[ENH] ModernTCN: A Modern Pure Convolution Structure for General Time Series Analysis added

[Muhammad-Rebaal]

Fixes #2358

This PR introduces v2 interface support for ModernTCN (ICLR 2024), a highly efficient, pure convolutional time series forecasting model that leverages large-kernel depthwise convolutions and patching to achieve state-of-the-art performance.

Reference

• ModernTCN: A Modern Pure Convolutional Structure for General Time Series Analysis (Luo & Wang, ICLR 2024)

[codecov]

Codecov Report

❌ Patch coverage is 99.42857% with 1 line in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (main@7bcf66c). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
...ch_forecasting/models/modern_tcn/_modern_tcn_v2.py	98.38%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2363   +/-   ##
=======================================
  Coverage        ?   87.75%           
=======================================
  Files           ?      183           
  Lines           ?    10339           
  Branches        ?        0           
=======================================
  Hits            ?     9073           
  Misses          ?     1266           
  Partials        ?        0           

Flag	Coverage Δ
cpu	87.75% <99.42%> (?)
pytest	87.75% <99.42%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Muhammad-Rebaal]

Hi @phoeenniixx,

I've implemented the ModernTCN model.

Here are some changes, in the pytorch_forecasting\layers:

• Added re-usable block as ModernTCNBlock
• Added two new layers:

1. _convolution that contains _reparam_large_kernel_conv.py
2. _head that contains _flatten_head.py

The purpose of adding these layers separately is for reusability in future models.