PyTorch Forecasting - NBEATS, DeepAR#
PyTorch Forecasting is a package/repository that provides convenient implementations of several leading deep learning-based forecasting models, namely Temporal Fusion Transformers, N-BEATS, and DeepAR. PyTorch Forecasting is built using PyTorch Lightning, making it easier to train in multi-GPU compute environments, out-of-the-box.
Note for Colab users: Run the following cell to install PyTorch Forecasting. After installation completes, you will likely need to restart the Colab runtime. If this is the case, a button RESTART RUNTIME will appear at the bottom of the next cell’s output.
if 'google.colab' in str(get_ipython()):
!pip install pytorch-forecasting==0.10.3
!pip install pytorch-lightning==1.5.9
!pip install torch==1.11.0 torchaudio==0.11.0 torchtext==0.6.0 torchvision==0.12.0
import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
import seaborn as sns
import torch
import pytorch_lightning as pl
from pytorch_lightning.callbacks import EarlyStopping, LearningRateMonitor, Callback
from pytorch_forecasting import TimeSeriesDataSet, NBeats, Baseline, DeepAR, GroupNormalizer, MultiNormalizer, EncoderNormalizer
from pytorch_forecasting.data import NaNLabelEncoder
from pytorch_forecasting.metrics import MAPE, MAE, MASE, RMSE
from pytorch_forecasting.metrics import NormalDistributionLoss
class EpochCallback(Callback):
def on_validation_epoch_end(self, trainer, module):
if 'val_MASE' in trainer.callback_metrics:
print("Validation MASE", trainer.callback_metrics['val_MASE'])
if 'google.colab' in str(get_ipython()):
from google.colab import drive
drive.mount('/content/drive')
Data Loading#
We will continue with the same example dataset and task as the previous demo.
# data_filename = "/content/drive/MyDrive/bootcamp_datasets/boc_exchange/dataset.csv"
data_filename = "/ssd003/projects/forecasting_bootcamp/bootcamp_datasets/boc_exchange/dataset.csv"
data_df = pd.read_csv(data_filename, index_col=0)
data_df.index = pd.to_datetime(data_df.index)
data_df = data_df.reset_index().rename({'index':'date'}, axis=1)
data_df
| date | AUD_CLOSE | DKK_CLOSE | EUR_CLOSE | HKD_CLOSE | JPY_CLOSE | MXN_CLOSE | NZD_CLOSE | NOK_CLOSE | SEK_CLOSE | CHF_CLOSE | GBP_CLOSE | USD_CLOSE | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 2007-05-01 | 0.9200 | 0.2000 | 1.5100 | 0.141970 | 0.009271 | 0.10000 | 0.8200 | 0.1860 | 0.1700 | 0.9100 | 2.2199 | 1.1105 |
| 1 | 2007-05-02 | 0.9100 | 0.2000 | 1.5100 | 0.141740 | 0.009232 | 0.10000 | 0.8200 | 0.1854 | 0.1700 | 0.9100 | 2.2055 | 1.1087 |
| 2 | 2007-05-03 | 0.9100 | 0.2000 | 1.5000 | 0.141496 | 0.009190 | 0.10000 | 0.8100 | 0.1847 | 0.1600 | 0.9100 | 2.1999 | 1.1066 |
| 3 | 2007-05-04 | 0.9100 | 0.2000 | 1.5100 | 0.141616 | 0.009218 | 0.10000 | 0.8100 | 0.1854 | 0.1600 | 0.9100 | 2.2075 | 1.1075 |
| 4 | 2007-05-05 | 0.9100 | 0.2000 | 1.5100 | 0.141616 | 0.009218 | 0.10000 | 0.8100 | 0.1854 | 0.1600 | 0.9100 | 2.2075 | 1.1075 |
| ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... |
| 3646 | 2017-04-24 | 1.0218 | 0.1974 | 1.4684 | 0.173684 | 0.012310 | 0.07215 | 0.9477 | 0.1579 | 0.1525 | 1.3568 | 1.7280 | 1.3511 |
| 3647 | 2017-04-25 | 1.0224 | 0.1994 | 1.4837 | 0.174374 | 0.012210 | 0.07183 | 0.9426 | 0.1586 | 0.1548 | 1.3661 | 1.7407 | 1.3565 |
| 3648 | 2017-04-26 | 1.0176 | 0.1995 | 1.4847 | 0.174965 | 0.012260 | 0.07098 | 0.9382 | 0.1585 | 0.1554 | 1.3707 | 1.7493 | 1.3612 |
| 3649 | 2017-04-27 | 1.0176 | 0.1992 | 1.4815 | 0.175103 | 0.012250 | 0.07151 | 0.9369 | 0.1590 | 0.1543 | 1.3704 | 1.7584 | 1.3624 |
| 3650 | 2017-04-28 | 1.0222 | 0.1999 | 1.4870 | 0.175485 | 0.012250 | 0.07253 | 0.9373 | 0.1590 | 0.1541 | 1.3719 | 1.7679 | 1.3650 |
3651 rows × 13 columns
Data Splitting#
We immediately split the data into training and test sets. We can later split the training set into training and validation.
