bk99.de entertain the web since 1997

bk99.de
HF · MLOpsCurated

Machine learning as versionable code

Summary

The article describes the shift from isolated models to collaborative, versioned artefacts. Models, data sets and applications are meant to be developed with working methods similar to those for source code. The Hub uses Git-based repositories for models and data sets.

Ideas

  • Traceable ML work needs versions for data and weights, not just for Python code.
  • Collaboration becomes easier when artefacts remain directly executable and verifiable.

Insights

  • Reproducibility requires immutable references to all artefacts, not just to source code.

Facts

  • Model cards document purpose, limits and evaluation.

Critique

  • The code metaphor underestimates privacy, data provenance and the high cost of binary artefacts.

Recommendations

  • Version the data reference, model revision, configuration and evaluation code in one reproducible run.

References

Read the original article on Hugging Face

Search the Web Archive

HF · 2025

An MCP agent in fifty lines of code

Tiny Agents shows a deliberately small agent that discovers and calls tools via the Model Context Protocol. The code makes the planning and execution loop directly comprehensible. The example implementation is about 50 lines of code.

HF · 2024

Simple agents that write actions as code

smolagents reduces agent logic to small, comprehensible components and lets models control tools through generated code. The approach favours readability over extensive orchestration frameworks. smolagents supports code agents and classic tool calls.

HF · 2024

Quantising and accelerating StarCoder on Xeon

The article combines 8-bit and 4-bit quantisation with speculative decoding for StarCoder on Xeon processors. It shows several independent routes to faster code generation. StarCoder is optimised for CPU inference with Optimum Intel.