Module monitor 

Real-time Training Monitoring Module

Provides low-overhead metrics collection with SIMD-accelerated aggregation.

Architecture

• MetricsCollector: Collects metrics during training (trueno SIMD)
• MetricsSummary: Statistical summary (mean, std, min, max)
• MetricRecord: Individual metric record with timestamp

Example

use entrenar::monitor::{MetricsCollector, Metric};

let mut collector = MetricsCollector::new();
collector.record(Metric::Loss, 0.5);
collector.record(Metric::Accuracy, 0.85);

let summary = collector.summary();
let loss_stats = summary.get(&Metric::Loss).expect("loss metric must exist");
println!("Mean loss: {}", loss_stats.mean);

Toyota Way: 現地現物 (Genchi Genbutsu)

All metrics are measured, not inferred. Every value comes from actual training.

Re-exports

pub use andon::Alert;

pub use andon::AlertLevel;

pub use andon::AndonConfig;

pub use andon::AndonSystem;

pub use dashboard::Dashboard;

pub use dashboard::DashboardConfig;

pub use drift::Anomaly;

pub use drift::AnomalySeverity;

pub use drift::DriftDetector;

pub use drift::DriftStatus;

pub use drift::SlidingWindowBaseline;

pub use export::ExportFormat;

pub use export::MetricsExporter;

pub use lineage::ChangeType;

pub use lineage::Derivation;

pub use lineage::ModelLineage;

pub use lineage::ModelMetadata;

pub use llm::EvalResult;

pub use llm::InMemoryLLMEvaluator;

pub use llm::LLMError;

pub use llm::LLMEvaluator;

pub use llm::LLMMetrics;

pub use llm::PromptId;

pub use llm::PromptVersion;

pub use params::ParamDiff;

pub use params::ParamLogger;

pub use params::ParamValue;

pub use report::HanseiAnalyzer;

pub use report::IssueSeverity;

pub use report::MetricSummary;

pub use report::PostTrainingReport;

pub use report::TrainingIssue;

pub use report::Trend;

pub use storage::InMemoryStore;

pub use storage::JsonFileStore;

pub use storage::MetricsStore;

pub use storage::StorageError;

pub use storage::StorageResult;

pub use tui::BrailleChart;

pub use tui::GpuTelemetry;

pub use tui::SamplePeek;

pub use tui::TrainingSnapshot;

pub use tui::TrainingState;

pub use tui::TrainingStateWriter;

pub use tui::TrainingStatus;

pub use tui::TuiMonitor;

pub use tui::TuiMonitorConfig;

pub use wasm::WasmDashboard;

pub use wasm::WasmDashboardOptions;

pub use wasm::WasmMetricsCollector;

Modules

andon

Andon Alerting System (ENT-045)

dashboard

Dashboard Module (ENT-043)

drift

Drift Detection Module (ENT-044)

export

Metrics Export Module (ENT-047)

gpu

GPU Monitoring Module (MLOPS-005)

inference

Real-Time Audit Log & Explainability for APR Format Models

lineage

Model Lineage Tracking (ENT-046)

llm

LLM Evaluation Metrics Module (#71)

params

Parameter Logging API (GH-73)

prometheus

Prometheus Metrics Export Module (MLOPS-006)

report

Hansei (反省) Post-Training Report Generator

storage

Metrics Storage Module

tui

Detached TUI Monitor (SPEC-FT-001 Section 10)

wasm

WebAssembly bindings for training monitor.

Structs

MetricRecord

A single metric record with timestamp

MetricStats

Statistical summary for a single metric

MetricsCollector

Collects training metrics with SIMD-accelerated aggregation

Enums

Metric

Standard training metrics

Type Aliases

MetricsSummary

Type alias for metrics summary