train_size = 0.8
train_df = data_df.iloc[:int(len(data_df)*train_size)]
test_df = data_df.iloc[int(len(data_df)*train_size):]
Data Formatting#
PyTorch Forecasting expects data to be formatted using its own TimeSeriesDataSet objects. Building a TimeSeriesDataSet begins with a Pandas DataFrame, and like with Prophet and NeuralProphet, we need to add certain custom columns.
For this demo, we are once again specifying a lead time of 30 days and we will withhold the last 20% of data for testing. In the code below, we are very careful to ensure that when training and validating the model, it does not have access to the withheld data.
Time Series Data Formatting#
PyTorch Forecasting models can accomodate datasets consisting of multiple, coincident time series in several ways. As per the documentation, a combination of group_id and time_idx identify a sample in the data, and that if we have only one time series, to set group_id to a constant.
time_idx is an integer column denoting the time index. This, as opposed to the date column, is used to determine the temporal sequence of samples.
lag_time = 60
lead_time = 30
train_df = train_df.reset_index().rename({'index':'time_idx'}, axis=1)
test_df = test_df.reset_index().rename({'index':'time_idx'}, axis=1)
train_df['group_ids'] = 0
test_df['group_ids'] = 0
train_df
| time_idx | date | AUD_CLOSE | DKK_CLOSE | EUR_CLOSE | HKD_CLOSE | JPY_CLOSE | MXN_CLOSE | NZD_CLOSE | NOK_CLOSE | SEK_CLOSE | CHF_CLOSE | GBP_CLOSE | USD_CLOSE | group_ids | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 0 | 2007-05-01 | 0.9200 | 0.2000 | 1.5100 | 0.141970 | 0.009271 | 0.10000 | 0.8200 | 0.1860 | 0.1700 | 0.9100 | 2.2199 | 1.1105 | 0 |
| 1 | 1 | 2007-05-02 | 0.9100 | 0.2000 | 1.5100 | 0.141740 | 0.009232 | 0.10000 | 0.8200 | 0.1854 | 0.1700 | 0.9100 | 2.2055 | 1.1087 | 0 |
| 2 | 2 | 2007-05-03 | 0.9100 | 0.2000 | 1.5000 | 0.141496 | 0.009190 | 0.10000 | 0.8100 | 0.1847 | 0.1600 | 0.9100 | 2.1999 | 1.1066 | 0 |
| 3 | 3 | 2007-05-04 | 0.9100 | 0.2000 | 1.5100 | 0.141616 | 0.009218 | 0.10000 | 0.8100 | 0.1854 | 0.1600 | 0.9100 | 2.2075 | 1.1075 | 0 |
| 4 | 4 | 2007-05-05 | 0.9100 | 0.2000 | 1.5100 | 0.141616 | 0.009218 | 0.10000 | 0.8100 | 0.1854 | 0.1600 | 0.9100 | 2.2075 | 1.1075 | 0 |
| ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... |
| 2915 | 2915 | 2015-04-24 | 0.9524 | 0.1773 | 1.3226 | 0.157028 | 0.010240 | 0.07916 | 0.9250 | 0.1564 | 0.1409 | 1.2761 | 1.8473 | 1.2170 | 0 |
| 2916 | 2916 | 2015-04-25 | 0.9524 | 0.1773 | 1.3226 | 0.157028 | 0.010240 | 0.07916 | 0.9250 | 0.1564 | 0.1409 | 1.2761 | 1.8473 | 1.2170 | 0 |
| 2917 | 2917 | 2015-04-26 | 0.9524 | 0.1773 | 1.3226 | 0.157028 | 0.010240 | 0.07916 | 0.9250 | 0.1564 | 0.1409 | 1.2761 | 1.8473 | 1.2170 | 0 |
| 2918 | 2918 | 2015-04-27 | 0.9500 | 0.1766 | 1.3177 | 0.156142 | 0.010160 | 0.07882 | 0.9249 | 0.1567 | 0.1408 | 1.2655 | 1.8426 | 1.2101 | 0 |
