rawsh | Dark Hacker News

1.	Flash-MSA: Accelerating Million-Token Training with Sparse Attention Kernels(nanduruganesh.github.io) 41 points by rawsh 48 days ago \| 5 comments
2.	Zyphra ZAYA1-base: First large-scale model trained on AMD(zyphra.com) 6 points by rawsh 278 days ago \| 1 comment
3.	278 days ago \| discuss
4.	Debugging divergence between engine and transformers logprobs for RL(gist.github.com) 2 points by rawsh 349 days ago \| 0 comments
5.	Batched reward model inference and Best-of-N sampling(raw.sh) 34 points by rawsh 1 year ago \| 0 comments
6.	Teaching LLMs to solve chess puzzles with DSPy and Finetuning(raw.sh) 1 points by rawsh 1 year ago \| 0 comments
7.	Teaching chat models to solve chess puzzles(raw.sh) 4 points by rawsh 2 years ago \| 0 comments
8.	Ask HN: Why aren't there Open source embedding models with context length > 512? 3 points by rawsh 2 years ago \| 2 comments
9.	Show HN: DankGPT – Chat with Your Documents(dankgpt.com) 17 points by rawsh 3 years ago \| 9 comments
10.	Show HN: Search PDFs in the browser using PDFgrep compiled to WebAssembly(pdfgrep.com) 4 points by rawsh 3 years ago \| 1 comment
11.	Show HN: Search PDFs using WASM in the browser(pdfgrep.com) 2 points by rawsh 3 years ago \| 0 comments