20B parameter GPT-3 model trained on a single Cerebras chipcerebras.net2 points·keveman··0 commentsOpen articleSaveView on HN