Cybertron AI

Cybertron AI's repositories

Make huge neural nets fit in memory

Language:PythonMIT2635 81 42

Code to reproduce "imagenet in 18 minutes" DAWN-benchmark entry

Language:PythonUnlicense845 51 24

Implementation of https://arxiv.org/abs/1904.00962

Language:PythonMIT362 7 10

Language:PythonUnlicense149 7 13

Training Transformer-XL on 128 GPUs

Language:PythonApache-2.0138 4 21

PyTorch-SSO: Scalable Second-Order methods in PyTorch

Language:PythonMIT137 5 5

Train ImageNet in 18 minutes on AWS

Language:PythonUnlicense123 5 7

Lightweight interface to AWS

Language:PythonMIT47 4 88

Language:PythonMIT32 4 7

Language:Python15 4 2

Ongoing research training transformer language models at scale, including: BERT

Language:PythonNOASSERTION15 30

Implementation of fluctuation dissipation relations for automatic learning rate annealing.

Language:PythonMIT8 6 1

Tools to benchmark AWS network performance

Language:Jupyter Notebook5 50

Example code for "PyTorch on AWS made easy"

Language:Python5 10