Training Data for 2000-Assembly MKCad Dataset
Budget: $1,500 – $3,000 USD
Training Data Plan for 2000-Assembly MKCad Dataset
I want to create a text to MKCad assembly website with a trained Llama-4 model. The model needs to interpret natural language requests and turn it into a physical assembly. For example, “Create a simple drive chain that uses [motor name], [specific axle name], and 2 gears with 36 teeth.” The website should then create this with gears and axles that can spin via different joints like a revolute mate, etc. To make it simpler, the LLM only needs to be trained on MKCad components.
Dataset Size and Distribution
Total Examples: 2000 Assemblies
By Complexity:
Simple mechanisms (2-5 parts): 800-900 examples (40-45%)
Medium assemblies (6-15 parts): 800-900 examples (40-45%)
Complex assemblies (16+ parts): 300-400 examples (15-20%)
By Mechanism Category:
Structural systems: 400-450 examples (20-22%)
Frames, chassis, brackets, mounts
Load-bearing structures
Component housings and enclosures
Motion systems: 450-500 examples (22-25%)
Various joint types (revolute, prismatic, cylindrical, spherical)
Linear motion mechanisms
Multi-axis positioning systems
Compound motion mechanisms
Manipulation systems: 350-400 examples (17-20%)
Various gripper types
End effectors with different functions
Tool changers and adapters
Power transmission: 350-400 examples (17-20%)
Gear systems (various types and configurations)
Belt and chain drives
Linkages and cams
Shaft couplings and connections
Integration examples: 300-350 examples (15-17%)
Subsystem interfaces
Complete robotic systems
Multi-function assemblies
Data Collection Strategy
1. Systematic Augmentation (600-800 examples)
Use parametric variations of base designs
Create related designs with different implementation approaches
Generate examples with different constraint types achieving the same function
Data Quality Standards
For Each Assembly:
Complete Featurescript code (validated to be functional)
This code, if run, should create the exact assembly
Comprehensive text description (200-400 words) including:
Purpose and function
Component relationships
Design constraints and considerations
Assembly procedure
Metadata tags:
Complexity level
Primary and secondary mechanism types
Part count and types
Constraint types used
I want to create a text to MKCad assembly website with a trained Llama-4 model. The model needs to interpret natural language requests and turn it into a physical assembly. For example, “Create a simple drive chain that uses [motor name], [specific axle name], and 2 gears with 36 teeth.” The website should then create this with gears and axles that can spin via different joints like a revolute mate, etc. To make it simpler, the LLM only needs to be trained on MKCad components.
Dataset Size and Distribution
Total Examples: 2000 Assemblies
By Complexity:
Simple mechanisms (2-5 parts): 800-900 examples (40-45%)
Medium assemblies (6-15 parts): 800-900 examples (40-45%)
Complex assemblies (16+ parts): 300-400 examples (15-20%)
By Mechanism Category:
Structural systems: 400-450 examples (20-22%)
Frames, chassis, brackets, mounts
Load-bearing structures
Component housings and enclosures
Motion systems: 450-500 examples (22-25%)
Various joint types (revolute, prismatic, cylindrical, spherical)
Linear motion mechanisms
Multi-axis positioning systems
Compound motion mechanisms
Manipulation systems: 350-400 examples (17-20%)
Various gripper types
End effectors with different functions
Tool changers and adapters
Power transmission: 350-400 examples (17-20%)
Gear systems (various types and configurations)
Belt and chain drives
Linkages and cams
Shaft couplings and connections
Integration examples: 300-350 examples (15-17%)
Subsystem interfaces
Complete robotic systems
Multi-function assemblies
Data Collection Strategy
1. Systematic Augmentation (600-800 examples)
Use parametric variations of base designs
Create related designs with different implementation approaches
Generate examples with different constraint types achieving the same function
Data Quality Standards
For Each Assembly:
Complete Featurescript code (validated to be functional)
This code, if run, should create the exact assembly
Comprehensive text description (200-400 words) including:
Purpose and function
Component relationships
Design constraints and considerations
Assembly procedure
Metadata tags:
Complexity level
Primary and secondary mechanism types
Part count and types
Constraint types used