2 papers
cs.CV2022
L3: Accelerator-Friendly Lossless Image Format for High-Resolution, High-Throughput DNN Training
Jonghyun Bae, Woohyeon Baek, Tae Jun Ham +1
The training process of deep neural networks (DNNs) is usually pipelined with stages for data preparation on CPUs followed by gradient computation on accelerators like GPUs. In an…
cond-mat.mtrl-sci2019
Pele's hairs and exotic multiply twinned graphite closed-shell microcrystals in meteoritic dust of Chelyabinsk superbolide
S. Taskaev, K. Skokov, V. Khovaylo +10
When a space body enters Earth's atmosphere, its surface is exposed to high pressure and temperatures. The airflow tears off small droplets from the meteoroid forming a cloud of me…