Train State Space ML Skill

What problem does it solve?

This Skill addresses the complexity of training advanced state-space models (SSMs) like IQUMamba, enabling efficient and scalable machine learning model development.

Core Features & Use Cases

• State-Space Model Training: Facilitates the training of S6 selective state-space blocks using a WGPU backend.
• Latent Space Learning: Supports latent space training with phase-coherent embeddings for richer representations.
• Graph-Aware Encoders: Integrates graph context for enhanced inference.
• Use Case: Train a new IQUMamba-1D model for sequence prediction tasks, leveraging its efficient state-space architecture and graph-aware capabilities.

Quick Start

Train the mamba model using the provided configuration and synthetic data.