| 2919 | 2919 | 2015-04-28 | 0.9642 | 0.1769 | 1.3202 | 0.155221 | 0.010120 | 0.07881 | 0.9288 | 0.1573 | 0.1409 | 1.2582 | 1.8437 | 1.2030 | 0 |
2920 rows × 15 columns
Preparing a TimeSeriesDataSet for N-BEATS#
N-BEATS is a univariate forecasting method. As such, it can only process one variable at a time.
train_df_nbeats = train_df.melt(id_vars=['time_idx', 'date'], value_vars=data_df.columns, var_name='group_ids')
train_df_nbeats.head()
| time_idx | date | group_ids | value | |
|---|---|---|---|---|
| 0 | 0 | 2007-05-01 | AUD_CLOSE | 0.92 |
| 1 | 1 | 2007-05-02 | AUD_CLOSE | 0.91 |
| 2 | 2 | 2007-05-03 | AUD_CLOSE | 0.91 |
| 3 | 3 | 2007-05-04 | AUD_CLOSE | 0.91 |
| 4 | 4 | 2007-05-05 | AUD_CLOSE | 0.91 |
train_df_nbeats.tail()
| time_idx | date | group_ids | value | |
|---|---|---|---|---|
| 35035 | 2915 | 2015-04-24 | USD_CLOSE | 1.2170 |
| 35036 | 2916 | 2015-04-25 | USD_CLOSE | 1.2170 |
| 35037 | 2917 | 2015-04-26 | USD_CLOSE | 1.2170 |
| 35038 | 2918 | 2015-04-27 | USD_CLOSE | 1.2101 |
| 35039 | 2919 | 2015-04-28 | USD_CLOSE | 1.2030 |
The following code defines the training/validation split and important parts of the prediction task - the lag time and lead time. Here we are specifying that 25% of the training data should be used for validation. In total, this results in a 60/20/20 train/valid/test split (since 25% of 80% is 20% of the initial 100%).
# define dataset
valid_size = 0.25
max_encoder_length = lag_time
max_prediction_length = lead_time
training_cutoff = train_df_nbeats["time_idx"][:int((1 - valid_size)*train_df_nbeats["time_idx"].max())].values[-1]
context_length = max_encoder_length
prediction_length = max_prediction_length
Since NBEATS is actually a univariate forecasting model, it makes sense that pytorch-forecasting does not support specifying explanatory variables as time_varying_unknown_reals. Earlier, we used the Pandas function melt to reshape the data into a single value column that is uniquely indexed by pairs of (time_idx, group_ids).
training = TimeSeriesDataSet(
train_df_nbeats.loc[train_df_nbeats.time_idx <= training_cutoff], # Applying the training cutoff
time_idx="time_idx",
target="value",
categorical_encoders={"group_ids": NaNLabelEncoder().fit(train_df_nbeats.group_ids)},
group_ids=['group_ids'],
max_encoder_length=context_length,
max_prediction_length=max_prediction_length,
time_varying_unknown_reals=["value"],
)
validation = TimeSeriesDataSet.from_dataset(training, train_df_nbeats, min_prediction_idx=training_cutoff + 1)
batch_size = 128
train_dataloader = training.to_dataloader(train=True, batch_size=batch_size, num_workers=0)
val_dataloader = validation.to_dataloader(train=False, batch_size=batch_size, num_workers=0)
Training the model#
In our first effort to train the N-BEATS model, we adopt the default configuration suggested by the pytorch-forecasting tutorials.
