------------------\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 17ms/step\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 19ms/step\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 16ms/step\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 17ms/step\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 19ms/step\nNothings Gains For Republicans On Bold\n\n------------- Generated Text -----------------------\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 17ms/step\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 16ms/step\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 16ms/step\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 16ms/step\n\u001b[1m1/1\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 16ms/step\nWhats Going On In This Picture\n\n","output_type":"stream"}]},{"cell_type":"markdown","source":"# Streamlit App","metadata":{}},{"cell_type":"markdown","source":"## 1. Download the streamlit and local tunnel","metadata":{}},{"cell_type":"code","source":"!pip install streamlit","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"!npm install -g localtunnel","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## 1.1 Create app.py file","metadata":{}},{"cell_type":"code","source":"%%writefile app.py\n\nimport tensorflow as tf\nfrom tensorflow.keras.preprocessing.sequence import pad_sequences\nfrom tensorflow.keras.layers import Embedding, LSTM, Dense, Dropout\nfrom tensorflow.keras.preprocessing.text import Tokenizer\nfrom tensorflow.keras.models import Sequential\nfrom tensorflow.keras.callbacks import EarlyStopping\nfrom tensorflow.keras.preprocessing.text import tokenizer_from_json\nimport tensorflow.keras.utils as ku\nimport streamlit as st\nimport tensorflow as tf\nimport json\n\n# Load the tokenizer from the JSON file\nwith open(\"tokenizer.json\", \"r\", encoding=\"utf-8\") as f:\n    tokenizer_json = json.loads(f.read())\n    tokenizer = tokenizer_from_json(tokenizer_json)\n\n# Function to generate text based on input sequence\ndef generate_text(seed_text, num_next_words, model, max_sequence_length):\n    for num in range(num_next_words):\n        \n        # Tokenize the input sequence\n        token_list = tokenizer.texts_to_sequences([seed_text])[0]\n        \n        # Pad the sequence\n        token_list = pad_sequences([token_list], maxlen=max_sequence_length, padding='pre')\n        \n        # Feed the preprocessed input data into the model to generate the next word\n        predicted_probs = model.predict(token_list)[0]\n        \n        # Find the index of the word with the highest probability\n        predicted_index = np.argmax(predicted_probs)\n        \n        # Find the word corresponding to the predicted index\n        output_word = ''\n        for word, index in tokenizer.word_index.items():\n            if index == predicted_index:\n                output_word = word\n                break\n            \n        # Update the seed text (input taken + generated word --> to predict next word)\n        seed_text += \" \" + output_word\n    \n    return seed_text.title()\n\n# Streamlit app\ndef main():\n    st.title(\"Text Generation App\")\n    \n    # load the model\n    model = tf.keras.models.load_model('model.h5')\n    \n    # Input sequence from the user\n    input_sequence = st.text_input(\"Enter the input sequence:\")\n    \n    # Number of words to generate\n    num_next_words = st.slider(\"Number of next words to generate:\", min_value=1, max_value=10, value=5)\n    \n    # Specifying the max sequence length\n    max_seq_len = 17\n    \n    # Button to generate text\n    if st.button(\"Generate Text\"):\n        # Generate text based on the input sequence\n        generated_text = generate_text(input_sequence, num_next_words, model, max_seq_len)\n        \n        # Display the generated text\n        st.write(\"Generated Text:\", generated_text)\n\nif __name__ == \"__main__\":\n    main()","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## 1.3 Run Streamlit app ","metadata":{}},{"cell_type":"code","source":"import tensorflow as tf\n\nprint(\"TensorFlow version:\", tf.__version__)\n","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"! streamlit --version","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"!streamlit run app.py & npx localtunnel --port 8501","metadata":{"trusted":true},"execution_count":null,"outputs":[]}]}