Command Line Interface¤
Command-line tools for running and benchmarking Datarax pipelines without writing Python code.
Commands¤
| Command | Purpose | Example |
|---|---|---|
datarax run |
Execute pipeline | datarax run config.yaml |
datarax benchmark |
Measure performance | datarax benchmark pipeline.yaml |
★ Insight ─────────────────────────────────────
- CLI is useful for quick experiments and CI/CD
- Config files define pipelines declaratively
- Benchmark command includes warmup automatically
- Use Python API for complex workflows
─────────────────────────────────────────────────
Quick Start¤
# Run a pipeline from config
datarax run config.yaml
# Benchmark pipeline performance
datarax benchmark config.yaml --batches 100
# Get help
datarax --help
Modules¤
Config File Format¤
# config.yaml
source:
type: hf
name: mnist
split: train
pipeline:
batch_size: 32
transforms:
- type: normalize
- type: augment
probability: 0.5
output:
format: tfrecord
path: ./output
Benchmark Output¤
$ datarax benchmark config.yaml --batches 100
Pipeline Benchmark Results
==========================
Warmup batches: 10
Measured batches: 100
Throughput: 1234.56 samples/sec
Latency (mean): 25.6 ms
Latency (p99): 32.1 ms
See Also¤
- Config - Configuration system
- Benchmarking - Programmatic benchmarking
- Installation - Installing CLI