pl.seed_everything(42)
early_stop_callback = EarlyStopping(monitor="val_loss", min_delta=1e-4, patience=5, verbose=False, mode="min")
epoch_callback = EpochCallback()
trainer = pl.Trainer(
max_epochs=100,
gpus=1,
weights_summary="top",
gradient_clip_val=0.01,
callbacks=[early_stop_callback, epoch_callback],
limit_train_batches=1.0,
)
net = NBeats.from_dataset(
training,
learning_rate=1e-4,
log_interval=10,
log_val_interval=1,
weight_decay=1e-2,
# widths=[16, 16],
# backcast_loss_ratio=1.0,
)
trainer.fit(
net,
train_dataloader=train_dataloader,
val_dataloaders=val_dataloader,
)
Global seed set to 42
GPU available: True, used: True
TPU available: False, using: 0 TPU cores
IPU available: False, using: 0 IPUs
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/models/nbeats/sub_modules.py:154: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:201.)
coefficients = torch.tensor([backcast_linspace ** i for i in range(thetas_dim)], dtype=torch.float32)
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_lightning/trainer/trainer.py:735: LightningDeprecationWarning: `trainer.fit(train_dataloader)` is deprecated in v1.4 and will be removed in v1.6. Use `trainer.fit(train_dataloaders)` instead. HINT: added 's'
rank_zero_deprecation(
LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0]
Set SLURM handle signals.
| Name | Type | Params
-----------------------------------------------
0 | loss | MASE | 0
1 | logging_metrics | ModuleList | 0
2 | net_blocks | ModuleList | 1.7 M
-----------------------------------------------
1.7 M Trainable params
0 Non-trainable params
1.7 M Total params
6.913 Total estimated model params size (MB)
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_lightning/callbacks/model_checkpoint.py:631: UserWarning: Checkpoint directory /scratch/ssd001/home/ethanj/forecasting-bootcamp/demos/lightning_logs/version_6641049/checkpoints exists and is not empty.
rank_zero_warn(f"Checkpoint directory {dirpath} exists and is not empty.")
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_lightning/trainer/data_loading.py:132: UserWarning: The dataloader, val_dataloader 0, does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` (try 32 which is the number of cpus on this machine) in the `DataLoader` init to improve performance.
rank_zero_warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_lightning/utilities/data.py:59: UserWarning: Trying to infer the `batch_size` from an ambiguous collection. The batch size we found is 128. To avoid any miscalculations, use `self.log(..., batch_size=batch_size)`.
warning_cache.warn(
Global seed set to 42
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_lightning/trainer/data_loading.py:132: UserWarning: The dataloader, train_dataloader, does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` (try 32 which is the number of cpus on this machine) in the `DataLoader` init to improve performance.
rank_zero_warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_lightning/utilities/data.py:59: UserWarning: Trying to infer the `batch_size` from an ambiguous collection. The batch size we found is 104. To avoid any miscalculations, use `self.log(..., batch_size=batch_size)`.
warning_cache.warn(
Validation MASE tensor(5.5188, device='cuda:0')
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
Validation MASE tensor(4.9181, device='cuda:0')
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
Validation MASE tensor(4.7710, device='cuda:0')
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
Validation MASE tensor(4.6807, device='cuda:0')
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
Validation MASE tensor(4.6517, device='cuda:0')
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
Validation MASE tensor(4.6694, device='cuda:0')
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
Validation MASE tensor(4.6186, device='cuda:0')
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
Validation MASE tensor(4.6373, device='cuda:0')
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
Validation MASE tensor(4.6552, device='cuda:0')
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
Validation MASE tensor(4.6492, device='cuda:0')
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
Validation MASE tensor(4.6546, device='cuda:0')
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_forecasting/data/encoders.py:373: UserWarning: scale is below 1e-7 - consider not centering the data or using data with higher variance for numerical stability
warnings.warn(
Validation MASE tensor(4.6444, device='cuda:0')
Testing the model#
We can now evaluate the model using the test set. Even though we trained the N-BEATS model on all variables in the training set, we will only evaluate performance on USD_CLOSE as a direct comparison to earlier experiments with Prophet and NeuralProphet. The following code creates a dataloader for the test set.
