Download train_model.py from d-e-e-k-11/LSTM1: direct link, hf CLI and curl.
- Browser
- Download file 4.44 kB
-
https://huggingface.co/d-e-e-k-11/LSTM1/resolve/main/train_model.py
- Command line
-
hf download hf://d-e-e-k-11/LSTM1/train_model.py
-
curl -L -o train_model.py https://huggingface.co/d-e-e-k-11/LSTM1/resolve/main/train_model.py
4.44 kB
| import pandas as pd | |
| import numpy as np | |
| import tensorflow as tf | |
| from tensorflow.keras.preprocessing.text import Tokenizer | |
| from tensorflow.keras.preprocessing.sequence import pad_sequences | |
| from tensorflow.keras.models import Sequential | |
| from tensorflow.keras.layers import Embedding, LSTM, Dense, SpatialDropout1D, Bidirectional, Dropout | |
| from sklearn.model_selection import train_test_split | |
| from sklearn.metrics import classification_report, confusion_matrix | |
| import pickle | |
| import os | |
| import pandas as pd | |
| import numpy as np | |
| import tensorflow as tf | |
| from tensorflow.keras.preprocessing.text import Tokenizer | |
| from tensorflow.keras.preprocessing.sequence import pad_sequences | |
| from tensorflow.keras.models import Model | |
| from tensorflow.keras.layers import Input, Embedding, LSTM, Dense, SpatialDropout1D, Bidirectional, Dropout, Layer, Concatenate | |
| import tensorflow.keras.backend as K | |
| import pickle | |
| import os | |
| # Custom Attention Layer | |
| class Attention(Layer): | |
| def __init__(self, **kwargs): | |
| super(Attention, self).__init__(**kwargs) | |
| def build(self, input_shape): | |
| self.W = self.add_weight(name='attention_weight', | |
| shape=(input_shape[-1], 1), | |
| initializer='random_normal', | |
| trainable=True) | |
| self.b = self.add_weight(name='attention_bias', | |
| shape=(input_shape[1], 1), | |
| initializer='zeros', | |
| trainable=True) | |
| super(Attention, self).build(input_shape) | |
| def call(self, x): | |
| e = K.tanh(K.dot(x, self.W) + self.b) | |
| a = K.softmax(e, axis=1) | |
| output = x * a | |
| return K.sum(output, axis=1) | |
| def train_advanced_model(file_path): | |
| print("Loading data for advanced model...") | |
| df = pd.read_csv(file_path) | |
| # Fill missing facts | |
| df['related_facts'] = df['related_facts'].fillna("No context provided.") | |
| # Advanced Preprocessing: Combine facts, question, and response | |
| # Structure: [FACTS] facts [SEP] [QUERY] question [SEP] [RES] response | |
| df['text'] = "[FACTS] " + df['related_facts'].astype(str) + \ | |
| " [QUERY] " + df['question'].astype(str) + \ | |
| " [RES] " + df['engine_response'].astype(str) | |
| y = df['best'].astype(int).values | |
| X_text = df['text'].astype(str).str.lower().values | |
| max_words = 15000 | |
| max_len = 300 | |
| tokenizer = Tokenizer(num_words=max_words, lower=True, split=' ') | |
| tokenizer.fit_on_texts(X_text) | |
| X_seq = tokenizer.texts_to_sequences(X_text) | |
| X_pad = pad_sequences(X_seq, maxlen=max_len) | |
| X_train, X_test, y_train, y_test = train_test_split(X_pad, y, test_size=0.15, random_state=42, stratify=y) | |
| # Save tokenizer immediately so it's available as soon as model starts saving checkpoints | |
| with open('tokenizer_advanced.pickle', 'wb') as handle: | |
| pickle.dump(tokenizer, handle, protocol=pickle.HIGHEST_PROTOCOL) | |
| print("Tokenizer saved.") | |
| # Advanced Arch: Bi-LSTM + Attention | |
| inputs = Input(shape=(max_len,)) | |
| embed = Embedding(max_words, 128)(inputs) | |
| drop1 = SpatialDropout1D(0.3)(embed) | |
| lstm = Bidirectional(LSTM(64, return_sequences=True))(drop1) | |
| attn = Attention()(lstm) | |
| dense1 = Dense(64, activation='relu')(attn) | |
| drop2 = Dropout(0.4)(dense1) | |
| outputs = Dense(1, activation='sigmoid')(drop2) | |
| model = Model(inputs=inputs, outputs=outputs) | |
| model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy']) | |
| print(model.summary()) | |
| # Training with Checkpointing | |
| batch_size = 128 | |
| epochs = 2 | |
| class_weight = {0: 1.0, 1: len(y[y==0]) / len(y[y==1])} | |
| checkpoint = tf.keras.callbacks.ModelCheckpoint( | |
| 'chatbot_performance_advanced.h5', | |
| monitor='val_accuracy', | |
| save_best_only=True, | |
| mode='max', | |
| verbose=1 | |
| ) | |
| print("Training advanced model with Attention...") | |
| model.fit( | |
| X_train, y_train, | |
| epochs=epochs, | |
| batch_size=batch_size, | |
| validation_split=0.1, | |
| class_weight=class_weight, | |
| callbacks=[checkpoint], | |
| verbose=1 | |
| ) | |
| print("Training complete.") | |
| if __name__ == "__main__": | |
| train_advanced_model('BP_MHS_V1.csv') | |