mirror of
https://github.com/wassname/pytorch-transformer-ts.git
synced 2026-08-07 11:27:21 +08:00
1.5 MiB
1.5 MiB
In [6]:
%matplotlib inlineIn [80]:
from typing import List, Optional, Iterable, Dict, Any
from itertools import islice
from functools import lru_cache, partial
from dataclasses import dataclass
import pandas as pd
from pandas.tseries.frequencies import to_offset
import numpy as np
from matplotlib import pyplot as plt
import matplotlib.dates as mdates
import tqdm.auto as tqdm
import pytorch_lightning as pl
from pytorch_lightning.loggers import CSVLogger
from datasets import load_dataset, interleave_datasets
from datasets.iterable_dataset import RandomlyCyclingMultiSourcesExamplesIterable
import torch
import torch.nn as nn
from torch.utils.data import DataLoader
from gluonts.core.component import validated
from gluonts.dataset.common import Dataset, ListDataset, DatasetCollection, ProcessDataEntry
from gluonts.dataset.field_names import FieldName
from gluonts.itertools import Cyclic, PseudoShuffled, IterableSlice
from gluonts.time_feature import (
time_features_from_frequency_str,
TimeFeature,
MinuteOfHour,
HourOfDay,
DayOfWeek,
DayOfMonth,
DayOfYear,
)
from gluonts.torch.modules.loss import DistributionLoss, NegativeLogLikelihood
from gluonts.transform import (
Transformation,
Chain,
RemoveFields,
SetField,
AsNumpyArray,
AddObservedValuesIndicator,
AddTimeFeatures,
AddAgeFeature,
VstackFeatures,
InstanceSplitter,
ValidationSplitSampler,
TestSplitSampler,
ExpectedNumInstanceSampler,
SelectFields,
InstanceSampler,
)
from gluonts.torch.util import (
IterableDataset,
)
from gluonts.evaluation import make_evaluation_predictions, Evaluator
from gluonts.torch.model.estimator import PyTorchLightningEstimator
from gluonts.torch.model.predictor import PyTorchPredictor
from gluonts.torch.distributions import (
DistributionOutput,
StudentTOutput,
)
from gluonts.torch.util import weighted_average
from gluonts.torch.modules.scaler import MeanScaler, NOPScaler
from gluonts.torch.modules.feature import FeatureEmbedder
from gluonts.time_feature import get_lags_for_frequency
from gluonts.dataset.repository.datasets import get_datasetIn [81]:
time_features=[
MinuteOfHour(),
HourOfDay(),
DayOfWeek(),
DayOfMonth(),
DayOfYear(),
]
def add_time_feature(data, freq):
length = len(data[FieldName.TARGET])
start = pd.Period(data[FieldName.START], freq)
index = pd.period_range(start, periods=length, freq=freq)
data[FieldName.FEAT_TIME] = np.vstack(
[feat(index) for feat in time_features]
).astype(np.float32)
age = np.log10(2.0 + np.arange(length, dtype=np.float32))
data[FieldName.FEAT_AGE] = age.reshape((1, length))
return dataIn [82]:
@lru_cache(10_000)
def _as_period(val, freq):
return pd.Period(val, freq)
@dataclass
class HFDataset(Dataset):
def __init__(self, dataset, freq, shuffle=False):
super().__init__()
transform = partial(add_time_feature, freq=freq)
self.dataset = dataset.with_format("np").map(
transform, num_proc=8, keep_in_memory=False,
)
if shuffle:
self.dataset = self.dataset.shuffle()
self.freq = to_offset(freq)
def __iter__(self):
for data in self.dataset:
yield {
FieldName.START: _as_period(data[FieldName.START], self.freq),
FieldName.TARGET: data[FieldName.TARGET],
FieldName.ITEM_ID: data[FieldName.ITEM_ID],
FieldName.FEAT_TIME: np.stack(data[FieldName.FEAT_TIME],0),
FieldName.FEAT_AGE: np.stack(data[FieldName.FEAT_AGE],0),
}
def __len__(self) -> int:
return len(self.dataset)In [83]:
dataset_1 = load_dataset("monash_tsf", "traffic_hourly")WARNING:datasets.builder:Reusing dataset monash_tsf (/home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050)
0%| | 0/3 [00:00<?, ?it/s]
In [84]:
dataset_2 = load_dataset("monash_tsf", "weather")WARNING:datasets.builder:Reusing dataset monash_tsf (/home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050)
0%| | 0/3 [00:00<?, ?it/s]
In [85]:
dataset_3 = load_dataset("monash_tsf", "tourism_monthly")WARNING:datasets.builder:Reusing dataset monash_tsf (/home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050)
0%| | 0/3 [00:00<?, ?it/s]
In [227]:
dataset_4 = load_dataset("monash_tsf", "london_smart_meters")WARNING:datasets.builder:Reusing dataset monash_tsf (/home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050)
0%| | 0/3 [00:00<?, ?it/s]
In [276]:
len(dataset_1["train"])Out [276]:
862
In [178]:
len(dataset_2["train"])Out [178]:
3010
In [179]:
len(dataset_3["train"])Out [179]:
366
In [279]:
len(dataset_4["train"])Out [279]:
5560
In [229]:
train_ds_1 = HFDataset(dataset_1["train"], freq="1H", shuffle=True)WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-8c3c0af03e5a73c5.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-13aba5b43ce908ab.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-68ddd416ed320a07.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-f0233ec5142eaa72.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-96dd364063c02cd3.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-89d31905648bc17a.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-70b8acf2f33f30e2.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-db6aa8a5bfed75e7.arrow
In [230]:
train_ds_2 = HFDataset(dataset_2["train"], freq="1D", shuffle=True)WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-cc05812313cd40b2.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-3ce08266efcaf8cb.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-a61774851ee48381.arrow WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-eae34bf137f66a70.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-eb4223ce934fc339.arrow WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-4b8e22f982d21676.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-43ad5320d89ce30e.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-980d162d54c7a5bb.arrow
In [88]:
train_ds_3 = HFDataset(dataset_3["train"], freq="1M", shuffle=True)WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-ad0a32eb8a39667c.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-81ec2e328ae3b844.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-d7b61bf51c9577b1.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-5954e6e8c6a89469.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-45b9e4c77db99750.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-b6684acd7d4b90a4.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-f5501b4793483be0.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-e576b680b5886c8c.arrow
In [231]:
train_ds_4 = HFDataset(dataset_4["train"], freq="30T", shuffle=True)WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-9bd0929b257b8a70.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-ce12c2c0ad527ce0.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-38ec19ddd65abfb3.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-48f96869b2835eb4.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-97473647197065d0.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-df155ca2bdf08a05.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-66d9c2ddc149606f.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-a5028a679a9da0a9.arrow
In [180]:
val_ds_1 = HFDataset(dataset_1["validation"], freq="1H", shuffle=True)WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-3596030a3da34a6c.arrow WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-8ade8fb025e410ef.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-655b67348586ad9c.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-6432e6e6eba4a097.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-76994d19762a78c0.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-bea5b99b93c5b53f.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-ad098c9df0a3698d.arrow WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-3e6c936f1b6652e2.arrow
In [181]:
val_ds_2 = HFDataset(dataset_2["validation"], freq="1D", shuffle=True)WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-1069c6339983775a.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-3508de63c10e5a53.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-4003c98c902f06de.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-d440f1db9a9e2c35.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-b5a29535f3f0b57e.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-41260549fad83e1a.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-3c1123ab6a16d5d3.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-297a37373ed51b67.arrow
