Meaning
A distributed pipeline scheduling technique that hides inter-GPU network latency behind active forward and backward calculations.
Articles Featuring This Term
Related Terms
- Optical Circuit Switch Fiber-Optic Network Reconfiguration
- Multi-Head Latent Attention Compressed Latent Attention Architecture
- MuJoCo Physics Engine Multi-Joint Dynamics Simulation
- Text-to-Speech (TTS)
- Area Under the ROC Curve (AUC) Overall Classifier Accuracy Score
- Process Reward Model Step-by-Step Logic Evaluator
- Native Multimodal Architecture Unified Sensory Foundation Model
- Dataset Shift Hospital Generalization Drift
- Self-Attention Dynamic Contextual Weighting