test_dataset = TimeSeriesDataSet(
test_df,
time_idx="time_idx",
target="USD_CLOSE",
group_ids=['group_ids'],
max_encoder_length=context_length,
max_prediction_length=max_prediction_length,
time_varying_unknown_reals=["USD_CLOSE"],
)
test_dataloader = test_dataset.to_dataloader(train=False, batch_size=1, num_workers=0)
Collect test set predictions#
best_model_path = trainer.checkpoint_callback.best_model_path
best_model = NBeats.load_from_checkpoint(best_model_path)
actuals = torch.cat([y[0] for x, y in iter(test_dataloader)])
predictions = best_model.predict(test_dataloader)
Plot model forecasts at max lead time#
Like in Demo 1, we are plotting the predictions at max lead time as well as one individual example forecast. Feel free to modify the start variable to plot different examples.
actuals_np = actuals.detach().numpy()
predictions_np = predictions.detach().numpy()
indexes = []
preds = []
trues = []
for time_idx in range(actuals_np.shape[0]):
indexes.append(time_idx)
preds.append(predictions_np[time_idx][-1])
trues.append(actuals_np[time_idx][-1])
fig, ax = plt.subplots()
ax.plot(indexes, preds, color='red', label='forecast')
ax.plot(indexes, trues, color='blue', label='ground truth')
ax.set_title(f"Forecasts at max lead time ({lead_time} samples) - N-BEATS")
ax.legend()
<matplotlib.legend.Legend at 0x7fcc1a0eae20>
# plot at single time index
fig, axs = plt.subplots(4, 4, figsize=(20,20))
axs = axs.flatten()
for fig_index, example_index in enumerate(range(lag_time, len(predictions_np), lead_time)):
if fig_index > len(axs) - 1:
break
n_prev_observations = lag_time
start_index = example_index - n_prev_observations
start_index = 0 if start_index < 0 else start_index
ground_truth = actuals_np[example_index]
forecast = predictions_np[example_index]
axs[fig_index].plot(ground_truth, color='blue', label='ground truth')
axs[fig_index].plot(forecast, color='red', label='forecast')
axs[fig_index].legend()
Compare results to previous experiment#
If you have run through experiments using Demo 1, you can load the results file here and continue the experiment.
results_df = pd.read_csv("./exchange_rate_mean_test_metrics.csv", index_col=0)
results_df.sort_values('mae')
| mse | rmse | mae | mape | |
|---|---|---|---|---|
| persistence_mean_metrics | 0.000544 | 0.020612 | 0.017749 | 0.013385 |
| neural_prophet_last_sample_only_mean_metrics | 0.000655 | 0.022720 | 0.019883 | 0.014971 |
| mean_window_mean_metrics | 0.000914 | 0.026235 | 0.023700 | 0.017844 |
| neural_prophet_baseline_mean_metrics | 0.001107 | 0.031964 | 0.026256 | 0.019803 |
| neural_prophet_sparse_ar_mean_metrics | 0.004883 | 0.056420 | 0.048945 | 0.036973 |
| prophet_mean_metrics | 0.004835 | 0.061034 | 0.058992 | 0.044357 |
To ensure consistency, we will redefine the same metrics from Demo 1.
