Hugging Face's logo
neuralink's picture
In a Training Loop ๐Ÿ”„
7 34 84

neuralink

neuralink
ltim's profile picture Leyo's profile picture clem's profile picture
ยท
https://phucnguyen.dev

AI & ML interests

distributed training @nous research. ex-nanotron @ hf

Recent Activity

liked a model 25 days ago
Qwen/Qwen3-30B-A3B
upvoted a paper about 2 months ago
R-Horizon: How Far Can Your Large Reasoning Model Really Go in Breadth and Depth?
reacted to nouamanetazi's post with ๐Ÿค— about 2 months ago
After training ๐’๐ฆ๐จ๐ฅ๐‹๐Œ๐Ÿ‘ on ๐Ÿ‘๐Ÿ–๐Ÿ’ ๐‡๐Ÿ๐ŸŽ๐ŸŽ๐ฌ for nearly a month, I've come to realize something most people overlook: ๐ข๐ง๐Ÿ๐ซ๐š๐ฌ๐ญ๐ซ๐ฎ๐œ๐ญ๐ฎ๐ซ๐ž ๐ข๐ฌ ๐ญ๐ก๐ž ๐ฆ๐š๐ค๐ž-๐จ๐ซ-๐›๐ซ๐ž๐š๐ค ๐Ÿ๐š๐œ๐ญ๐จ๐ซ ๐ข๐ง ๐‹๐‹๐Œ ๐ญ๐ซ๐š๐ข๐ง๐ข๐ง๐ . ๐Ÿ”ฅ Everyone talks about model architecture and data quality. And yes, those matter immensely. But here's what nobody tells you: when your training run fails at 2 AM because of mysterious ๐๐‚๐‚๐‹ ๐ž๐ซ๐ซ๐จ๐ซ๐ฌ, or when your expensive GPU cluster is running at ๐Ÿ”๐ŸŽ% ๐ž๐Ÿ๐Ÿ๐ข๐œ๐ข๐ž๐ง๐œ๐ฒ, the problem isn't your model. It's most probably a ๐ฆ๐ข๐ฌ๐ฎ๐ฌ๐ž ๐จ๐Ÿ ๐ญ๐ก๐ž ๐ก๐š๐ซ๐๐ฐ๐š๐ซ๐ž. ๐Ÿ› ๏ธ Questions that seemed simple but had no clear answers: Why is ๐Œ๐จ๐„ ๐ญ๐ซ๐š๐ข๐ง๐ข๐ง๐  ๐ฌ๐ฅ๐จ๐ฐ๐ž๐ซ ๐ญ๐ก๐š๐ง ๐๐ž๐ง๐ฌ๐ž ๐ฆ๐จ๐๐ž๐ฅ๐ฌ? Which ๐๐‚๐‚๐‹ ๐Ÿ๐ฅ๐š๐ ๐ฌ should we actually set? How often should we checkpoint without killing throughput? That's why we built ๐“๐ก๐ž ๐’๐ฆ๐จ๐ฅ ๐“๐ซ๐š๐ข๐ง๐ข๐ง๐  ๐๐ฅ๐š๐ฒ๐›๐จ๐จ๐ค ๐Ÿ“–: a complete guide covering everything from model architecture and data curation to the SmolLM3 training marathon, post-training techniques, and crucially, the ๐ข๐ง๐Ÿ๐ซ๐š๐ฌ๐ญ๐ซ๐ฎ๐œ๐ญ๐ฎ๐ซ๐ž ๐ฅ๐š๐ฒ๐ž๐ซ that most teams get wrong. We validated real vs theoretical bandwidth across the entire stack: ๐‡๐๐Œ๐Ÿ‘ ๐ก๐ข๐ญ๐ญ๐ข๐ง๐  ๐Ÿ‘ ๐“๐/๐ฌ, ๐๐•๐‹๐ข๐ง๐ค ๐Ÿ’.๐ŸŽ ๐ซ๐ž๐š๐œ๐ก๐ข๐ง๐  ๐Ÿ•๐Ÿ–๐Ÿ” ๐†๐/๐ฌ, ๐๐‚๐ˆ๐ž ๐†๐ž๐ง๐Ÿ’ ๐š๐ญ ๐Ÿ๐Ÿ’.๐Ÿ ๐†๐/๐ฌ. Then we ran collective operations across ๐Ÿ๐Ÿ๐Ÿ– ๐†๐๐”๐ฌ (16 nodes, 8xH100s each) and measured how performance degrades at scale: all-reduce drops from ๐Ÿ’๐Ÿ–๐ŸŽ ๐†๐/๐ฌ on a single node to ๐Ÿ‘๐Ÿ๐ŸŽ-๐Ÿ‘๐Ÿ“๐ŸŽ ๐†๐/๐ฌ across 16 nodes. If you've ever wondered why your training runs are slower than they should be, or you're planning to scale up and want to avoid expensive mistakes, this guide might save you weeks of debugging. ๐“๐ก๐ž ๐’๐ฆ๐จ๐ฅ ๐“๐ซ๐š๐ข๐ง๐ข๐ง๐  ๐๐ฅ๐š๐ฒ๐›๐จ๐จ๐ค: https://lnkd.in/e5MKXUHS Shared with โค๏ธ by the HuggingFace team
View all activity

Organizations

NousResearch's profile picture BigCode's profile picture wut?'s profile picture