Add more model and chat template

Fix typo

Update starcoder

set tp for bias

Fix rope for qwen and add internlm

Update readme

Add qwen2 and new quants

Add q8 kernel

use quantized embedding

Expert parallel for MOE

Fix tp for mixtral

Add Mixtral

Update tensor parallel

Initial support for new 2-bit quants

Fix rope base