from sklearn.metrics import mean_squared_error, mean_absolute_error, mean_absolute_percentage_error
metrics = {
'mse': mean_squared_error,
'rmse': lambda y_true, y_pred: np.sqrt(mean_squared_error(y_true, y_pred)),
'mae': mean_absolute_error,
'mape': mean_absolute_percentage_error
}
def compute_error_statistics(error_metrics_dict, exp_name):
return {
'mean': pd.DataFrame(error_metrics_dict).mean(axis=0).rename(f'{exp_name}_mean_metrics'),
'std': pd.DataFrame(error_metrics_dict).std(axis=0).rename(f'{exp_name}_std_metrics'),
'max': pd.DataFrame(error_metrics_dict).max(axis=0).rename(f'{exp_name}_max_metrics'),
}
def compute_baseline_error_metrics(predictions, actuals):
errors = {metric_name:[] for metric_name in metrics.keys()}
for i in range(len(predictions)):
for metric_name, metric_fn in metrics.items():
errors[metric_name].append(metric_fn(y_true=actuals[i], y_pred=predictions[i]))
return errors
actuals_np = actuals.detach().numpy()
predictions_np = predictions.detach().numpy()
nbeats_errors = compute_baseline_error_metrics(predictions_np, actuals_np)
nbeats_stats = compute_error_statistics(nbeats_errors, 'nbeats')
nbeats_stats['mean']
mse 0.000585
rmse 0.021233
mae 0.018360
mape 0.013796
Name: nbeats_mean_metrics, dtype: float32
results_df = results_df.append(nbeats_stats['mean'])
results_df.sort_values('mae')
| mse | rmse | mae | mape | |
|---|---|---|---|---|
| persistence_mean_metrics | 0.000544 | 0.020612 | 0.017749 | 0.013385 |
| nbeats_mean_metrics | 0.000585 | 0.021233 | 0.018360 | 0.013796 |
| neural_prophet_last_sample_only_mean_metrics | 0.000655 | 0.022720 | 0.019883 | 0.014971 |
| mean_window_mean_metrics | 0.000914 | 0.026235 | 0.023700 | 0.017844 |
| neural_prophet_baseline_mean_metrics | 0.001107 | 0.031964 | 0.026256 | 0.019803 |
| neural_prophet_sparse_ar_mean_metrics | 0.004883 | 0.056420 | 0.048945 | 0.036973 |
| prophet_mean_metrics | 0.004835 | 0.061034 | 0.058992 | 0.044357 |
DeepAR#
We will now add pytorch-forecasting’s implementation of DeepAR to our experiment. The implementation is based on the original paper. DeepAR is a method for probabilistic forecasting with autoregressive recurrent neural networks. DeepAR learns a global model from all historical data in the dataset, similar to N-BEATS.
deepar_training = TimeSeriesDataSet(
train_df_nbeats.loc[train_df_nbeats.time_idx <= training_cutoff], # Applying the training cutoff
time_idx="time_idx",
target="value",
group_ids=['group_ids'],
min_encoder_length=context_length,
max_encoder_length=context_length,
min_prediction_length=1,
max_prediction_length=max_prediction_length,
categorical_encoders={"group_ids": NaNLabelEncoder().fit(train_df_nbeats.group_ids)},
time_varying_unknown_reals=["value"],
time_varying_known_reals=["time_idx"],
target_normalizer=GroupNormalizer(groups=["group_ids"]),
add_relative_time_idx=False,
add_target_scales=True,
randomize_length=None,
)
deepar_validation = TimeSeriesDataSet.from_dataset(deepar_training,
train_df_nbeats,
min_prediction_idx=training_cutoff + 1)
batch_size = 64
train_dataloader = deepar_training.to_dataloader(train=True, batch_size=batch_size, num_workers=0)
val_dataloader = deepar_validation.to_dataloader(train=False, batch_size=batch_size, num_workers=0)
pl.seed_everything(42)
trainer = pl.Trainer(gpus=1)
net = DeepAR.from_dataset(deepar_training)
Global seed set to 42
GPU available: True, used: True
TPU available: False, using: 0 TPU cores
IPU available: False, using: 0 IPUs
early_stop_callback = EarlyStopping(monitor="val_MASE", min_delta=1e-4, patience=5, verbose=False, mode="min")
epoch_callback = EpochCallback()
trainer = pl.Trainer(
max_epochs=100,
gpus=1,
gradient_clip_val=0.1,
# weights_summary="top",
callbacks=[early_stop_callback, epoch_callback],
limit_train_batches=1.0,
)
net = DeepAR.from_dataset(
deepar_training,
hidden_size=32,
dropout=0.1,
loss=NormalDistributionLoss(),
learning_rate=1e-4,
log_interval=10,
log_val_interval=1,
)
trainer.fit(
net,
train_dataloader=train_dataloader,
val_dataloaders=val_dataloader,
)
GPU available: True, used: True
TPU available: False, using: 0 TPU cores
IPU available: False, using: 0 IPUs
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_lightning/trainer/trainer.py:735: LightningDeprecationWarning: `trainer.fit(train_dataloader)` is deprecated in v1.4 and will be removed in v1.6. Use `trainer.fit(train_dataloaders)` instead. HINT: added 's'
rank_zero_deprecation(
LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0]
Set SLURM handle signals.
