# AUTOGENERATED! DO NOT EDIT! File to edit: ../../nbs/models.tcn.ipynb.
# %% auto 0
__all__ = ['TCN']
# %% ../../nbs/models.tcn.ipynb 5
from typing import List, Optional
import torch
import torch.nn as nn
from ..losses.pytorch import MAE
from ..common._base_recurrent import BaseRecurrent
from ..common._modules import MLP, TemporalConvolutionEncoder
# %% ../../nbs/models.tcn.ipynb 7
class TCN(BaseRecurrent):
"""TCN
Temporal Convolution Network (TCN), with MLP decoder.
The historical encoder uses dilated skip connections to obtain efficient long memory,
while the rest of the architecture allows for future exogenous alignment.
**Parameters:**
`h`: int, forecast horizon.
`input_size`: int, maximum sequence length for truncated train backpropagation. Default -1 uses all history.
`inference_input_size`: int, maximum sequence length for truncated inference. Default -1 uses all history.
`kernel_size`: int, size of the convolving kernel.
`dilations`: int list, ontrols the temporal spacing between the kernel points; also known as the à trous algorithm.
`encoder_hidden_size`: int=200, units for the TCN's hidden state size.
`encoder_activation`: str=`tanh`, type of TCN activation from `tanh` or `relu`.
`context_size`: int=10, size of context vector for each timestamp on the forecasting window.
`decoder_hidden_size`: int=200, size of hidden layer for the MLP decoder.
`decoder_layers`: int=2, number of layers for the MLP decoder.
`futr_exog_list`: str list, future exogenous columns.
`hist_exog_list`: str list, historic exogenous columns.
`stat_exog_list`: str list, static exogenous columns.
`loss`: PyTorch module, instantiated train loss class from [losses collection](https://nixtla.github.io/neuralforecast/losses.pytorch.html).
`max_steps`: int=1000, maximum number of training steps.
`learning_rate`: float=1e-3, Learning rate between (0, 1).
`valid_batch_size`: int=None, number of different series in each validation and test batch.
`num_lr_decays`: int=-1, Number of learning rate decays, evenly distributed across max_steps.
`early_stop_patience_steps`: int=-1, Number of validation iterations before early stopping.
`val_check_steps`: int=100, Number of training steps between every validation loss check.
`batch_size`: int=32, number of differentseries in each batch.
`scaler_type`: str='robust', type of scaler for temporal inputs normalization see [temporal scalers](https://nixtla.github.io/neuralforecast/common.scalers.html).
`random_seed`: int=1, random_seed for pytorch initializer and numpy generators.
`num_workers_loader`: int=os.cpu_count(), workers to be used by `TimeSeriesDataLoader`.
`drop_last_loader`: bool=False, if True `TimeSeriesDataLoader` drops last non-full batch.
`alias`: str, optional, Custom name of the model.
`optimizer`: Subclass of 'torch.optim.Optimizer', optional, user specified optimizer instead of the default choice (Adam).
`optimizer_kwargs`: dict, optional, list of parameters used by the user specified `optimizer`.
`**trainer_kwargs`: int, keyword trainer arguments inherited from [PyTorch Lighning's trainer](https://pytorch-lightning.readthedocs.io/en/stable/api/pytorch_lightning.trainer.trainer.Trainer.html?highlight=trainer).