In [182]:
val_ds_3 = HFDataset(dataset_3["validation"], freq="1M", shuffle=True)WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-461bcd71d6d2d30c.arrow WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-7ef229e9d60bb9ff.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-0e83671840bcd195.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-73dfe605cb0d6a6a.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-3b9a22aa0fd2bc74.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-74cbdda51b82b75d.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-34553b27db7c7a9a.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-4bfec924652065e3.arrow
In [232]:
val_ds_4 = HFDataset(dataset_4["validation"], freq="30T", shuffle=True)WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-db7c9c157de0715a.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-ec61a6194bc2c76b.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-b21798d1e7aa77c1.arrow WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-3955fa385a4d2085.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-e917407729770037.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-d0aae877d1c9d26f.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-ecd2edc01013ad70.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-4343a936c634e6ee.arrow
In [92]:
test_ds_1 = HFDataset(dataset_1["test"], freq="1H",)WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-9055f3808d86232d.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-ead11dae3f44b7c5.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-3febf1d14efbf1a8.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-31c84bfe42d6d7f4.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-64767e4d3eaa944d.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-d0b7ad4afbe8c427.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-a2bb9d6dbbedb939.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/traffic_hourly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-f1c0bfaaf321500e.arrow
In [93]:
test_ds_2 = HFDataset(dataset_2["test"], freq="1D",)WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-0b25ebe60e9e4e5a.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-8fb6953cc8f42087.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-46b318263d275b59.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-e68415a06dff3a69.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-13c18c9a228d4f61.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-b452474d5ebe0ec2.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-02cd57facb7b8b48.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/weather/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-417f88230cab327b.arrow
In [94]:
test_ds_3 = HFDataset(dataset_3["test"], freq="1M",)WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-40cd4f2d4ee75163.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-f719d2d360988094.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-9aaf3cded6dac594.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-3d1a61d7b7b8c178.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-8277985664f5bc6c.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-bb9be5ac08e6259c.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-020872a8f99d7ca0.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/tourism_monthly/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-4fb935b373299e6c.arrow
In [233]:
test_ds_4 = HFDataset(dataset_4["test"], freq="30T",)WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-e6671ddd677d538c.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-601aee543ca82f63.arrow WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-5e185211b0341000.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-e01fe2c465df9a49.arrow WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-0bd87771d0f4e38d.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-7ff4f6f0df4c0c92.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-165fa1f3b7b0aa98.arrow
WARNING:datasets.arrow_dataset:Loading cached processed dataset at /home/kashif/.cache/huggingface/datasets/monash_tsf/london_smart_meters/1.0.0/82998723d55d6edbc664cfaebe371004d849846ee5f61b57f1f974000d44c050/cache-6464a15aadbbd181.arrow
In [281]:
train_ds_list = [train_ds_1, train_ds_2, train_ds_3, train_ds_4]
train_ds_size = np.array([len(ds) for ds in train_ds_list])
raw_weights = 1/train_ds_size
normalization_factor = 1/sum(raw_weights)
probablities = raw_weights * normalization_factorIn [283]:
train_ds = RandomlyCyclingMultiSourcesExamplesIterable(
train_ds_list,
generator=np.random.default_rng(),
probabilities=probablities,
)In [284]:
val_ds = DatasetCollection(datasets=[val_ds_1, val_ds_2, val_ds_3], interleave=True)In [246]:
class TransformerModel(nn.Module):
@validated()
def __init__(
self,
context_length: int,
prediction_length: int,
num_feat_dynamic_real: int,
num_feat_static_real: int,
num_feat_static_cat: int,
cardinality: List[int],
# transformer arguments
nhead: int,
num_encoder_layers: int,
num_decoder_layers: int,
dim_feedforward: int,
activation: str = "gelu",
dropout: float = 0.1,
# univariate input
input_size: int = 1,
embedding_dimension: Optional[List[int]] = None,
distr_output: DistributionOutput = StudentTOutput(),
freq: Optional[str] = None,
lags_seq: Optional[List[int]] = None,
scaling: bool = True,
num_parallel_samples: int = 100,
) -> None:
super().__init__()
assert (freq is not None) or (lags_seq is not None), "either freq or lags_seq must be given"
self.input_size = input_size
self.target_shape = distr_output.event_shape
self.num_feat_dynamic_real = num_feat_dynamic_real
self.num_feat_static_cat = num_feat_static_cat
self.num_feat_static_real = num_feat_static_real
self.embedding_dimension = (
embedding_dimension
if embedding_dimension is not None or cardinality is None
else [min(50, (cat + 1) // 2) for cat in cardinality]
)
self.lags_seq = lags_seq or get_lags_for_frequency(freq_str=freq)
self.num_parallel_samples = num_parallel_samples
self.history_length = context_length + max(self.lags_seq)
self.embedder = FeatureEmbedder(
cardinalities=cardinality,
embedding_dims=self.embedding_dimension,
)
if scaling:
self.scaler = MeanScaler(dim=1, keepdim=True)
else:
self.scaler = NOPScaler(dim=1, keepdim=True)
# total feature size
d_model = self.input_size * len(self.lags_seq) + self._number_of_features
self.context_length = context_length
self.prediction_length = prediction_length
self.distr_output = distr_output
self.param_proj = distr_output.get_args_proj(d_model)
# transformer enc-decoder and mask initializer
self.transformer = nn.Transformer(
d_model=d_model,
nhead=nhead,
num_encoder_layers=num_encoder_layers,
num_decoder_layers=num_decoder_layers,
dim_feedforward=dim_feedforward,
dropout=dropout,
activation=activation,
batch_first=True,
)
# causal decoder tgt mask
self.register_buffer(
"tgt_mask",
self.transformer.generate_square_subsequent_mask(prediction_length),
)
@property
def _number_of_features(self) -> int:
return (
sum(self.embedding_dimension)
+ self.num_feat_dynamic_real
+ self.num_feat_static_real
+ self.input_size # the log(scale)
)
@property
def _past_length(self) -> int:
return self.context_length + max(self.lags_seq)
def get_lagged_subsequences(
self,
sequence: torch.Tensor,
subsequences_length: int,
shift: int = 0
) -> torch.Tensor:
"""
Returns lagged subsequences of a given sequence.
Parameters
----------
sequence : Tensor
the sequence from which lagged subsequences should be extracted.
Shape: (N, T, C).
subsequences_length : int
length of the subsequences to be extracted.
shift: int
shift the lags by this amount back.
Returns
--------
lagged : Tensor
a tensor of shape (N, S, C, I), where S = subsequences_length and
I = len(indices), containing lagged subsequences. Specifically,
lagged[i, j, :, k] = sequence[i, -indices[k]-S+j, :].
