How arefin994/mlbench-lite works
mlbench-lite is a Python library and CLI/TUI for benchmarking classification and regression models on tabular datasets. The main workflow accepts built-in or user-provided data, selects a task and models, optionally recommends or applies scaling, evaluates models by holdout or cross-validation, and returns or saves metric results. README capabilities include model discovery, GPU/deep-model options, reproducibility, and statistical comparisons. Some internals of benchmark.py and CLI/TUI handlers were not sampled; only documented or excerpt-supported connections are shown.
Main entry points: benchmark.py Benchmark runner · tui.py Interactive TUI · cli.py Command-line interface · __init__.py Python API
Components 15
User interfaces
- Python API
mlbench_lite/__init__.py - Command-line interface
mlbench_lite/cli.py - Interactive TUI
mlbench_lite/tui.py
Benchmarking
- Benchmark runner
mlbench_lite/benchmark.py - Model discovery
mlbench_lite/benchmark.py - Scaling tools
mlbench_lite/preprocessing.py
Data and analysis
- Clover dataset
mlbench_lite/datasets.py - Significance tests
mlbench_lite/stats.py - Reproducibility
mlbench_lite/utils.py - Pandas results
Model integrations
- Deep tabular models
mlbench_lite/deep_models.py - ML estimator libraries
Other components
- Python / CLI / TUI user
- User dataset
- CSV / JSON output
Connections 27
- Python / CLI / TUI user → Python API: calls
- Python / CLI / TUI user → Command-line interface: invokes
- Python / CLI / TUI user → Interactive TUI: invokes
- Python API → Benchmark runner: exports
- Python API → Clover dataset: exports
- Python API → Scaling tools: exports
- Python API → Significance tests: exports
- Python API → Reproducibility: exports
- Command-line interface → User dataset: loads
- Command-line interface → Benchmark runner: calls
- Command-line interface → Model discovery: lists
- Command-line interface → Pandas results: prints
- Command-line interface → CSV / JSON output: saves
- Interactive TUI → Command-line interface: loads demo
- Interactive TUI → Benchmark runner: calls
- Interactive TUI → Scaling tools: requests advice
- Interactive TUI → Model discovery: lists
- Interactive TUI → Pandas results: displays
- Interactive TUI → CSV / JSON output: saves
- Benchmark runner → Scaling tools: applies scaling
- Benchmark runner → Reproducibility: seeds
- Benchmark runner → Model discovery: selects
- Benchmark runner → ML estimator libraries: evaluates with
- Benchmark runner → Deep tabular models: uses optional models
- Benchmark runner → Pandas results: returns
- Significance tests → Pandas results: reads fold scores
- Clover dataset → Python / CLI / TUI user: provides data