CLI Reference
This document provides a complete reference for all warpt commands, options, and available tests.
warpt offers a consistent command-line interface for hardware discovery, stress testing, monitoring, and benchmarking across diverse computational environments.
Commands Overview
| Command | Description |
|---|---|
warpt list |
List CPU, GPU, memory, storage, and framework information |
warpt stress |
Run hardware stress tests |
warpt monitor |
Real-time system monitoring |
warpt power |
Power consumption monitoring |
warpt benchmark |
Performance benchmarking |
warpt carbon |
Energy and carbon tracking |
warpt integrate |
AI-assisted hardware backend integration |
warpt version |
Display version information |
warpt list
List CPU and GPU information, installed frameworks, and system details.
Usage:
warpt list [OPTIONS]
Options:
| Option | Description |
|---|---|
--export |
Export results to JSON file with default filename |
--export-file TEXT |
Export results to JSON file with custom filename |
--help |
Show help message |
Examples:
| Command | Description |
|---|---|
warpt list |
Display all system info |
warpt list --export |
Export to JSON with auto-generated filename |
warpt list --export-file my.json |
Export to specific file |
Output includes:
- CPU information (make, model, architecture, cores, frequencies)
- GPU information (if NVIDIA GPU detected)
- CUDA information
- Docker status
- ML frameworks (PyTorch, TensorFlow, JAX, NumPy, etc.)
- Core libraries (MKL, OpenBLAS, cuDNN, etc.)
- Memory information
- Storage devices
warpt stress
Run hardware stress tests across CPU, GPU, RAM, storage, and network.
Usage:
warpt stress [OPTIONS]
Options:
| Option | Description |
|---|---|
-l, --list |
List available stress tests |
-c, --category TEXT |
Filter by category: cpu, accelerator, ram, storage, network, all |
-t, --test TEXT |
Run specific test(s) by name (e.g. GPUFP32ComputeTest) |
-d, --duration INTEGER |
Duration in seconds per test (default: 30) |
-w, --warmup INTEGER |
Warmup period in seconds (default: 5) |
--device-id TEXT |
Device ID(s) for accelerator tests (comma-separated) |
-o, --output TEXT |
Output file(s) - format auto-detected (.json/.yaml). Repeatable. |
| `-f, --format [json | yaml |
--config PATH |
YAML config file with per-test settings |
-v, --verbose |
Verbose output |
--target TEXT |
Target IP(s) for network tests (comma-separated) |
--payload INTEGER |
Payload size in bytes for network tests |
| `--network-mode [latency | bandwidth |
--read-ratio FLOAT |
Read ratio for StorageMixedTest (0.0-1.0) |
--help |
Show help message |
Examples:
| Command | Description |
|---|---|
warpt stress --list |
List available tests |
warpt stress --list -c cpu |
List CPU tests only |
warpt stress -c all |
Run all available tests |
warpt stress -c cpu |
Run CPU category, all tests |
warpt stress -t GPUFP32ComputeTest |
Run specific test |
warpt stress -o results.json |
Save to JSON |
Available Stress Tests
CPU
| Test | Description |
|---|---|
CPUHashingTest |
CPU hashing performance |
CPUMatMulTest |
Matrix multiplication (NumPy/BLAS) |
CPUZlibTest |
Compression performance |
Accelerator (requires NVIDIA GPU)
| Test | Description |
|---|---|
GPUCFDSimulationTest |
CFD simulation workload |
GPUFP64ComputeTest |
FP64 compute performance |
GPUFP32ComputeTest |
FP32 matrix multiplication (TF32 supported) |
GPUMemoryBandwidthTest |
GPU memory bandwidth |
GPUPrecisionTest |
Multi-precision testing |
RAM
| Test | Description |
|---|---|
RAMBandwidthTest |
Memory bandwidth |
RAMLatencyTest |
Memory latency |
RAMSwapPressureTest |
Swap behavior under pressure |
RAMWorkloadSimulationTest |
Realistic workload patterns |
Storage
| Test | Description |
|---|---|
StorageMixedTest |
Mixed read/write patterns |
StorageRandomReadTest |
Random read performance |
StorageRandomWriteTest |
Random write performance |
StorageSequentialReadTest |
Sequential read performance |
StorageSequentialWriteTest |
Sequential write performance |
Network
| Test | Description |
|---|---|
NetworkBidirectionalTest |
Bidirectional throughput (requires remote server) |
NetworkLoopbackTest |
Local loopback latency |
NetworkPointToPointTest |
Point-to-point testing (requires remote server) |
warpt monitor
Monitor system performance in real-time with a TUI dashboard or CLI output.