"""
sequence_length = sequence.shape[1]
indices = [l - shift for l in self.lags_seq]
assert max(indices) + subsequences_length <= sequence_length, (
f"lags cannot go further than history length, found lag {max(indices)} "
f"while history length is only {sequence_length}"
)
lagged_values = []
for lag_index in indices:
begin_index = -lag_index - subsequences_length
end_index = -lag_index if lag_index > 0 else None
lagged_values.append(sequence[:, begin_index:end_index, ...])
return torch.stack(lagged_values, dim=-1)
def _check_shapes(
self,
prior_input: torch.Tensor,
inputs: torch.Tensor,
features: Optional[torch.Tensor],
) -> None:
assert len(prior_input.shape) == len(inputs.shape)
assert (
len(prior_input.shape) == 2 and self.input_size == 1
) or prior_input.shape[2] == self.input_size
assert (len(inputs.shape) == 2 and self.input_size == 1) or inputs.shape[
-1
] == self.input_size
assert (
features is None or features.shape[2] == self._number_of_features
), f"{features.shape[2]}, expected {self._number_of_features}"
def create_network_inputs(
self,
feat_static_cat: torch.Tensor,
feat_static_real: torch.Tensor,
past_time_feat: torch.Tensor,
past_target: torch.Tensor,
past_observed_values: torch.Tensor,
future_time_feat: Optional[torch.Tensor] = None,
future_target: Optional[torch.Tensor] = None,
):
# time feature
time_feat = (
torch.cat(
(
past_time_feat[:, self._past_length - self.context_length :, ...],
future_time_feat,
),
dim=1,
)
if future_target is not None
else past_time_feat[:, self._past_length - self.context_length :, ...]
)
# target
context = past_target[:, -self.context_length :]
observed_context = past_observed_values[:, -self.context_length :]
_, scale = self.scaler(context, observed_context)
inputs = (
torch.cat((past_target, future_target), dim=1) / scale
if future_target is not None
else past_target / scale
)
inputs_length = (
self._past_length + self.prediction_length
if future_target is not None
else self._past_length
)
assert inputs.shape[1] == inputs_length
subsequences_length = (
self.context_length + self.prediction_length
if future_target is not None
else self.context_length
)
# embeddings
embedded_cat = self.embedder(feat_static_cat)
log_scale = scale.log() if self.input_size == 1 else scale.squeeze(1).log()
static_feat = torch.cat(
(embedded_cat, feat_static_real, log_scale),
dim=1,
)
expanded_static_feat = static_feat.unsqueeze(1).expand(
-1, time_feat.shape[1], -1
)
features = torch.cat((expanded_static_feat, time_feat), dim=-1)
#self._check_shapes(prior_input, inputs, features)
#sequence = torch.cat((prior_input, inputs), dim=1)
lagged_sequence = self.get_lagged_subsequences(
sequence=inputs,
subsequences_length=subsequences_length,
)
lags_shape = lagged_sequence.shape
reshaped_lagged_sequence = lagged_sequence.reshape(
lags_shape[0], lags_shape[1], -1
)
transformer_inputs = torch.cat((reshaped_lagged_sequence, features), dim=-1)
return transformer_inputs, scale, static_feat
def output_params(self, transformer_inputs):
enc_input = transformer_inputs[:, :self.context_length, ...]
dec_input = transformer_inputs[:, self.context_length:, ...]
enc_out = self.transformer.encoder(
enc_input
)
dec_output = self.transformer.decoder(
dec_input,
enc_out,
tgt_mask=self.tgt_mask
)
return self.param_proj(dec_output)
@torch.jit.ignore
def output_distribution(
self, params, scale=None, trailing_n=None
) -> torch.distributions.Distribution:
sliced_params = params
if trailing_n is not None:
sliced_params = [p[:, -trailing_n:] for p in params]
return self.distr_output.distribution(sliced_params, scale=scale)
# for prediction
def forward(
self,
feat_static_cat: torch.Tensor,
feat_static_real: torch.Tensor,
past_time_feat: torch.Tensor,
past_target: torch.Tensor,
past_observed_values: torch.Tensor,
future_time_feat: torch.Tensor,
num_parallel_samples: Optional[int] = None,
) -> torch.Tensor:
if num_parallel_samples is None:
num_parallel_samples = self.num_parallel_samples
encoder_inputs, scale, static_feat = self.create_network_inputs(
feat_static_cat,
feat_static_real,
past_time_feat,
past_target,
past_observed_values,
)
enc_out = self.transformer.encoder(encoder_inputs)
repeated_scale = scale.repeat_interleave(
repeats=self.num_parallel_samples, dim=0
)
repeated_past_target = (
past_target.repeat_interleave(
repeats=self.num_parallel_samples, dim=0
)
/ repeated_scale
)
expanded_static_feat = static_feat.unsqueeze(1).expand(
-1, future_time_feat.shape[1], -1
)
features = torch.cat((expanded_static_feat, future_time_feat), dim=-1)
repeated_features = features.repeat_interleave(
repeats=self.num_parallel_samples, dim=0
)
repeated_enc_out = enc_out.repeat_interleave(
repeats=self.num_parallel_samples, dim=0
)
future_samples = []
# greedy decoding
for k in range(self.prediction_length):
#self._check_shapes(repeated_past_target, next_sample, next_features)
#sequence = torch.cat((repeated_past_target, next_sample), dim=1)
lagged_sequence = self.get_lagged_subsequences(
sequence=repeated_past_target,
subsequences_length=1+k,
shift=1,
)
lags_shape = lagged_sequence.shape
reshaped_lagged_sequence = lagged_sequence.reshape(
lags_shape[0], lags_shape[1], -1
)
decoder_input = torch.cat((reshaped_lagged_sequence, repeated_features[:, : k+1]), dim=-1)
output = self.transformer.decoder(decoder_input, repeated_enc_out)
params = self.param_proj(output[:,-1:])
distr = self.output_distribution(params, scale=repeated_scale)
next_sample = distr.sample()
repeated_past_target = torch.cat(
(repeated_past_target, next_sample / repeated_scale), dim=1
)
future_samples.append(next_sample)
concat_future_samples = torch.cat(future_samples, dim=1)
return concat_future_samples.reshape(
(-1, self.num_parallel_samples, self.prediction_length)
+ self.target_shape,
)In [247]:
class TransformerLightningModule(pl.LightningModule):
def __init__(
self,
model: TransformerModel,
loss: DistributionLoss = NegativeLogLikelihood(),
lr: float = 1e-3,
weight_decay: float = 1e-8,
) -> None:
super().__init__()
self.save_hyperparameters()
self.model = model
self.loss = loss
self.lr = lr
self.weight_decay = weight_decay
def training_step(self, batch, batch_idx: int):
"""Execute training step"""
train_loss = self(batch)
self.log(
"train_loss",
train_loss,
on_epoch=True,
on_step=False,
prog_bar=True,
)
return train_loss
def validation_step(self, batch, batch_idx: int):
"""Execute validation step"""
with torch.inference_mode():
val_loss = self(batch)
self.log(
"val_loss", val_loss, on_epoch=True, on_step=False, prog_bar=True
