When should I use this skill?

Use this skill at the very beginning of a project when you need to decide which neural network architecture best fits your data and performance requirements.

Does this skill help with PyTorch implementation?

No, this skill focuses on high-level architecture selection and design principles. You should use implementation-specific packs for coding details.

Can it help fix training instability?

Yes, it provides specific guidance on normalization techniques and architectural stability for deep networks with more than 20 layers.

Does it support generative models like GANs or Diffusion?

Yes, it includes a dedicated route for generative model families to help you choose between GANs, VAEs, and Diffusion models based on your quality and speed needs.

Neural Architecture Selection Router

Name: Neural Architecture Selection Router
Author: tachyon-beep

bytachyon-beep

•

Data Science & ML

Guides developers in choosing the optimal neural network architecture based on data modality, problem constraints, and performance requirements.

About

This skill serves as a strategic decision-making framework for selecting neural network architectures before the training phase begins. It provides a structured routing logic that evaluates data types—such as images, sequences, and graphs—against specific constraints like dataset size, latency requirements, and computational resources. By systematically directing users to the appropriate model families, including CNNs, Transformers, GNNs, and Generative models, it helps prevent architectural mismatches that cannot be fixed by later optimization.

Key Features

5 GitHub stars
Architecture comparison guides (e.g., RNN vs. Transformer)
Custom topology and design principle guidance
Modality-based routing for images, sequences, and graphs
Structural troubleshooting for deep network instability
Constraint-aware selection based on compute and latency

Use Cases

Deciding between CNNs and Vision Transformers for computer vision tasks
Selecting the right model family for a new machine learning project
Identifying the best sequential model for time-series forecasting or NLP

About

Key Features

5 GitHub stars
Architecture comparison guides (e.g., RNN vs. Transformer)
Custom topology and design principle guidance
Modality-based routing for images, sequences, and graphs
Structural troubleshooting for deep network instability
Constraint-aware selection based on compute and latency

Use Cases

Deciding between CNNs and Vision Transformers for computer vision tasks
Selecting the right model family for a new machine learning project
Identifying the best sequential model for time-series forecasting or NLP