Usage:
warpt monitor [OPTIONS]
Options:
| Option | Description |
|---|---|
-i, --interval FLOAT |
Sampling interval in seconds (default: 1.0) |
-d, --duration FLOAT |
Stop after this many seconds (default: run until interrupted) |
--no-tui |
Run CLI output instead of curses dashboard |
--help |
Show help message |
warpt power
Monitor system power consumption for CPU, GPU, and other components.
Usage:
warpt power [OPTIONS]
Options:
| Option | Description |
|---|---|
-i, --interval FLOAT |
Sampling interval in seconds (default: 1.0) |
-d, --duration FLOAT |
Stop after this many seconds (default: run until interrupted) |
--no-processes |
Don't show per-process power attribution |
-n, --top INTEGER |
Number of top processes to display (default: 10) |
--json |
Output in JSON format |
-o, --output TEXT |
Write results to JSON file |
-c, --continuous |
Run continuously (vs single snapshot) |
--sources |
Show available power sources and exit |
--help |
Show help message |
Platform support:
- Linux: Intel/AMD RAPL via
/sys/class/powercap/ - macOS:
powermetrics(requiressudo) - All: NVIDIA GPUs via NVML
warpt benchmark
Run and manage performance benchmarks.
Usage:
warpt benchmark [OPTIONS] [list|validate|build|run]
Subcommands:
| Subcommand | Description |
|---|---|
list |
List available benchmarks |
validate |
Validate configuration files |
build |
Build benchmark container |
run |
Run a benchmark |
Options:
| Option | Description |
|---|---|
-b, --benchmark TEXT |
Benchmark name (e.g., hpl) |
--system PATH |
System configuration file (JSON, YAML, or HCL) |
--benchmark-config PATH |
Benchmark configuration file |
--cluster PATH |
Cluster configuration file |
--set TEXT |
Override configuration value (KEY=VALUE format) |
--dry-run |
Validate and print actions without executing |
--debug |
Enable debug output and stack traces |
-o, --output PATH |
Write benchmark results to file |
--force |
Force rebuild even if cached image exists |
--no-monitor |
Disable power and resource monitoring |
--help |
Show help message |
warpt carbon
Track energy consumption, CO2 emissions, and cost for any workload.
Usage:
warpt carbon [SUBCOMMAND] [OPTIONS]
Subcommands:
| Subcommand | Description |
|---|---|
start |
Start background energy tracking |
stop |
Stop tracking and show results |
status |
Show daemon status (default if no subcommand) |
history |
Show recent tracking sessions |
summary |
Aggregate summary over time window |
regions |
List available grid regions and carbon intensities |
set-region |
Set your grid region (persisted to ~/.warpt/config.json) |
intensity |
Set a custom carbon intensity in gCO2/kWh |
kwh-price |
Set electricity price in USD per kWh |
Options:
| Option | Description |
|---|---|
-l, --label TEXT |
Session label for start |
-i, --interval FLOAT |
Sampling interval in seconds |
-n, --limit INTEGER |
Max sessions to show in history (default: 20) |
-d, --days INTEGER |
Time window in days for summary (default: 30) |
--json |
Output in JSON format |
--value TEXT |
Value for set-region (region code), intensity (gCO2/kWh), or kwh-price ($/kWh) |
--help |
Show help message |
Examples:
| Command | Description |
|---|---|
warpt carbon set-region --value EU-DE |
Set grid region to Germany |
warpt carbon intensity --value 500 |
Set custom carbon intensity |
warpt carbon kwh-price --value 0.21 |
Set electricity rate |
warpt carbon start -l "training" |
Start tracking with a label |
warpt carbon stop |
Stop and print results |
warpt carbon history |
Show recent sessions |
warpt carbon summary -d 7 |
Last 7 days summary |
warpt carbon regions |
List grid regions |
warpt integrate
AI-assisted hardware backend integration. Uses Claude Code to generate backend code from vendor SDK documentation.
Usage:
warpt integrate [OPTIONS] [status|validate|reset]
Subcommands:
| Subcommand | Description |
|---|---|
status |
Check integration status |
validate |
Run linting and tests on generated code |
reset |
Reset and start over |
Options:
| Option | Description |
|---|---|
--vendor TEXT |
Vendor name (lowercase, underscores allowed) |
--sdk-docs PATH |
Path to SDK docs (local dir, PDF, or URL) |
--vendor-context TEXT |
Optional notes about the hardware |
--dry-run |
Validate setup without starting agent |
--help |
Show help message |
Examples:
| Command | Description |
|---|---|
warpt integrate --vendor amd --sdk-docs ./rocm-docs/ |
Start new integration |
warpt integrate |
Continue existing integration |
warpt integrate --dry-run |
Validate without running |
warpt integrate validate |
Lint and test generated code |
warpt integrate reset |
Start over |
warpt version
Display warpt version information.
Usage:
warpt version [OPTIONS]
Options:
| Option | Description |
|---|---|
-v, --verbose |
Show detailed version information |
--help |
Show help message |