identify-architecture

Analyze ML model architecture from papers and code. Use when understanding model structure for implementation.

⭐8stars🔀4forks📁View on GitHub🕐Updated Jan 3, 2026

When & Why to Use This Skill

The Identify Architecture skill empowers developers and machine learning researchers to systematically deconstruct complex ML model structures from research papers and source code. By extracting critical data such as layer specifications, connection maps, and information flow, it bridges the gap between theoretical research and practical implementation, providing a clear blueprint for building or documenting neural networks.

Use Cases

  • Academic Research Translation: Automatically convert complex neural network diagrams and descriptions from PDF research papers into structured, layer-by-layer technical specifications.
  • Model Implementation Planning: Generate detailed implementation blueprints, including kernel sizes, strides, and activation functions, to streamline the coding process in frameworks like Mojo or Python.
  • Legacy Code Documentation: Analyze existing machine learning codebases to create comprehensive architecture documentation and data flow diagrams for better team onboarding and maintenance.
  • Architecture Comparison & Auditing: Systematically compare different model variations or versions to identify structural changes, optimization points, and design inconsistencies.
nameidentify-architecture
description"Analyze ML model architecture from papers and code. Use when understanding model structure for implementation."
mcp_fallbacknone
categoryanalysis
tier2
user-invocablefalse

Identify Architecture

Analyze and document machine learning model architectures including layers, connections, and information flow.

When to Use

  • Understanding paper model designs
  • Planning model implementation
  • Comparing architecture variations
  • Documenting neural network structure

Quick Reference

# Extract architecture from paper
# Look for: "Figure X: Architecture of [Model]"
# Check for: Table with layer specifications
# Find: Layer descriptions (Conv2D, FC, BatchNorm, etc.)

# Visualize model structure (Mojo)
# var model: SimpleNet = ...
# print(model)  # Should show layer information

Workflow

  1. Locate architecture diagram: Find visual architecture representation in paper
  2. List layers: Enumerate all layers with type and parameters
  3. Document connections: Map data flow between layers (skip connections, merges)
  4. Extract layer parameters: For each layer record size, activation, normalization
  5. Create implementation plan: Translate to Mojo struct/function definitions

Output Format

Architecture documentation:

  • Model name and source
  • Layer-by-layer breakdown
  • Layer type (Conv2D, Dense, etc.)
  • Parameters (kernel size, stride, padding, activation)
  • Input/output shapes
  • Data flow diagram (text or ASCII)
  • Special components (skip connections, attention)

References

  • See extract-hyperparameters skill for model configuration
  • See CLAUDE.md > Mojo Syntax Standards for implementation patterns
  • See /notes/review/mojo-ml-patterns.md for architecture patterns