Предварительная обработка данных

import numpy as np
import tensorflow as tf
 
data = np.arange(1,1001,1)
data = np.expand_dims(data, axis=1)
data = np.hstack([data, data + 100])
 
n = data.shape[0]
train_data = data[0:int(n*0.7)]
val_data = data[int(n*0.7):int(n*0.9)]
test_data = data[int(n*0.9):]
 
mean = np.mean(data, axis=0)
std = np.std(data, axis=0)
 
train_data = (train_data - mean) / std
val_data = (val_data - mean) / std
test_data = (test_data - mean) / std
 
def split_window(features): 
    inputs = features[:, :6, :]  
    labels = features[:, 6:, :]  
    labels = tf.stack([labels[:, :, 1]], axis=-1)  
    inputs.set_shape([None, 6, None])   
    labels.set_shape([None, 1, None])   
    return inputs, labels
 
def make_dataset(data):
    data = np.array(data, dtype=np.float32)
    ds = tf.keras.utils.timeseries_dataset_from_array(
        data=data,
        targets=None,
        sequence_length=7,
        sequence_stride=1,
        shuffle=True,
        batch_size=32,
    )
    ds = ds.map(split_window)
    return ds
 
model = tf.keras.Sequential([
    tf.keras.layers.Dense(units=1)
])
 
model.compile(
        loss=tf.losses.MeanSquaredError(),
        optimizer=tf.optimizers.Adam(),
        metrics=[tf.metrics.MeanAbsoluteError()]
)
 
history = model.fit(
    make_dataset(train_data),
    epochs=100,
    validation_data=make_dataset(val_data)
)
 
scores = model.evaluate(make_dataset(test_data))