Quantized Training with Deep Networks
How to cut neural net training time in half with minimal effort…
Saga of the Lottery Ticket Hypothesis
How winning tickets were discovered, debunked, and re-discovered.
Using CLIP to Classify Images without any Labels
How to achieve 76.2% test accuracy on ImageNet without observing a single label
Understanding the Open Pre-Trained Transformers (OPT) Library
The release of the OPT library is a true step towards improved transparency and general understanding in large language modeling and provides significant benefit to the research community as a whole.
Training a Random Forest to Identify Malignant Breast Cancer Tumors
Using machine learning to more accurately diagnose serious medical conditions.