| Name | Type | Params
------------------------------------------------------------------
0 | loss | NormalDistributionLoss | 0
1 | logging_metrics | ModuleList | 0
2 | embeddings | MultiEmbedding | 0
3 | rnn | LSTM | 13.3 K
4 | distribution_projector | Linear | 66
------------------------------------------------------------------
13.4 K Trainable params
0 Non-trainable params
13.4 K Total params
0.054 Total estimated model params size (MB)
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_lightning/callbacks/model_checkpoint.py:631: UserWarning: Checkpoint directory /scratch/ssd001/home/ethanj/forecasting-bootcamp/demos/lightning_logs/version_6641049/checkpoints exists and is not empty.
rank_zero_warn(f"Checkpoint directory {dirpath} exists and is not empty.")
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_lightning/trainer/data_loading.py:132: UserWarning: The dataloader, val_dataloader 0, does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` (try 32 which is the number of cpus on this machine) in the `DataLoader` init to improve performance.
rank_zero_warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_lightning/utilities/data.py:59: UserWarning: Trying to infer the `batch_size` from an ambiguous collection. The batch size we found is 64. To avoid any miscalculations, use `self.log(..., batch_size=batch_size)`.
warning_cache.warn(
Global seed set to 42
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_lightning/trainer/data_loading.py:132: UserWarning: The dataloader, train_dataloader, does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` (try 32 which is the number of cpus on this machine) in the `DataLoader` init to improve performance.
rank_zero_warn(
/ssd003/projects/aieng/public/forecasting_unified/lib/python3.8/site-packages/pytorch_lightning/utilities/data.py:59: UserWarning: Trying to infer the `batch_size` from an ambiguous collection. The batch size we found is 32. To avoid any miscalculations, use `self.log(..., batch_size=batch_size)`.
warning_cache.warn(
Validation MASE tensor(24.1788, device='cuda:0')
Validation MASE tensor(22.4341, device='cuda:0')
Validation MASE tensor(15.3717, device='cuda:0')
Validation MASE tensor(12.1220, device='cuda:0')
Validation MASE tensor(9.8014, device='cuda:0')
Validation MASE tensor(7.2970, device='cuda:0')
Validation MASE tensor(6.3345, device='cuda:0')
Validation MASE tensor(5.6729, device='cuda:0')
Validation MASE tensor(5.2699, device='cuda:0')
Validation MASE tensor(5.0503, device='cuda:0')
Validation MASE tensor(4.8650, device='cuda:0')
Validation MASE tensor(4.7308, device='cuda:0')
Validation MASE tensor(4.6130, device='cuda:0')
Validation MASE tensor(4.5380, device='cuda:0')
Validation MASE tensor(4.5167, device='cuda:0')
Validation MASE tensor(4.4723, device='cuda:0')
Validation MASE tensor(4.4954, device='cuda:0')
Validation MASE tensor(4.4755, device='cuda:0')
Validation MASE tensor(4.5482, device='cuda:0')
Validation MASE tensor(4.5403, device='cuda:0')
Validation MASE tensor(4.5327, device='cuda:0')
# save datasets
deepar_training.save("training.pkl")
deepar_validation.save("validation.pkl")
best_model_path = trainer.checkpoint_callback.best_model_path
best_model = DeepAR.load_from_checkpoint(best_model_path)
print(best_model_path)
/scratch/ssd001/home/ethanj/forecasting-bootcamp/demos/lightning_logs/version_6641049/checkpoints/epoch=20-step=8483.ckpt
Load the model from a saved checkpoint (optional)#
best_model = DeepAR.load_from_checkpoint(best_model_path)
Prepare the test dataset with necessary preprocessing options#
Since we only want to evaluate on USD_CLOSE, the test dataset should only contain data from this series (as opposed to all of them vertically stacked like we did for training).
