Patent Leaderboard
USPTO Patent Rankings Data through Dec 31, 2025
Home › Highly performant pipeline parallel deep neural network training

Highly performant pipeline parallel deep neural network training

US Patent 12056604 · Granted Aug 6, 2024

Estimated economic value: $310,621,000 ⓘ

Assignee

Inventors

View full patent text on Google Patents →

More Patents by These Inventors

Training neural networks based on dual pipeline architectures2025
Neural network activation compression with narrow block floating-point2025
Deep learning scheduler toolkit2025
Neural network activation compression with non-uniform mantissas2025
Adjusting activation compression for neural network training2024
Neural network activation compression with non-uniform mantissas2024
Neural network activation compression with outlier block floating-point2024
Mitigating communication bottlenecks during parameter exchange in data-parallel 2024
Mitigating communication bottlenecks during parameter exchange in data-parallel 2024
Efficient data encoding for deep neural network training2023
Neural network activation compression with non-uniform mantissas2023
Data amelioration and reformation system2022