YAML Examples Catalog

This page catalogs all 30 YAML configuration examples, organized by category. Each example demonstrates a specific training scenario.

Overview

Section A: Basic Training & Data

mnist_cpu.yaml

MNIST baseline training on CPU.

entrenar train examples/yaml/mnist_cpu.yaml

QA Focus: Verify alimentar downloads and caches correctly.

csv_data.yaml

Training on local CSV tabular data.

entrenar train examples/yaml/csv_data.yaml

QA Focus: CSV parsing robustness (headers, types).

parquet_data.yaml

High-throughput columnar data loading.

entrenar train examples/yaml/parquet_data.yaml

QA Focus: Parquet read performance.

multiworker.yaml

Multi-worker data loading.

entrenar train examples/yaml/multiworker.yaml

QA Focus: No data corruption with parallel workers.

dropout.yaml

Regularization with dropout layers.

entrenar train examples/yaml/dropout.yaml

QA Focus: Dropout disabled during validation.

deterministic.yaml

Bit-exact reproducible training.

entrenar train examples/yaml/deterministic.yaml

QA Focus: Same seed produces identical results.

Section B: Compiler-in-the-Loop (CITL)

citl_suggest.yaml

CITL optimization suggestions.

entrenar train examples/yaml/citl_suggest.yaml

QA Focus: Suggestions are actionable.

citl_workspace.yaml

CITL workspace management.

entrenar train examples/yaml/citl_workspace.yaml

QA Focus: Workspace isolation.

Section C: Model Architecture

custom_arch.yaml

Custom model architecture definition.

entrenar train examples/yaml/custom_arch.yaml

QA Focus: Layer connections validated.

llama2_mock.yaml

LLaMA-2 mock model for testing.

entrenar train examples/yaml/llama2_mock.yaml

QA Focus: Architecture matches real LLaMA.

lora.yaml

LoRA fine-tuning configuration.

entrenar train examples/yaml/lora.yaml

QA Focus: Only adapter weights updated.

qlora.yaml

QLoRA 4-bit fine-tuning.

entrenar train examples/yaml/qlora.yaml

QA Focus: VRAM usage < 50% of full fine-tune.

Section D: Optimization & Schedulers

grad_clip.yaml

Gradient clipping for stability.

entrenar train examples/yaml/grad_clip.yaml

QA Focus: Gradient norms bounded.

grad_accum.yaml

Gradient accumulation for large effective batch.

entrenar train examples/yaml/grad_accum.yaml

QA Focus: Accumulation count matches config.

lr_schedule.yaml

Learning rate scheduling (cosine).

entrenar train examples/yaml/lr_schedule.yaml

QA Focus: LR follows expected curve.

distillation.yaml

Knowledge distillation from teacher.

entrenar train examples/yaml/distillation.yaml

QA Focus: Student approaches teacher quality.

Section E: Monitoring & Alerts

andon.yaml

Andon alerting system (Jidoka).

entrenar train examples/yaml/andon.yaml

QA Focus: Alerts trigger on anomalies.

outlier.yaml

Outlier detection during training.

entrenar train examples/yaml/outlier.yaml

QA Focus: Outliers flagged, not silently ignored.

bias.yaml

Bias detection and mitigation.

entrenar train examples/yaml/bias.yaml

QA Focus: Demographic parity metrics tracked.

drift.yaml

Data/model drift detection.

entrenar train examples/yaml/drift.yaml

QA Focus: Drift alerts when distribution shifts.

Section F: Reliability & Checkpoints

checkpoint.yaml

Checkpoint saving and resumption.

entrenar train examples/yaml/checkpoint.yaml

QA Focus: Resume from checkpoint is exact.

config_validate.yaml

Strict configuration validation.

entrenar validate examples/yaml/config_validate.yaml

QA Focus: Invalid configs rejected early.

long_run.yaml

Extended training duration test.

entrenar train examples/yaml/long_run.yaml

QA Focus: No memory leaks over hours.

locked.yaml

Lockfile for reproducibility.

entrenar train examples/yaml/locked.yaml

QA Focus: Lockfile pins all dependencies.

Section G: Inference & Output

latency.yaml

Inference latency benchmarking.

entrenar bench examples/yaml/latency.yaml

QA Focus: Latency meets SLA.

json_output.yaml

JSON format output generation.

entrenar train examples/yaml/json_output.yaml

QA Focus: JSON is valid and complete.

Section H: Research & Privacy

dp.yaml

Differential privacy training.

entrenar train examples/yaml/dp.yaml

QA Focus: Privacy budget (epsilon) tracked.

release.yaml

Production release configuration.

entrenar train examples/yaml/release.yaml

QA Focus: All 25 QA points pass.

Section I: Ecosystem Integration

session.yaml

Session management with Ruchy.

entrenar train examples/yaml/session.yaml

QA Focus: Session state persists correctly.

Section J: Edge Cases

soak.yaml

Soak test for extended stability.

entrenar train examples/yaml/soak.yaml

QA Focus: System stable over extended period.

Running All Examples

Validation Only

# Validate all YAML configs
for f in examples/yaml/*.yaml; do
  echo "Validating $f..."
  entrenar validate "$f"
done

Integration Tests

# Run all integration tests
cargo test --test yaml_mode_integration

Quick Reference Table

Next Steps

• QA Process - 25-point checklist
• YAML Mode Training - Complete schema reference
• Schema Reference - All configuration options