Posts in: Blog

Announcing local gradient aggregation, a new performance optimization for open-source distributed training with Horovod!

Slurm is a great HPC job scheduler but lacks key deep learning capabilities, such as hyperparameter tuning, distributed training, and experiment management.

How do Determined and Kubeflow compare? Determined is better for deep learning training, but Determined and Kubeflow can also be effectively used together.

Grad school would have been easier if we had access to Determined’s deep learning training platform.

Takeaways from our experience building state-of-the-art hyperparameter tuning in Determined AI’s integrated deep learning training platform.

Dramatically reduce the time to train GANs with Determined AI.

Learn how to do production-grade MLOps with scalable, automated machine learning training and deployment using Determined, Kubeflow Pipelines, and Seldon Core.

Training deep learning models for NLP tasks typically requires many hours or days to complete on a single GPU. In this post, we leverage Determined’s distributed training capability to reduce BERT for SQuAD model training time from hours to minutes, without sacrificing model accuracy.

How to build an end-to-end deep learning pipeline, including data preprocessing with Spark, versioned data storage with Delta Lake, distributed training with Determined, and batch inference with Spark.

A better approach to loading data for deep learning models.