deep_ar_test_dataset = TimeSeriesDataSet(
test_df,
time_idx="time_idx",
target="USD_CLOSE",
group_ids=['group_ids'],
max_encoder_length=context_length,
max_prediction_length=max_prediction_length,
categorical_encoders={"group_ids": NaNLabelEncoder().fit(test_df.group_ids)},
time_varying_unknown_reals=["USD_CLOSE"],
time_varying_known_reals=["time_idx"],
target_normalizer=GroupNormalizer(groups=["group_ids"]),
add_relative_time_idx=False,
add_target_scales=True,
randomize_length=None,
)
deepar_test_dataloader = deep_ar_test_dataset.to_dataloader(train=False, batch_size=1, num_workers=0)
Same visualization and evaluation code that we used previously#
actuals = torch.cat([y[0] for x, y in iter(deepar_test_dataloader) if y[0].shape[1] == lead_time])
print(actuals.shape)
torch.Size([642, 30])
predictions = best_model.predict(deepar_test_dataloader)
print(predictions.shape)
torch.Size([642, 30])
actuals_np = actuals.detach().numpy()
predictions_np = predictions.detach().numpy()
indexes = []
preds = []
trues = []
for time_idx in range(actuals_np.shape[0]):
indexes.append(time_idx)
preds.append(predictions_np[time_idx][-1])
trues.append(actuals_np[time_idx][-1])
fig, ax = plt.subplots()
ax.plot(indexes, preds, color='red', label='forecast')
ax.plot(indexes, trues, color='blue', label='ground truth')
ax.set_title(f"Forecasts at max lead time ({lead_time} samples) - N-BEATS")
ax.legend()
<matplotlib.legend.Legend at 0x7fcbdd60b550>
# plot at single time index
fig, axs = plt.subplots(4, 4, figsize=(20,20))
axs = axs.flatten()
for fig_index, example_index in enumerate(range(lag_time, len(predictions_np), lead_time)):
if fig_index > len(axs) - 1:
break
n_prev_observations = lag_time
start_index = example_index - n_prev_observations
start_index = 0 if start_index < 0 else start_index
ground_truth = actuals_np[example_index]
forecast = predictions_np[example_index]
axs[fig_index].plot(ground_truth, color='blue', label='ground truth')
axs[fig_index].plot(forecast, color='red', label='forecast')
axs[fig_index].legend()
deepar_errors = compute_baseline_error_metrics(predictions_np, actuals_np)
deepar_stats = compute_error_statistics(deepar_errors, 'deepAR')
deepar_stats['mean']
mse 0.000513
rmse 0.020122
mae 0.017302
mape 0.013006
Name: deepAR_mean_metrics, dtype: float32
results_df = results_df.append(deepar_stats['mean'])
results_df.sort_values('mae')
| mse | rmse | mae | mape | |
|---|---|---|---|---|
| deepAR_mean_metrics | 0.000513 | 0.020122 | 0.017302 | 0.013006 |
| deepAR_mean_metrics | 0.000513 | 0.020122 | 0.017302 | 0.013006 |
| persistence_mean_metrics | 0.000544 | 0.020612 | 0.017749 | 0.013385 |
| nbeats_mean_metrics | 0.000585 | 0.021233 | 0.018360 | 0.013796 |
| neural_prophet_last_sample_only_mean_metrics | 0.000655 | 0.022720 | 0.019883 | 0.014971 |
| mean_window_mean_metrics | 0.000914 | 0.026235 | 0.023700 | 0.017844 |
| neural_prophet_baseline_mean_metrics | 0.001107 | 0.031964 | 0.026256 | 0.019803 |
| neural_prophet_sparse_ar_mean_metrics | 0.004883 | 0.056420 | 0.048945 | 0.036973 |
| prophet_mean_metrics | 0.004835 | 0.061034 | 0.058992 | 0.044357 |