🚀Major Update: Introducing WizardLM-13B-V1.2 trained from Llama-2

Compared with WizardLM-13B-V1.1, there are the following updates:

• Context Window: 2048 -> 4096
• AlpacaEval Leaderboard: 86.32% -> 89.17%
• MT-Bench Leaderboard: 6.76 -> 7.06
• WizardLM Eval: 99.3% -> 101.4%