WizardLM-2 adopts the prompt format from Vicuna and supports multi-switch dialogue. The prompt should be as pursuing:Individuals top quality controls included both of those heuristic and NSFW filters, in addition to information deduplication, and textual content classifiers accustomed to predict the standard of the information just before training.