nlpxucan/WizardLM
WizardLM is a collection of large language models trained using Evol-Instruct to follow complex instructions.

Velocity · 7d
+8.3
★ / day
Trend
→steady
star history
WizardLM provides a suite of LLMs including WizardLM (general instruction following), WizardCoder (code generation), and WizardMath (mathematical reasoning). These models are built using the Evol-Instruct method which automatically generates and filters training data to improve instruction following capabilities. The project includes training code, data generation utilities, and model weights for several model sizes.