"""
# Class attributes
SAMPLING_TYPE = "recurrent"
def __init__(
self,
h: int,
input_size: int = -1,
inference_input_size: int = -1,
kernel_size: int = 2,
dilations: List[int] = [1, 2, 4, 8, 16],
encoder_hidden_size: int = 200,
encoder_activation: str = "ReLU",
context_size: int = 10,
decoder_hidden_size: int = 200,
decoder_layers: int = 2,
futr_exog_list=None,
hist_exog_list=None,
stat_exog_list=None,
loss=MAE(),
valid_loss=None,
max_steps: int = 1000,
learning_rate: float = 1e-3,
num_lr_decays: int = -1,
early_stop_patience_steps: int = -1,
val_check_steps: int = 100,
batch_size: int = 32,
valid_batch_size: Optional[int] = None,
scaler_type: str = "robust",
random_seed: int = 1,
num_workers_loader=0,
drop_last_loader=False,
optimizer=None,
optimizer_kwargs=None,
**trainer_kwargs
):
super(TCN, self).__init__(
h=h,
input_size=input_size,
inference_input_size=inference_input_size,
loss=loss,
valid_loss=valid_loss,
max_steps=max_steps,
learning_rate=learning_rate,
num_lr_decays=num_lr_decays,
early_stop_patience_steps=early_stop_patience_steps,
val_check_steps=val_check_steps,
batch_size=batch_size,
valid_batch_size=valid_batch_size,
scaler_type=scaler_type,
futr_exog_list=futr_exog_list,
hist_exog_list=hist_exog_list,
stat_exog_list=stat_exog_list,
num_workers_loader=num_workers_loader,
drop_last_loader=drop_last_loader,
random_seed=random_seed,
optimizer=optimizer,
optimizer_kwargs=optimizer_kwargs,
**trainer_kwargs
)
# ----------------------------------- Parse dimensions -----------------------------------#
# TCN
self.kernel_size = kernel_size
self.dilations = dilations
self.encoder_hidden_size = encoder_hidden_size
self.encoder_activation = encoder_activation
# Context adapter
self.context_size = context_size
# MLP decoder
self.decoder_hidden_size = decoder_hidden_size
self.decoder_layers = decoder_layers
self.futr_exog_size = len(self.futr_exog_list)
self.hist_exog_size = len(self.hist_exog_list)
self.stat_exog_size = len(self.stat_exog_list)
# TCN input size (1 for target variable y)
input_encoder = 1 + self.hist_exog_size + self.stat_exog_size
# ---------------------------------- Instantiate Model -----------------------------------#
# Instantiate historic encoder
self.hist_encoder = TemporalConvolutionEncoder(
in_channels=input_encoder,
out_channels=self.encoder_hidden_size,
kernel_size=self.kernel_size, # Almost like lags
dilations=self.dilations,
activation=self.encoder_activation,
)
# Context adapter
self.context_adapter = nn.Linear(
in_features=self.encoder_hidden_size + self.futr_exog_size * h,
out_features=self.context_size * h,
)
# Decoder MLP
self.mlp_decoder = MLP(
in_features=self.context_size + self.futr_exog_size,
out_features=self.loss.outputsize_multiplier,
hidden_size=self.decoder_hidden_size,
num_layers=self.decoder_layers,
activation="ReLU",
dropout=0.0,
)
def forward(self, windows_batch):
# Parse windows_batch
encoder_input = windows_batch["insample_y"] # [B, seq_len, 1]
futr_exog = windows_batch["futr_exog"]
hist_exog = windows_batch["hist_exog"]
stat_exog = windows_batch["stat_exog"]
# Concatenate y, historic and static inputs
# [B, C, seq_len, 1] -> [B, seq_len, C]
# Contatenate [ Y_t, | X_{t-L},..., X_{t} | S ]
batch_size, seq_len = encoder_input.shape[:2]
if self.hist_exog_size > 0:
hist_exog = hist_exog.permute(0, 2, 1, 3).squeeze(
-1
) # [B, X, seq_len, 1] -> [B, seq_len, X]
encoder_input = torch.cat((encoder_input, hist_exog), dim=2)
if self.stat_exog_size > 0:
stat_exog = stat_exog.unsqueeze(1).repeat(
1, seq_len, 1
) # [B, S] -> [B, seq_len, S]
encoder_input = torch.cat((encoder_input, stat_exog), dim=2)
# TCN forward
hidden_state = self.hist_encoder(
encoder_input
) # [B, seq_len, tcn_hidden_state]
if self.futr_exog_size > 0:
futr_exog = futr_exog.permute(0, 2, 3, 1)[
:, :, 1:, :
] # [B, F, seq_len, 1+H] -> [B, seq_len, H, F]
hidden_state = torch.cat(
(hidden_state, futr_exog.reshape(batch_size, seq_len, -1)), dim=2
)
# Context adapter
context = self.context_adapter(hidden_state)
context = context.reshape(batch_size, seq_len, self.h, self.context_size)
# Residual connection with futr_exog
if self.futr_exog_size > 0:
context = torch.cat((context, futr_exog), dim=-1)
# Final forecast
output = self.mlp_decoder(context)
output = self.loss.domain_map(output)
return output