view article Article Async GRPO with LoRA across HF Jobs: a bucket, a proxy, and no NCCL +2 aminediroHF, qgallouedec, kashif, sergiopaniego • 11 days ago • 44
view article Article Making open-source AI weather forecasting models easy to run hugging-science • 13 days ago • 37
Running 10 Why Adaptive Optimizers Exist — a Tangible lesson. 🏔 10 Why momentum/AdamW can outperform SGD — a Tangible lesson.