bidiptas13.bsky.social @bidiptas13.bsky.social · 21/02/2026Check out our new work on autonomous driving in new cities with map data + MARL! 010
bidiptas13.bsky.social @bidiptas13.bsky.social · 21/11/2025Evolve at the hyperscale! Work co-led with Mattie Fellows and Juan Agustin Duque. Made possible by #Isambard and AIRR 🌐 Website: eshyperscale.github.io 📝 Paper: alphaxiv.org/abs/2511.16652 💻 Code: github.com/ESHyperscale... 🥚NanoEgg : github.com/ESHyperscale... (train in int 😉)eshyperscale.github.ioEvolution Strategies at the HyperscaleGeneral ML Training Made as Fast and Easy as Inference 030
bidiptas13.bsky.social @bidiptas13.bsky.social · 21/11/2025Scaling LLM Reasoning with EGGROLL 🥚🧠📝 Using 🥚 to finetune RWKV-7 language models outperforms GRPO on Countdown and GSM8K ❗ 🥚significantly outperformed GRPO on the Countdown task, achieving a 35% validation accuracy compared to GRPO's 23%❗ 100
bidiptas13.bsky.social @bidiptas13.bsky.social · 21/11/2025EGGROLL 🥚for RL 🎮🤖 🥚 is competitive with, and in many cases, better than OpenES performance, even before considering the vast speed-up! 🥚 matched OpenES on 7/16 environments and outperformed it on another 7/16 🥚's low-rank approach does not compromise ES performance 100
bidiptas13.bsky.social @bidiptas13.bsky.social · 21/11/2025🥚EGGROLLing in the Deep with🚀 💯✕ Speedup 🥚 speed nearly reaches the throughput of pure batch inference, leaving OpenES far behind 🥚 reaches 91% of pure batch inference speed vs. OpenES reaching only 0.41% 100
bidiptas13.bsky.social @bidiptas13.bsky.social · 21/11/2025The EGGROLL Recipe 🧠🛠️ We replace full-rank perturbations with low-rank ones. Each update is still high rank, maintaining expressivity with faster training 🥚 EGGROLL converges to the full-rank update at a fast rate of 1/rank. The method is effective even with a rank of 1 100
bidiptas13.bsky.social @bidiptas13.bsky.social · 21/11/2025We use EGGROLL 🥚to train RNN language models from scratch using only integer datatypes (and no activation functions!), scaling population size from 64 to 262144 2 (🐔🐔) orders of magnitude larger than prior ES works❗ 100
bidiptas13.bsky.social @bidiptas13.bsky.social · 21/11/2025Introducing 🥚EGGROLL 🥚(Evolution Guided General Optimization via Low-rank Learning)! 🚀 Scaling backprop-free Evolution Strategies (ES) for billion-parameter models at large population sizes ⚡100x Training Throughput 🎯Fast Convergence 🔢Pure Int8 Pretraining of RNN LLMs 1268
