Hackernews posts about bfloat16

1. llm.c: multi-GPU, bfloat16, flash attention, ~7% faster than PyTorch (twitter.com)
   121 points by tosh 7 months ago | 10 comments
2. Bfloat16 support coming to Apple's Metal and PyTorch [video] (developer.apple.com)
   93 points by dlewis1788 over 1 year ago | 53 comments
3. Supporting BFLOAT16 in Fortran: "Not Recommended"? (fortran-lang.discourse.group)
   7 points by ggraphilia 8 months ago | 2 comments
4. BFloat16: The secret to high performance on Cloud TPUs (cloud.google.com)
   3 points by mfiguiere 10 months ago | 1 comments
5. Brain Floating Point (Bfloat16) (en.wikipedia.org)
   3 points by ml_basics over 1 year ago | 1 comments
6. RISC-V Support for BFloat16 (fprox.substack.com)
   3 points by camel-cdr about 1 year ago | discuss
7. llama.cpp bfloat16 support (github.com)
   2 points by indigodaddy 7 months ago | discuss
8. Supporting BFLOAT16 in ISO_Fortran_env: not recommended (j3-fortran.org)
   1 points by zaikunzhang 8 months ago | 7 comments
9. Converging Harmonic Series with bfloat16 (github.com)
   1 points by octocop about 1 year ago | 1 comments
10. Show HN: Dive into Transformers and LLM World – Llama 3.1 in Go, Step by Step (github.com)
    4 points by adalkiran 3 months ago | 2 comments
11. Show HN: LLaMA Nuts and Bolts, A holistic way of understanding how LLMs run (github.com)
    3 points by adalkiran 8 months ago | discuss
12. Show HN: Retrieval Augmented Generation Optimised LLM's (huggingface.co)
    2 points by alextttty over 1 year ago | discuss
13. Ask HN: Clarify VRAM usage during LLM forward pass
    1 points by furiousteabag 12 months ago | discuss
14. Show HN: Model2Vec: make sentence transformers 500x faster on CPU, 15x smaller (github.com)
    6 points by stephantul 2 months ago | 2 comments
15. Tell HN: I learnt some new things while working on my project
    1 points by behnamoh over 1 year ago | discuss