Hugging Face's logo
iliass ayaou's picture
3 1

iliass ayaou

datalyes
Roman190928's profile picture sbrandeis's profile picture cbobed's profile picture
ยท

AI & ML interests

information retrieval, patent retrieval, knowledge management, data engineering and architecture, NLP

Recent Activity

new activity 2 days ago
datalyes/DAPFAM_patent:Add task category: text-retrieval
reacted to nouamanetazi's post with ๐Ÿš€ 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
reacted to piercus's post with ๐Ÿ‘€ about 2 months ago
Starts erasing! ๐ŸŽ‰ ๐ŸŽ‰ ๐ŸŽ‰ This is made with a one-step SD1.5 LBM [1]ย eraser ! Data is open. Data pipeline is open. Training code is open. On our LBM fork : https://github.com/finegrain-ai/LBM [1] https://huggingface.co/papers/2503.07535
View all activity

Organizations

datalyesorg's profile picture