Moe
NVIDIA details three inference optimizations for DGX Spark
NVIDIA says three DGX Spark optimizations lifted Ling-3.0-flash throughput and cut latency in GSM8K testing, with a caveat on stacking them.
Moe
NVIDIA's Transformer Engine Boosts MoE Training in JAX by 10x
NVIDIA's Transformer Engine accelerates Dropless Mixture-of-Experts (MoE) training in JAX, achieving a 10x performance gain and 97% scaling efficiency.
Moe
NVIDIA Enhances PyTorch with NeMo Automodel for Efficient MoE Training
NVIDIA introduces NeMo Automodel to facilitate large-scale mixture-of-experts (MoE) model training in PyTorch, offering enhanced efficiency, accessibility, and scalability for developers.