Spaces:
Sleeping
Sleeping
File size: 1,763 Bytes
4ad3b10 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 | import numpy as np
import tensorflow as tf
from tensorflow.keras.models import Model
from tensorflow.keras.layers import Input, LSTM, RepeatVector, TimeDistributed, Dense
from sklearn.preprocessing import StandardScaler, MinMaxScaler
import joblib
import logging
import os
logger = logging.getLogger(__name__)
def create_sequences(data, time_steps=30):
"""
Creates overlapping sequences for LSTM training.
"""
xs = []
for i in range(len(data) - time_steps):
xs.append(data[i:(i + time_steps)])
return np.array(xs)
def build_lstm_autoencoder(input_shape):
"""
Standard LSTM Autoencoder Architecture.
"""
inputs = Input(shape=input_shape)
# Encoder
L1 = LSTM(128, activation='relu', return_sequences=True)(inputs)
L2 = LSTM(64, activation='relu', return_sequences=False)(L1)
# Decoder
L3 = RepeatVector(input_shape[0])(L2)
L4 = LSTM(64, activation='relu', return_sequences=True)(L3)
L5 = LSTM(128, activation='relu', return_sequences=True)(L4)
output = TimeDistributed(Dense(input_shape[1]))(L5)
model = Model(inputs=inputs, outputs=output)
model.compile(optimizer='adam', loss='mse')
return model
def save_artifact(obj, path):
"""
Saves a model or scaler to the specified path.
"""
os.makedirs(os.path.dirname(path), exist_ok=True)
if path.endswith('.h5') or path.endswith('.keras'):
obj.save(path)
else:
joblib.dump(obj, path)
def load_artifact(path):
"""
Loads a model or scaler from the specified path.
"""
if not os.path.exists(path):
return None
if path.endswith('.h5') or path.endswith('.keras'):
return tf.keras.models.load_model(path)
else:
return joblib.load(path)
|