)
return val_loss
def configure_optimizers(self):
"""Returns the optimizer to use"""
return torch.optim.Adam(
self.model.parameters(),
lr=self.lr,
weight_decay=self.weight_decay,
capturable=True,
)
def forward(self, batch):
feat_static_cat = batch["feat_static_cat"]
feat_static_real = batch["feat_static_real"]
past_time_feat = batch["past_time_feat"]
past_target = batch["past_target"]
future_time_feat = batch["future_time_feat"]
future_target = batch["future_target"]
past_observed_values = batch["past_observed_values"]
future_observed_values = batch["future_observed_values"]
transformer_inputs, scale, _ = self.model.create_network_inputs(
feat_static_cat,
feat_static_real,
past_time_feat,
past_target,
past_observed_values,
future_time_feat,
future_target,
)
params = self.model.output_params(transformer_inputs)
distr = self.model.output_distribution(params, scale)
loss_values = self.loss(distr, future_target)
if len(self.model.target_shape) == 0:
loss_weights = future_observed_values
else:
loss_weights, _ = future_observed_values.min(dim=-1, keepdim=False)
return weighted_average(loss_values, weights=loss_weights)In [248]:
PREDICTION_INPUT_NAMES = [
"feat_static_cat",
"feat_static_real",
"past_time_feat",
"past_target",
"past_observed_values",
"future_time_feat",
]
TRAINING_INPUT_NAMES = PREDICTION_INPUT_NAMES + [
"future_target",
"future_observed_values",
]In [249]:
class TransformerEstimator(PyTorchLightningEstimator):
@validated()
def __init__(
self,
prediction_length: int,
# Transformer arguments
nhead: int,
num_encoder_layers: int,
num_decoder_layers: int,
dim_feedforward: int,
input_size: int = 1,
activation: str = "gelu",
dropout: float = 0.1,
context_length: Optional[int] = None,
num_feat_dynamic_real: int = 0,
num_feat_static_cat: int = 0,
num_feat_static_real: int = 0,
cardinality: Optional[List[int]] = None,
embedding_dimension: Optional[List[int]] = None,
distr_output: DistributionOutput = StudentTOutput(),
loss: DistributionLoss = NegativeLogLikelihood(),
scaling: bool = True,
freq: Optional[str] = None,
lags_seq: Optional[List[int]] = None,
time_features: Optional[List[TimeFeature]] = None,
num_parallel_samples: int = 100,
batch_size: int = 32,
num_batches_per_epoch: int = 50,
trainer_kwargs: Optional[Dict[str, Any]] = dict(),
train_sampler: Optional[InstanceSampler] = None,
validation_sampler: Optional[InstanceSampler] = None,
) -> None:
trainer_kwargs = {
"max_epochs": 100,
**trainer_kwargs,
}
super().__init__(trainer_kwargs=trainer_kwargs)
self.freq = freq
self.context_length = (
context_length if context_length is not None else prediction_length
)
self.prediction_length = prediction_length
self.distr_output = distr_output
self.loss = loss
self.input_size = input_size
self.nhead = nhead
self.num_encoder_layers = num_encoder_layers
self.num_decoder_layers = num_decoder_layers
self.activation = activation
self.dim_feedforward = dim_feedforward
self.dropout = dropout
self.num_feat_dynamic_real = num_feat_dynamic_real
self.num_feat_static_cat = num_feat_static_cat
self.num_feat_static_real = num_feat_static_real
self.cardinality = (
cardinality if cardinality and num_feat_static_cat > 0 else [1]
)
self.embedding_dimension = embedding_dimension
self.scaling = scaling
self.lags_seq = lags_seq
self.time_features = (
time_features
if time_features is not None
else time_features_from_frequency_str(self.freq)
)
self.num_parallel_samples = num_parallel_samples
self.batch_size = batch_size
self.num_batches_per_epoch = num_batches_per_epoch
self.train_sampler = train_sampler or ExpectedNumInstanceSampler(
num_instances=1.0, min_future=prediction_length
)
self.validation_sampler = validation_sampler or ValidationSplitSampler(
min_future=prediction_length
)
def create_transformation(self) -> Transformation:
remove_field_names = []
if self.num_feat_static_real == 0:
remove_field_names.append(FieldName.FEAT_STATIC_REAL)
if self.num_feat_dynamic_real == 0:
remove_field_names.append(FieldName.FEAT_DYNAMIC_REAL)
return Chain(
[RemoveFields(field_names=remove_field_names)]
+ (
[SetField(output_field=FieldName.FEAT_STATIC_CAT, value=[0])]
if not self.num_feat_static_cat > 0
else []
)
+ (
[
SetField(
output_field=FieldName.FEAT_STATIC_REAL, value=[0.0]
)
]
if not self.num_feat_static_real > 0
else []
)
+ [
AsNumpyArray(
field=FieldName.FEAT_STATIC_CAT,
expected_ndim=1,
dtype=int,
),
AsNumpyArray(
field=FieldName.FEAT_STATIC_REAL,
expected_ndim=1,
),
AsNumpyArray(
field=FieldName.TARGET,
# in the following line, we add 1 for the time dimension
expected_ndim=1 + len(self.distr_output.event_shape),
),
AddObservedValuesIndicator(
target_field=FieldName.TARGET,
output_field=FieldName.OBSERVED_VALUES,
),
# AddTimeFeatures(
# start_field=FieldName.START,
# target_field=FieldName.TARGET,
# output_field=FieldName.FEAT_TIME,
# time_features=self.time_features,
# pred_length=self.prediction_length,
# ),
# AddAgeFeature(
# target_field=FieldName.TARGET,
# output_field=FieldName.FEAT_AGE,
# pred_length=self.prediction_length,
# log_scale=True,
# ),
VstackFeatures(
output_field=FieldName.FEAT_TIME,
input_fields=[FieldName.FEAT_TIME, FieldName.FEAT_AGE]
+ (
[FieldName.FEAT_DYNAMIC_REAL]
if self.num_feat_dynamic_real > 0
else []
),
),
]
)
def _create_instance_splitter(
self, module: TransformerLightningModule, mode: str
):
assert mode in ["training", "validation", "test"]
instance_sampler = {
"training": self.train_sampler,
"validation": self.validation_sampler,
"test": TestSplitSampler(),
}[mode]
return InstanceSplitter(
target_field=FieldName.TARGET,
is_pad_field=FieldName.IS_PAD,
start_field=FieldName.START,
forecast_start_field=FieldName.FORECAST_START,
instance_sampler=instance_sampler,
past_length=module.model._past_length,
future_length=self.prediction_length,
time_series_fields=[
FieldName.FEAT_TIME,
FieldName.OBSERVED_VALUES,
],
dummy_value=self.distr_output.value_in_support,
)
def create_training_data_loader(
self,
data: Dataset,
module: TransformerLightningModule,
shuffle_buffer_length: Optional[int] = None,
**kwargs,
) -> Iterable:
transformation = self._create_instance_splitter(
module, "training"
) + SelectFields(TRAINING_INPUT_NAMES)
training_instances = transformation.apply(
Cyclic(data)
if shuffle_buffer_length is None
else PseudoShuffled(
Cyclic(data), shuffle_buffer_length=shuffle_buffer_length
)
)
return IterableSlice(
iter(
DataLoader(
IterableDataset(training_instances),
batch_size=self.batch_size,
pin_memory=True,
