research
∙
06/18/2023
DropCompute: simple and more robust distributed synchronous training via compute variance reduction
Background: Distributed training is essential for large scale training o...
share
research
∙
06/15/2020
Neural gradients are lognormally distributed: understanding sparse and quantized training
Neural gradient compression remains a main bottleneck in improving train...
research
∙
02/18/2020
Robust Quantization: One Model to Rule Them All
Neural network quantization methods often involve simulating the quantiz...
research
∙
09/17/2019