mlbenchkit 0.1.0

Cross-platform ML benchmarking toolkit (CPU/GPU + system specs) for Windows and macOS.

MachineLearningBenchMarkingToolkit

Cross-platform machine-learning benchmarking toolkit to compare Windows PCs and MacBook (Pro/Air) machines. It reports system specs (CPU cores/frequency, RAM) and runs lightweight ML-style benchmarks on:

• CPU (always)
• NVIDIA CUDA GPU (Windows/Linux, if available)
• Apple Silicon GPU via MPS (macOS, if available)

Outputs are saved as JSON files so you can easily compare multiple machines.

Features

• ✅ Machine specs: hostname, OS, CPU model, cores, CPU freq, RAM totals/available
• ✅ Accelerator info:
  • CUDA: GPU name, total VRAM, compute capability
  • MPS: Apple Silicon (unified memory note)
• ✅ Benchmarks:
  • PyTorch matmul on CPU
  • PyTorch matmul on CUDA/MPS (if available)
  • Optional: scikit-learn RandomForest training benchmark

Installation

Minimal

pip install mlbenchkit

With PyTorch benchmarks

# With PyTorch benchmarks
pip install "mlbenchkit[torch]"

With scikit-learn benchmark

pip install "mlbenchkit[sklearn]"

Everything

pip install "mlbenchkit[torch,sklearn]"

## How to use it?

```bash
mlbench specs

Run benchmark suite (saves a JSON file):

mlbench run
# Run with sklearn benchmark:
mlbench run --with-sklearn
## Customize workload:
mlbench run --cpu-N 2048 --gpu-N 4096 --iters 50 --warmup 20 --gpu-dtype float16