# Import to be able to import python package from src
import sys
sys.path.insert(0, '../src')

from datetime import datetime

import ontime as on
import pandas as pd
import numpy as np

import tensorflow as tf

from ontime.module.processing.pytorch.utils import create_dataset
from ontime.module.processing.common import train_test_split, normalize

PyTorch Dataset

Create test TimeSeries

v1 = np.arange(0,100)
v2 = np.arange(100,200)
arr = np.column_stack((v1,v2))

start_date = pd.Timestamp('2023-01-01')
time_indices = pd.date_range(start=start_date, periods=len(arr), freq='D')

ts = on.TimeSeries.from_times_and_values(time_indices, arr)

Create Splits

train, test = train_test_split(ts, test_split=0.3)
train_enc, transformer = normalize(train, return_transformer=True)
test_enc = transformer.transform(test)

Create DataSets

ds_train = create_dataset(
    train_enc,
    stride_length = 10,
    input_length = 6,
    target_length = 4
)

ds_test = create_dataset(
    test_enc,
    stride_length = 10,
    input_length = 6,
    target_length = 4
)

Check samples encoding

Get a sample

input = ds_train.data_array[0]
target = ds_train.labels_array[0]

Print input

transformer.inverse_transform(on.TimeSeries.from_data(input.T))

<TimeSeries (DataArray) (time: 6, component: 2, sample: 1)> Size: 96B
array([[[  0.],
        [100.]],

       [[  1.],
        [101.]],

       [[  2.],
        [102.]],

       [[  3.],
        [103.]],

       [[  4.],
        [104.]],

       [[  5.],
        [105.]]])
Coordinates:
  * time       (time) int64 48B 0 1 2 3 4 5
  * component  (component) <U1 8B '0' '1'
Dimensions without coordinates: sample
Attributes:
    static_covariates:  None
    hierarchy:          None

TimeSeries (DataArray)

• time: 6
• component: 2
• sample: 1

• 0.0 100.0 1.0 101.0 2.0 102.0 3.0 103.0 4.0 104.0 5.0 105.0

  array([[[  0.],
          [100.]],
  
         [[  1.],
          [101.]],
  
         [[  2.],
          [102.]],
  
         [[  3.],
          [103.]],
  
         [[  4.],
          [104.]],
  
         [[  5.],
          [105.]]])

• Coordinates: (2)
  • time
    (time)
    int64
    0 1 2 3 4 5

    array([0, 1, 2, 3, 4, 5])

  • component
    (component)
    <U1
    '0' '1'

    array(['0', '1'], dtype='<U1')

• Indexes: (2)
  • time
    PandasIndex

    PandasIndex(RangeIndex(start=0, stop=6, step=1, name='time'))

  • component
    PandasIndex

    PandasIndex(Index(['0', '1'], dtype='object', name='component'))

• Attributes: (2)

  static_covariates :

  None

  hierarchy :

  None

Print target

transformer.inverse_transform(on.TimeSeries.from_data(target.T))

<TimeSeries (DataArray) (time: 4, component: 2, sample: 1)> Size: 64B
array([[[  6.],
        [106.]],

       [[  7.],
        [107.]],

       [[  8.],
        [108.]],

       [[  9.],
        [109.]]])
Coordinates:
  * time       (time) int64 32B 0 1 2 3
  * component  (component) <U1 8B '0' '1'
Dimensions without coordinates: sample
Attributes:
    static_covariates:  None
    hierarchy:          None

TimeSeries (DataArray)

• time: 4
• component: 2
• sample: 1

• 6.0 106.0 7.0 107.0 8.0 108.0 9.0 109.0

  array([[[  6.],
          [106.]],
  
         [[  7.],
          [107.]],
  
         [[  8.],
          [108.]],
  
         [[  9.],
          [109.]]])

• Coordinates: (2)
  • time
    (time)
    int64
    0 1 2 3

    array([0, 1, 2, 3])

  • component
    (component)
    <U1
    '0' '1'

    array(['0', '1'], dtype='<U1')

• Indexes: (2)
  • time
    PandasIndex

    PandasIndex(RangeIndex(start=0, stop=4, step=1, name='time'))

  • component
    PandasIndex

    PandasIndex(Index(['0', '1'], dtype='object', name='component'))

• Attributes: (2)

  static_covariates :

  None

  hierarchy :

  None