persistent_workers=kwargs.get("num_workers", 0) > 0,
**kwargs,
)
),
self.num_batches_per_epoch,
)
def create_validation_data_loader(
self,
data: Dataset,
module: TransformerLightningModule,
**kwargs,
) -> Iterable:
transformation = self._create_instance_splitter(
module, "validation"
) + SelectFields(TRAINING_INPUT_NAMES)
validation_instances = transformation.apply(data)
return DataLoader(
IterableDataset(validation_instances),
batch_size=self.batch_size,
pin_memory=True,
persistent_workers=kwargs.get("num_workers", 0) > 0,
**kwargs,
)
def create_predictor(
self,
transformation: Transformation,
module: TransformerLightningModule,
) -> PyTorchPredictor:
prediction_splitter = self._create_instance_splitter(module, "test")
return PyTorchPredictor(
input_transform=transformation + prediction_splitter,
input_names=PREDICTION_INPUT_NAMES,
prediction_net=module.model,
batch_size=self.batch_size,
prediction_length=self.prediction_length,
device=torch.device('cuda' if torch.cuda.is_available() else 'cpu'),
)
def create_lightning_module(self) -> TransformerLightningModule:
model = TransformerModel(
freq=self.freq,
context_length=self.context_length,
prediction_length=self.prediction_length,
num_feat_dynamic_real=1 + self.num_feat_dynamic_real + len(self.time_features),
num_feat_static_real=max(1, self.num_feat_static_real),
num_feat_static_cat=max(1, self.num_feat_static_cat),
cardinality=self.cardinality,
embedding_dimension=self.embedding_dimension,
# transformer arguments
nhead=self.nhead,
num_encoder_layers=self.num_encoder_layers,
num_decoder_layers=self.num_decoder_layers,
activation=self.activation,
dropout=self.dropout,
dim_feedforward=self.dim_feedforward,
# univariate input
input_size=self.input_size,
distr_output=self.distr_output,
lags_seq=self.lags_seq,
scaling=self.scaling,
num_parallel_samples=self.num_parallel_samples,
)
return TransformerLightningModule(model=model, loss=self.loss)In [285]:
estimator = TransformerEstimator(
prediction_length=24,
context_length=24*10,
lags_seq=[1,2,3,4,5,6,7,24,30],
time_features=[
MinuteOfHour(),
HourOfDay(),
DayOfWeek(),
DayOfMonth(),
DayOfYear(),
],
nhead=2,
num_encoder_layers=6,
num_decoder_layers=2,
dim_feedforward=32,
activation="gelu",
scaling=True,
batch_size=256,
num_batches_per_epoch=100,
trainer_kwargs=dict(max_epochs=150, accelerator='gpu',),
)In [286]:
predictor = estimator.train(
training_data=train_ds,
validation_data=val_ds,
num_workers=5,
shuffle_buffer_length=2048,
ckpt_path='/mnt/scratch/kashif/pytorch-transformer-ts/transformer/lightning_logs/version_200/checkpoints/epoch=87-step=8800.ckpt',
)/home/kashif/.env/pytorch/lib/python3.8/site-packages/pytorch_lightning/utilities/parsing.py:261: UserWarning: Attribute 'model' is an instance of `nn.Module` and is already saved during checkpointing. It is recommended to ignore them using `self.save_hyperparameters(ignore=['model'])`. rank_zero_warn( /home/kashif/.env/pytorch/lib/python3.8/site-packages/pytorch_lightning/utilities/parsing.py:261: UserWarning: Attribute 'loss' is an instance of `nn.Module` and is already saved during checkpointing. It is recommended to ignore them using `self.save_hyperparameters(ignore=['loss'])`. rank_zero_warn( GPU available: True, used: True TPU available: False, using: 0 TPU cores IPU available: False, using: 0 IPUs HPU available: False, using: 0 HPUs Restoring states from the checkpoint path at /mnt/scratch/kashif/pytorch-transformer-ts/transformer/lightning_logs/version_200/checkpoints/epoch=87-step=8800.ckpt /home/kashif/.env/pytorch/lib/python3.8/site-packages/pytorch_lightning/callbacks/model_checkpoint.py:342: UserWarning: The dirpath has changed from '/mnt/scratch/kashif/pytorch-transformer-ts/transformer/lightning_logs/version_200/checkpoints' to '/mnt/scratch/kashif/pytorch-transformer-ts/transformer/lightning_logs/version_201/checkpoints', therefore `best_model_score`, `kth_best_model_path`, `kth_value`, `last_model_path` and `best_k_models` won't be reloaded. Only `best_model_path` will be reloaded. warnings.warn( LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0] | Name | Type | Params ------------------------------------------------ 0 | model | TransformerModel | 24.1 K 1 | loss | NegativeLogLikelihood | 0 ------------------------------------------------ 24.1 K Trainable params 0 Non-trainable params 24.1 K Total params 0.096 Total estimated model params size (MB) Restored all states from the checkpoint file at /mnt/scratch/kashif/pytorch-transformer-ts/transformer/lightning_logs/version_200/checkpoints/epoch=87-step=8800.ckpt
Sanity Checking: 0it [00:00, ?it/s]
Training: 100it [00:00, ?it/s]
Validation: 0it [00:00, ?it/s]
Epoch 88, global step 8900: 'val_loss' reached 1.10372 (best 1.10372), saving model to '/mnt/scratch/kashif/pytorch-transformer-ts/transformer/lightning_logs/version_201/checkpoints/epoch=88-step=8900.ckpt' as top 1
Validation: 0it [00:00, ?it/s]
Epoch 89, global step 9000: 'val_loss' reached 1.09097 (best 1.09097), saving model to '/mnt/scratch/kashif/pytorch-transformer-ts/transformer/lightning_logs/version_201/checkpoints/epoch=89-step=9000.ckpt' as top 1
Validation: 0it [00:00, ?it/s]
Epoch 90, global step 9100: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 91, global step 9200: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 92, global step 9300: 'val_loss' reached 1.08504 (best 1.08504), saving model to '/mnt/scratch/kashif/pytorch-transformer-ts/transformer/lightning_logs/version_201/checkpoints/epoch=92-step=9300.ckpt' as top 1
Validation: 0it [00:00, ?it/s]
Epoch 93, global step 9400: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 94, global step 9500: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 95, global step 9600: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 96, global step 9700: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 97, global step 9800: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 98, global step 9900: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 99, global step 10000: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 100, global step 10100: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 101, global step 10200: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 102, global step 10300: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 103, global step 10400: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 104, global step 10500: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 105, global step 10600: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 106, global step 10700: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 107, global step 10800: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 108, global step 10900: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 109, global step 11000: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 110, global step 11100: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 111, global step 11200: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 112, global step 11300: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 113, global step 11400: 'val_loss' was not in top 1
