Megatron-LM
Novita Sandbox Integrates With Miles for Agentic RL Training
Novita Sandbox is now a listed sandbox provider in Miles, RadixArk’s open-source framework for large-scale reinforcement learning and model post-training. The…
Lila Sciences and Anyscale Train GLM-5.3-Flash on 1M-Token Sequences
Lila Sciences and Anyscale have trained the open model GLM-5.3-Flash on sequences of 1 million tokens using 64 B200 GPUs. The teams say the work extends their training…
NVIDIA Cuts Cost of Bitwise-Deterministic Pretraining in Megatron Core to About 2%
NVIDIA says it has reduced the overhead of bitwise-deterministic pretraining in Megatron Core to approximately 2% at 2,432 GPUs, while keeping runs bitwise identical…
CoreWeave reports 51-52% MFU and 3.66-day MTTF in 1,024-GPU H100 test
CoreWeave says a six-week benchmark of up to 1,024 NVIDIA H100 GPUs reached 51-52% Model FLOPs Utilization (MFU) and a mean time to failure (MTTF) of 3.66 days. The…