AI Technology & Industry Review
Colossal-AI (https://github.com/hpcaitech/ColossalAI), the widely-used open-source library for training, inference and fine-tuning of large deep learning models, has released a new automatic parallelism feature and functionality that reduces hardware costs by up to 46 times for AI-Generate Content (AIGC) solutions.
A research team from UC Berkeley, Amazon Web Services, Google, Shanghai Jiao Tong University and Duke University proposes Alpa, a compiler system for distributed deep learning on GPU clusters that automatically generates parallelization plans that match or outperform hand-tuned model-parallel training systems even on the models they were designed for.
A research team from Microsoft and NVIDIA leverages the NVIDIA Megatron-LM and Microsoft’s DeepSpeed to create an efficient and scalable 3D parallel system that combines data, pipeline, and tensor-slicing based parallelism, achieving superior zero-, one-, and few-shot learning accuracies and new state-of-the-art results on NLP benchmarks.
A Google research team presents ALX, an open-source library that leverages Tensor Processing Units (TPUs) to enable efficient distributed matrix factorization using Alternating Least Squares.
A research team from ETH Zurich and NVIDIA proposes a training framework that enables fast policy generation for real-world robotic tasks — training time can be reduced by multiple orders of magnitude using massive parallelism on a single workstation GPU.
A research team from Google proposes GSPMD, an automatic parallelism system for ML computation graphs that uses simple tensor sharding annotations to achieve different parallelism paradigms in a unified way, including data parallelism, within-layer model parallelism, spatial partitioning, weight-update sharding, optimizer-state sharding and pipeline parallelism.
A research team from NVIDIA, Stanford University and Microsoft Research propose a novel pipeline parallelism approach that improves throughput by more than 10 percent with a comparable memory footprint, showing such strategies can achieve high aggregate throughput while training models with up to a trillion parameters.
“We want to be able to train a machine learning model, like Google Translate, in a few hours,” said Prof. Hwu. Best known for enabling the parallelism capability of the GPU to solve scientific problems like AI.