Exception ignored in: <function _MultiProcessingDataLoaderIter.__del__ at 0x7fb7e6ed1820>
Traceback (most recent call last):
File "/home/kashif/.env/pytorch/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 1471, in __del__
Exception ignored in: <function _MultiProcessingDataLoaderIter.__del__ at 0x7fb7e6ed1820>self._shutdown_workers()
Traceback (most recent call last):
File "/home/kashif/.env/pytorch/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 1471, in __del__
self._shutdown_workers() File "/home/kashif/.env/pytorch/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 1454, in _shutdown_workers
File "/home/kashif/.env/pytorch/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 1454, in _shutdown_workers
if w.is_alive():if w.is_alive():
File "/usr/lib/python3.8/multiprocessing/process.py", line 160, in is_alive
File "/usr/lib/python3.8/multiprocessing/process.py", line 160, in is_alive
assert self._parent_pid == os.getpid(), 'can only test a child process'
assert self._parent_pid == os.getpid(), 'can only test a child process'AssertionError
AssertionError: can only test a child process
: can only test a child process
Exception ignored in: <function _MultiProcessingDataLoaderIter.__del__ at 0x7fb7e6ed1820>
Traceback (most recent call last):
File "/home/kashif/.env/pytorch/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 1471, in __del__
self._shutdown_workers()
File "/home/kashif/.env/pytorch/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 1454, in _shutdown_workers
if w.is_alive():
File "/usr/lib/python3.8/multiprocessing/process.py", line 160, in is_alive
assert self._parent_pid == os.getpid(), 'can only test a child process'
AssertionError: can only test a child process
Exception ignored in: <function _MultiProcessingDataLoaderIter.__del__ at 0x7fb7e6ed1820>
Traceback (most recent call last):
File "/home/kashif/.env/pytorch/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 1471, in __del__
self._shutdown_workers()
File "/home/kashif/.env/pytorch/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 1454, in _shutdown_workers
if w.is_alive():
File "/usr/lib/python3.8/multiprocessing/process.py", line 160, in is_alive
assert self._parent_pid == os.getpid(), 'can only test a child process'
AssertionError: can only test a child process
Exception ignored in: <function _MultiProcessingDataLoaderIter.__del__ at 0x7fb7e6ed1820>
Traceback (most recent call last):
File "/home/kashif/.env/pytorch/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 1471, in __del__
self._shutdown_workers()
File "/home/kashif/.env/pytorch/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 1454, in _shutdown_workers
if w.is_alive():
File "/usr/lib/python3.8/multiprocessing/process.py", line 160, in is_alive
assert self._parent_pid == os.getpid(), 'can only test a child process'
AssertionError: can only test a child process
Validation: 0it [00:00, ?it/s]
Epoch 114, global step 11500: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 115, global step 11600: 'val_loss' reached 1.07203 (best 1.07203), saving model to '/mnt/scratch/kashif/pytorch-transformer-ts/transformer/lightning_logs/version_201/checkpoints/epoch=115-step=11600.ckpt' as top 1
Validation: 0it [00:00, ?it/s]
Epoch 116, global step 11700: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 117, global step 11800: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 118, global step 11900: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 119, global step 12000: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 120, global step 12100: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 121, global step 12200: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 122, global step 12300: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 123, global step 12400: 'val_loss' reached 1.05691 (best 1.05691), saving model to '/mnt/scratch/kashif/pytorch-transformer-ts/transformer/lightning_logs/version_201/checkpoints/epoch=123-step=12400.ckpt' as top 1
Validation: 0it [00:00, ?it/s]
Epoch 124, global step 12500: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 125, global step 12600: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 126, global step 12700: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 127, global step 12800: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 128, global step 12900: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 129, global step 13000: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 130, global step 13100: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 131, global step 13200: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 132, global step 13300: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 133, global step 13400: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 134, global step 13500: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 135, global step 13600: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 136, global step 13700: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 137, global step 13800: 'val_loss' reached 1.04551 (best 1.04551), saving model to '/mnt/scratch/kashif/pytorch-transformer-ts/transformer/lightning_logs/version_201/checkpoints/epoch=137-step=13800.ckpt' as top 1
Validation: 0it [00:00, ?it/s]
Epoch 138, global step 13900: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 139, global step 14000: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 140, global step 14100: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 141, global step 14200: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 142, global step 14300: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 143, global step 14400: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 144, global step 14500: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 145, global step 14600: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 146, global step 14700: 'val_loss' reached 1.03886 (best 1.03886), saving model to '/mnt/scratch/kashif/pytorch-transformer-ts/transformer/lightning_logs/version_201/checkpoints/epoch=146-step=14700.ckpt' as top 1
