muutils, stylized as "$\mu$utils" or "μutils", is a collection of miscellaneous python utilities, meant to be small and with no dependencies outside of standard python.
PyPi: muutils
pip install muutils
Optional dependencies:
pip install muutils[array] # numpy, torch, jaxtyping -- for mlutils, tensor_utils, tensor_info, ml, json_serialize array features
pip install muutils[notebook] # ipython -- for nbutils.configure_notebook
pip install muutils[parallel] # multiprocess, tqdm -- for parallel processing with progress
pip install muutils[web] # weasyprint -- for web/html_to_pdf
| Module | Description |
|---|---|
statcounter | Extension of collections.Counter with smart stats computation (mean, variance, percentiles) |
dictmagic | Dictionary utilities: dotlist conversion, DefaulterDict, tensor dict condensing |
kappa | Anonymous getitem (Kappa(lambda x: x**2)[2] returns 4) |
sysinfo | System information collection for logging |
misc | Utilities: stable_hash, list_join/list_split, filename sanitization, freeze |
interval | Mathematical intervals (open/closed/half-open) with containment, clamping, set operations |
errormode | Enum-based error handling (raise/warn/log/ignore) |
validate_type | Runtime type validation for basic and generic types |
console_unicode | Safe console output with Unicode/ASCII fallback |
spinner | Animated spinners with elapsed time and status updates |
timeit_fancy | Enhanced timing with multiple runs, profiling, and statistics |
dbg | Debug printing inspired by Rust's dbg! macro |
collect_warnings | Context manager to capture and summarize warnings |
parallel | Simplified parallel processing with progress bars |
jsonlines | Simple jsonl file reading/writing |
group_equiv | Group elements by equivalence relation (non-transitive) |
json_serialize | Serialize arbitrary Python objects to JSON (works with ZANJ) |
nbutils | Jupyter utilities: notebook conversion, configuration, mermaid/TeX display |
math | Binning functions and matrix power computation |
cli | CLI utilities: boolean argument parsing, flag actions |
web | HTML asset inlining for standalone documents |
logger | (deprecated) Logging framework, use trnbl instead |
mlutils | ML pipeline: device detection, seeding, checkpoints (requires array) |
tensor_utils | PyTorch/numpy type conversions (requires array) |
tensor_info | Tensor metadata extraction and formatting (requires array) |
ml | CUDA memory monitoring (requires array) |
ZANJ is a human-readable and simple format for ML models, datasets, and arbitrary objects. It's built around having a zip file with json and npy files, and has been spun off into its own project.
There are a couple work-in-progress utilities in _wip that aren't ready for anything, but nothing in this repo is suitable for production. Use at your own risk!