HN Explorer

stephantul

↗ HN profile [ 67.2 ms ] full profile
Karma
0
Created
()
Submissions
0
  1. From Chesterton's fence to Chesterton's gap (stephantul.github.io)
    86 points 11
  2. Why scikit learn's fit transform is probably not for you (stephantul.github.io)
    1 points 0
  3. Show HN: Semble – Code search for agents that uses 98% fewer tokens than grep (github.com)

    Hey HN! We (Stephan and Thomas) recently open-sourced Semble. We kept running into the same problem while using Claude Code on large codebases: when the agent can't find something directly, it falls back to grep,…

    8 points 0
  4. Show HN: Semble – Fast code search for agents with near-transformer accuracy (github.com)

    Hey HN! We've just open-sourced Semble, a fast and accurate code search library built for agents. We're also releasing potion-code-16M, a small code-specialized static embedding model that powers it. Most…

    7 points 0
  5. Show HN: Skeletoken, a Python package for editing model tokenizers (github.com)

    Hello! I just released a new version of Skeletoken, a package for editing tokenizers. New in this version is the ability to automatically adapt a model to an edited tokenizer. For example, you can a new token to your…

    1 points 0
  6. Show HN: PyNIFE. 400-900× speedup for embedding-based retrieval pipelines (github.com)

    Hey HN, I've been playing around with ways to make retrieval pipelines faster, and ended up building something I'm calling PyNIFE (Nearly Inference-Free Embeddings). The idea is simple: train a static embedding model…

    2 points 0
  7. Show HN: Skeletoken, a Package for Editing Tokenizers (github.com)

    Hello! I work on Hugging Face tokenizers a lot in my day job. Editing tokenizers, e.g., adding or removing tokens is super painful. This is why I wrote a library for working with the format. It contains many useful…

    1 points 0
  8. Turning any tokenizer into a greedy one (stephantul.github.io)
    2 points 1
  9. Decasing Transformers for Fun (stephantul.github.io)
    3 points 1
  10. Model2Vec as a Fasttext Alternative (minish.ai)
    5 points 1
  11. Using overloads to handle union return types in Python (stephantul.github.io)
    1 points 1
  12. Ask HN: Favourite resources for learning programming type theory?

    Hello HN, Recently, I've been massively enjoying diving into type checkers and how to effectively use them (mostly in Python). I lack foundational knowledge about the topic (I never studied CS), however, so I am looking…

    6 points 8
  13. Evaluating ML classifiers using relative error instead of absolute accuracy (stephantul.github.io)
    1 points 0
  14. Defeat stringly typing without making your users unhappy (stephantul.github.io)
    2 points 0
  15. Distilling ModernBERT into a static model doesn't work (minishlab.github.io)
    5 points 3
  16. Show HN: SemHash – Fast Semantic Text Deduplication for Cleaner Datasets (github.com)

    We’ve just open-sourced SemHash, a lightweight package for semantic text deduplication. It lets you effortlessly clean up your datasets and avoid pitfalls caused by duplicate samples in semantic search, RAG, and machine…

    5 points 0
  17. Train faster static embedding models with sentence transformers (huggingface.co)
    52 points 1
  18. Semhash: Fast deduplication and dataset multitool in Python (minishlab.github.io)
    3 points 1
  19. Model2Vec: Make sentence transformers 500x faster on CPU, 15x smaller (huggingface.co)
    5 points 0
  20. Show HN: Model2Vec: make sentence transformers 500x faster on CPU, 15x smaller (github.com)

    Hi HN! We (Thomas and Stéphan, hello!) recently released Model2Vec, a Python library for distilling any sentence transformer into a small set of static embeddings. This makes inference with such a model up to 500x…

    9 points 2
  21. Show HN: Model2Vec: make sentence transformers 500x faster on CPU, 15x smaller (github.com)

    Hi HN! We (Thomas and Stéphan, hello!) recently released Model2Vec, a Python library for distilling any sentence transformer into a small set of static embeddings. This makes inference with such a model up to 500x…

    6 points 2
  22. 1 points 0