Validation: 0it [00:00, ?it/s]
Epoch 147, global step 14800: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 148, global step 14900: 'val_loss' was not in top 1
Validation: 0it [00:00, ?it/s]
Epoch 149, global step 15000: 'val_loss' was not in top 1
In [287]:
forecast_it, ts_it = make_evaluation_predictions(
dataset=test_ds_4,
predictor=predictor
)In [288]:
forecasts = list(forecast_it)In [289]:
tss = list(ts_it)In [290]:
evaluator = Evaluator()In [291]:
agg_metrics, ts_metrics = evaluator(iter(tss), iter(forecasts))Running evaluation: 5560it [00:00, 8310.84it/s] /home/kashif/gluon-ts-PR/src/gluonts/evaluation/_base.py:352: RuntimeWarning: divide by zero encountered in float_scalars metrics["ND"] = cast(float, metrics["abs_error"]) / cast( /home/kashif/gluon-ts-PR/src/gluonts/evaluation/_base.py:352: RuntimeWarning: divide by zero encountered in float_scalars metrics["ND"] = cast(float, metrics["abs_error"]) / cast( /home/kashif/gluon-ts-PR/src/gluonts/evaluation/_base.py:352: RuntimeWarning: divide by zero encountered in float_scalars metrics["ND"] = cast(float, metrics["abs_error"]) / cast( /home/kashif/gluon-ts-PR/src/gluonts/evaluation/_base.py:352: RuntimeWarning: divide by zero encountered in float_scalars metrics["ND"] = cast(float, metrics["abs_error"]) / cast( /home/kashif/gluon-ts-PR/src/gluonts/evaluation/_base.py:352: RuntimeWarning: divide by zero encountered in float_scalars metrics["ND"] = cast(float, metrics["abs_error"]) / cast( /home/kashif/gluon-ts-PR/src/gluonts/evaluation/_base.py:352: RuntimeWarning: divide by zero encountered in float_scalars metrics["ND"] = cast(float, metrics["abs_error"]) / cast( /home/kashif/gluon-ts-PR/src/gluonts/evaluation/_base.py:352: RuntimeWarning: divide by zero encountered in float_scalars metrics["ND"] = cast(float, metrics["abs_error"]) / cast( /home/kashif/gluon-ts-PR/src/gluonts/evaluation/_base.py:352: RuntimeWarning: divide by zero encountered in float_scalars metrics["ND"] = cast(float, metrics["abs_error"]) / cast( /home/kashif/gluon-ts-PR/src/gluonts/evaluation/_base.py:352: RuntimeWarning: divide by zero encountered in float_scalars metrics["ND"] = cast(float, metrics["abs_error"]) / cast( /home/kashif/gluon-ts-PR/src/gluonts/evaluation/_base.py:352: RuntimeWarning: divide by zero encountered in float_scalars metrics["ND"] = cast(float, metrics["abs_error"]) / cast( /home/kashif/gluon-ts-PR/src/gluonts/evaluation/_base.py:352: RuntimeWarning: divide by zero encountered in float_scalars metrics["ND"] = cast(float, metrics["abs_error"]) / cast( /home/kashif/gluon-ts-PR/src/gluonts/evaluation/_base.py:352: RuntimeWarning: divide by zero encountered in float_scalars metrics["ND"] = cast(float, metrics["abs_error"]) / cast( /home/kashif/gluon-ts-PR/src/gluonts/evaluation/_base.py:352: RuntimeWarning: divide by zero encountered in float_scalars metrics["ND"] = cast(float, metrics["abs_error"]) / cast( /home/kashif/gluon-ts-PR/src/gluonts/evaluation/_base.py:352: RuntimeWarning: divide by zero encountered in float_scalars metrics["ND"] = cast(float, metrics["abs_error"]) / cast( /home/kashif/gluon-ts-PR/src/gluonts/evaluation/_base.py:352: RuntimeWarning: divide by zero encountered in float_scalars metrics["ND"] = cast(float, metrics["abs_error"]) / cast( /home/kashif/gluon-ts-PR/src/gluonts/evaluation/metrics.py:105: RuntimeWarning: divide by zero encountered in double_scalars return np.mean(np.abs(target - forecast)) / seasonal_error /home/kashif/gluon-ts-PR/src/gluonts/evaluation/metrics.py:154: RuntimeWarning: divide by zero encountered in double_scalars return numerator / seasonal_error /home/kashif/gluon-ts-PR/src/gluonts/evaluation/_base.py:252: FutureWarning: Could not cast to float64, falling back to object. This behavior is deprecated. In a future version, when a dtype is passed to 'DataFrame', either all columns will be cast to that dtype, or a TypeError will be raised. metrics_per_ts = pd.DataFrame(rows, dtype=np.float64) /home/kashif/.env/pytorch/lib/python3.8/site-packages/pandas/core/construction.py:784: UserWarning: Warning: converting a masked element to nan. subarr = np.array(arr, dtype=dtype, copy=copy)
In [271]:
# ds 1
agg_metricsOut [271]:
{'MSE': 0.0008509528534418956,
'abs_error': 263.7405581064522,
'abs_target_sum': 926.2648004814982,
'abs_target_mean': 0.04477304720038177,
'seasonal_error': 0.01542953681305969,
'MASE': 0.8322391929513296,
'MAPE': 0.44503156124074483,
'sMAPE': 0.3769144580540321,
'MSIS': 15.835981866315985,
'QuantileLoss[0.1]': 108.89879824975941,
'Coverage[0.1]': 0.1377126836813612,
'QuantileLoss[0.2]': 162.7133320433968,
'Coverage[0.2]': 0.25435034802784223,
'QuantileLoss[0.3]': 204.02928320299958,
'Coverage[0.3]': 0.361900618716164,
'QuantileLoss[0.4]': 237.48500103205325,
'Coverage[0.4]': 0.46263534416086616,
'QuantileLoss[0.5]': 263.74055804964587,
'Coverage[0.5]': 0.5542343387470997,
'QuantileLoss[0.6]': 282.4404990474631,
'Coverage[0.6]': 0.6274651972157773,
'QuantileLoss[0.7]': 293.3158605407789,
'Coverage[0.7]': 0.703692962103635,
'QuantileLoss[0.8]': 293.75111032472194,
'Coverage[0.8]': 0.7752320185614849,
'QuantileLoss[0.9]': 275.7946062145103,
'Coverage[0.9]': 0.8459010054137663,
'RMSE': 0.029171096198838595,
'NRMSE': 0.6515325183984766,
'ND': 0.2847355939352899,
'wQuantileLoss[0.1]': 0.11756767416094273,
'wQuantileLoss[0.2]': 0.17566610753082043,
'wQuantileLoss[0.3]': 0.22027101007934174,
'wQuantileLoss[0.4]': 0.2563899663558433,
'wQuantileLoss[0.5]': 0.2847355938739615,
'wQuantileLoss[0.6]': 0.30492414145570756,
'wQuantileLoss[0.7]': 0.3166652348100728,
'wQuantileLoss[0.8]': 0.3171351326014137,
'wQuantileLoss[0.9]': 0.2977492031124842,
'mean_absolute_QuantileLoss': 235.7965609672588,
'mean_wQuantileLoss': 0.2545671182200653,
'MAE_Coverage': 0.04231760763083271,
'OWA': nan}In [272]:
plt.figure(figsize=(20, 15))
date_formater = mdates.DateFormatter('%b, %d')
plt.rcParams.update({'font.size': 15})
for idx, (forecast, ts) in islice(enumerate(zip(forecasts, tss)),9):
ax = plt.subplot(3, 3, idx+1)
ts[-4 * 24:].plot(ax=ax, label="target", )
forecast.plot( color='g')
plt.xticks(rotation=60)
plt.gcf().tight_layout()
plt.legend()
plt.show()
In [219]:
# ds 2
agg_metricsOut [219]:
