New GPU Dependency Engine in Apache MXNet
In this short presentation, Serge Panev from NVIDIA introduces their latest work on improving MXNet's dependency engine / scheduler that reduces the synchronization and improves the overall performance.
Видео New GPU Dependency Engine in Apache MXNet канала Apache MXNet
Видео New GPU Dependency Engine in Apache MXNet канала Apache MXNet
Показать
Комментарии отсутствуют
Информация о видео
Другие видео канала
Matched-filtering Techniques and Deep Neural Networks for Gravitational Wave Astronomy
Automatic Mixed Precision in Apache MXNet
The Apache Way
Standardizing on a single N-dimensional array API for Python
Large Tensor Support
Using Apache MXNet in Production Deep Learning Streaming Pipelines
Neural Time Series Models with GluonTS
Towards the ultimate framework efficiency
GluonCV with Apache MXNet
Apache MXNet 2.0: Bridging Deep Learning and Machine Learning
Optimizing Inference for Neural Machine Translation using Sockeye 2
AutoGluon: Powerful and Easy-to-use AutoML for Text, Image, and Tabular Data
Apache MXNet: A Call to ARMs
Extending Apache MXNet for new features and performance
Serving 1 Million BERT inference requests for 20 cents
Optimizing inference on CPU in Apache MXNet 2.0
The History and the Future of Run-time Compilation in MXNet
Looking for change? Roll the Dice and demand Attention
CUDA Graphs support in Apache MXNet 1.8
Apache TVM and Machine Learing Compilation