A meta-classification for properties, characteristics, and measurable attributes of machine learning models including performance metrics (accuracy, latency, throughput), architectural properties (parameters, layers, context length), and operational characteristics (memory footprint, inference co…
Semantic Classification
Content
Property Types (Inferred by Reasoner)
-
Accuracy is-a ModelProperty
-
Latency is-a ModelProperty
-
Throughput is-a ModelProperty
-
Parameter Count is-a ModelProperty
-
Context Length is-a ModelProperty
-
Memory Footprint is-a ModelProperty
-
Inference Cost is-a ModelProperty
-
Training Compute is-a ModelProperty
Definition
ModelProperty serves as the meta-class for all measurable and descriptive properties of machine learning models. This classification enables structured representation of model characteristics for comparison, selection, and governance purposes.
Property Categories
Performance Properties
Property Type Unit Description Accuracy Metric % Correct predictions ratio Precision Metric % True positive ratio Recall Metric % Sensitivity/TPR F1 Score Metric 0-1 Harmonic mean of precision/recall Perplexity Metric scalar Language model uncertainty Latency Properties
Property Type Unit Description Inference Latency Metric ms Time per prediction Time to First Token Metric ms Streaming response start Tokens per Second Metric tok/s Generation throughput Batch Throughput Metric req/s Concurrent request handling Architectural Properties
Property Type Unit Description Parameter Count Size B Total trainable parameters Layer Count Structure int Network depth Context Length Capacity tokens Maximum input sequence Hidden Dimension Structure int Internal representation size Attention Heads Structure int Multi-head attention count Operational Properties
Property Type Unit Description Memory Footprint Resource GB VRAM/RAM required Inference Cost Economic $/1M tokens API pricing Training Compute Resource FLOP Total training compute Energy Consumption Resource kWh Power requirements Usage in Ontology
Property Governance
Model properties support:
-
Model Cards: Standardised documentation
-
Benchmarking: Performance comparison
-
Procurement: Selection criteria
-
Regulation: Compliance verification
-
Risk Assessment: Capability evaluation