Implementation:Sktime Pytorch forecasting mLSTMNetwork

Knowledge Sources	Sktime_Pytorch_forecasting
Domains	Time_Series, Forecasting, Deep_Learning
Last Updated	2026-02-08 08:00 GMT

Overview

mLSTMNetwork is a complete model combining stacked mLSTM layers with a fully connected output layer for sequence prediction.

Description

mLSTMNetwork extends nn.Module and composes an mLSTMLayer (stacked matrix LSTM layers) with a final nn.Linear fully connected layer. The mLSTM layers process the input sequence recurrently, and the FC layer maps the last time step's hidden output to the desired output dimension. This provides an end-to-end trainable network architecture based on the mLSTM cell.

Usage

Use mLSTMNetwork as a standalone recurrent forecasting model when you want a complete mLSTM-based architecture with configurable depth, hidden size, dropout, layer normalization, and residual connections.

Code Reference

Source Location

• Repository: Sktime_Pytorch_forecasting
• File: pytorch_forecasting/layers/_recurrent/_mlstm/network.py
• Lines: 1-99

Signature

class mLSTMNetwork(nn.Module):
    def __init__(
        self,
        input_size,
        hidden_size,
        num_layers,
        output_size,
        dropout=0.0,
        use_layer_norm=True,
        use_residual=True,
    ):
    def forward(self, x, h=None, c=None, n=None):
    def init_hidden(self, batch_size, device=None):

Import

from pytorch_forecasting.layers._recurrent._mlstm.network import mLSTMNetwork

I/O Contract

Inputs

__init__

Name	Type	Required	Description
input_size	int	Yes	Number of features in the input.
hidden_size	int	Yes	Number of features in the hidden state of each mLSTM layer.
num_layers	int	Yes	Number of mLSTM layers to stack.
output_size	int	Yes	Number of features in the output.
dropout	float	No	Dropout probability for the mLSTM layers. Defaults to 0.0.
use_layer_norm	bool	No	Whether to use layer normalization in the mLSTM layers. Defaults to True.
use_residual	bool	No	Whether to use residual connections in the mLSTM layers. Defaults to True.

forward

Name	Type	Required	Description
x	torch.Tensor	Yes	Input sequence tensor.
h	torch.Tensor or None	No	Initial hidden states for all layers. If None, initialized to zeros.
c	torch.Tensor or None	No	Initial cell states for all layers. If None, initialized to zeros.
n	torch.Tensor or None	No	Initial normalized states for all layers. If None, initialized to zeros.

Outputs

forward

Name	Type	Description
output	torch.Tensor	Final output tensor from the fully connected layer, produced from the last time step.
(h, c, n)	tuple of torch.Tensor	Final hidden, cell, and normalized states for all layers.

Usage Examples

import torch
from pytorch_forecasting.layers._recurrent._mlstm.network import mLSTMNetwork

model = mLSTMNetwork(
    input_size=32,
    hidden_size=64,
    num_layers=2,
    output_size=10,
    dropout=0.1,
    use_layer_norm=True,
    use_residual=True,
)

seq_len, batch_size = 20, 16
x = torch.randn(seq_len, batch_size, 32)
output, (h, c, n) = model(x)
# output shape: (16, 10)  -- last time step projected to output_size

Related Pages

• Principle:Sktime_Pytorch_forecasting_xLSTM_Architecture
• Implementation:Sktime_Pytorch_forecasting_mLSTMCell
• Implementation:Sktime_Pytorch_forecasting_mLSTMLayer