{'MSE': 23.940462105773744,
'abs_error': 169621.5951450821,
'abs_target_sum': 795403.1574395448,
'abs_target_mean': 11.010564194899567,
'seasonal_error': 2.6968875399609384,
'MASE': 1.0617574055775507,
'MAPE': 0.7603387070885905,
'sMAPE': 0.7029547654975161,
'MSIS': 13.830274620382172,
'QuantileLoss[0.1]': 74253.39198404923,
'Coverage[0.1]': 0.0724390919158361,
'QuantileLoss[0.2]': 109880.7456878802,
'Coverage[0.2]': 0.17782392026578073,
'QuantileLoss[0.3]': 135075.30077576044,
'Coverage[0.3]': 0.2966362126245847,
'QuantileLoss[0.4]': 154390.34527562038,
'Coverage[0.4]': 0.4288759689922481,
'QuantileLoss[0.5]': 169621.5953532845,
'Coverage[0.5]': 0.5738372093023256,
'QuantileLoss[0.6]': 177621.70033670584,
'Coverage[0.6]': 0.7098560354374307,
'QuantileLoss[0.7]': 181507.8876448301,
'Coverage[0.7]': 0.8369462901439645,
'QuantileLoss[0.8]': 172879.50066934875,
'Coverage[0.8]': 0.8887181616832779,
'QuantileLoss[0.9]': 142303.91879960973,
'Coverage[0.9]': 0.9146733111849392,
'RMSE': 4.892899151400297,
'NRMSE': 0.44438223734863996,
'ND': 0.21325235329855266,
'wQuantileLoss[0.1]': 0.09335315215880685,
'wQuantileLoss[0.2]': 0.1381447190146863,
'wQuantileLoss[0.3]': 0.16981992026606574,
'wQuantileLoss[0.4]': 0.1941032592485716,
'wQuantileLoss[0.5]': 0.21325235356030972,
'wQuantileLoss[0.6]': 0.22331027815942017,
'wQuantileLoss[0.7]': 0.22819608640870367,
'wQuantileLoss[0.8]': 0.2173482705623891,
'wQuantileLoss[0.9]': 0.17890791288495186,
'mean_absolute_QuantileLoss': 146392.709614121,
'mean_wQuantileLoss': 0.1840484391404339,
'MAE_Coverage': 0.05622308354866493,
'OWA': nan}In [113]:
plt.figure(figsize=(20, 15))
date_formater = mdates.DateFormatter('%b, %d')
plt.rcParams.update({'font.size': 15})
for idx, (forecast, ts) in islice(enumerate(zip(forecasts, tss)),9):
ax = plt.subplot(3, 3, idx+1)
ts[-4 * 24:].plot(ax=ax, label="target", )
forecast.plot( color='g')
plt.xticks(rotation=60)
plt.gcf().tight_layout()
plt.legend()
plt.show()
In [264]:
# ds 3
agg_metricsOut [264]:
{'MSE': 1115704921.4488733,
'abs_error': 41362382.66113281,
'abs_target_sum': 166958474.8984375,
'abs_target_mean': 19007.112351825763,
'seasonal_error': 1542.9874709095218,
'MASE': 2.359203822787185,
'MAPE': 0.31409288595391543,
'sMAPE': 0.2854564158832875,
'MSIS': 18.364880119159142,
'QuantileLoss[0.1]': 14798822.461064912,
'Coverage[0.1]': 0.08413023679417121,
'QuantileLoss[0.2]': 24240270.733774565,
'Coverage[0.2]': 0.20366575591985428,
'QuantileLoss[0.3]': 31439739.301414296,
'Coverage[0.3]': 0.31079234972677594,
'QuantileLoss[0.4]': 37215963.02638474,
'Coverage[0.4]': 0.4114298724954464,
'QuantileLoss[0.5]': 41362383.03600502,
'Coverage[0.5]': 0.5058060109289617,
'QuantileLoss[0.6]': 44064631.575115964,
'Coverage[0.6]': 0.5835610200364298,
'QuantileLoss[0.7]': 44516753.69809838,
'Coverage[0.7]': 0.6574453551912568,
'QuantileLoss[0.8]': 41446041.34827957,
'Coverage[0.8]': 0.7421448087431693,
'QuantileLoss[0.9]': 32230755.70360107,
'Coverage[0.9]': 0.8435792349726776,
'RMSE': 33402.169412313226,
'NRMSE': 1.757351079639655,
'ND': 0.2477405395940156,
'wQuantileLoss[0.1]': 0.08863774342732338,
'wQuantileLoss[0.2]': 0.1451874230914014,
'wQuantileLoss[0.3]': 0.18830873557355746,
'wQuantileLoss[0.4]': 0.2229055041921267,
'wQuantileLoss[0.5]': 0.24774054183931765,
'wQuantileLoss[0.6]': 0.26392569530789567,
'wQuantileLoss[0.7]': 0.26663368676060534,
'wQuantileLoss[0.8]': 0.2482416144103593,
'wQuantileLoss[0.9]': 0.1930465388067743,
'mean_absolute_QuantileLoss': 34590595.65374872,
'mean_wQuantileLoss': 0.20718083148992902,
'MAE_Coverage': 0.024537037037037055,
'OWA': nan}In [265]:
plt.figure(figsize=(20, 15))
date_formater = mdates.DateFormatter('%b, %d')
plt.rcParams.update({'font.size': 15})
for idx, (forecast, ts) in islice(enumerate(zip(forecasts, tss)),9):
ax = plt.subplot(3, 3, idx+1)
ts[-4 * 24:].plot(ax=ax, label="target", )
forecast.plot( color='g')
plt.xticks(rotation=60)
plt.gcf().tight_layout()
plt.legend()
plt.show()
In [292]:
# ds 4
agg_metricsOut [292]:
{'MSE': 0.07938089170859768,
'abs_error': 20232.472963708704,
'abs_target_sum': 34828.328015925596,
'abs_target_mean': 0.261003657193687,
'seasonal_error': 0.12278935134470352,
'MASE': inf,
'MAPE': 0.5928652279373905,
'sMAPE': 0.6665428828939403,
'MSIS': inf,
'QuantileLoss[0.1]': 5670.534526125589,
'Coverage[0.1]': 0.02965377697841726,
'QuantileLoss[0.2]': 10227.543230812102,
'Coverage[0.2]': 0.09668764988009591,
'QuantileLoss[0.3]': 14223.804547135831,
'Coverage[0.3]': 0.16566247002398082,
'QuantileLoss[0.4]': 17593.629550595586,
'Coverage[0.4]': 0.2318195443645084,
'QuantileLoss[0.5]': 20232.472946430178,
'Coverage[0.5]': 0.29769184652278174,
'QuantileLoss[0.6]': 22116.110551246515,
'Coverage[0.6]': 0.3615857314148681,
'QuantileLoss[0.7]': 22682.888253401314,
'Coverage[0.7]': 0.4422661870503597,
'QuantileLoss[0.8]': 21633.46849227681,
'Coverage[0.8]': 0.5449790167865708,
'QuantileLoss[0.9]': 17937.66250043737,
'Coverage[0.9]': 0.6864583333333334,
'RMSE': 0.2817461476375457,
'NRMSE': 1.0794720298821943,
'ND': 0.5809200187404118,
'wQuantileLoss[0.1]': 0.1628138601293947,
'wQuantileLoss[0.2]': 0.29365587765612683,
'wQuantileLoss[0.3]': 0.4083975705245413,
'wQuantileLoss[0.4]': 0.5051528612728904,
'wQuantileLoss[0.5]': 0.5809200182443062,
'wQuantileLoss[0.6]': 0.6350035103934276,
'wQuantileLoss[0.7]': 0.6512769789876028,
'wQuantileLoss[0.8]': 0.6211457662390424,
'wQuantileLoss[0.9]': 0.5150308246848714,
'mean_absolute_QuantileLoss': 16924.23495538459,
'mean_wQuantileLoss': 0.48593302979246705,
'MAE_Coverage': 0.1825772715161204,
'OWA': nan}In [293]:
plt.figure(figsize=(20, 15))
date_formater = mdates.DateFormatter('%b, %d')
plt.rcParams.update({'font.size': 15})
for idx, (forecast, ts) in islice(enumerate(zip(forecasts, tss)),9):
ax = plt.subplot(3, 3, idx+1)
ts[-4 * 24:].plot(ax=ax, label="target", )
forecast.plot( color='g')
plt.xticks(rotation=60)
plt.gcf().tight_layout()
plt.legend()
plt.show()
In [273]:
def plot_prob_forecasts(ts_entry, forecast_entry):
plot_length = 70
prediction_intervals = (50.0, 90.0)
legend = ["observations", "median prediction"] + [f"{k}% prediction interval" for k in prediction_intervals][::-1]
fig, ax = plt.subplots(1, 1, figsize=(10, 7))
ts_entry[-plot_length:].plot(ax=ax) # plot the time series
forecast_entry.plot(prediction_intervals=prediction_intervals, color='g')
plt.grid(which="both")
plt.legend(legend, loc="best")
plt.show()In [300]:
index = 2222
plot_prob_forecasts(tss[index], forecasts[index])
In [ ]: