Is the LLM part the same as Qwen7B-Chat?

[abussseee]

Is the LLM part the same as Qwen7B-Chat?

[simonJJJ]

The LLM part was initialized from Qwen7B (not Chat version). And we fine-tuned it in further training.

[tfka]

The LLM part was initialized from Qwen7B (not Chat version). And we fine-tuned it in further training.

In the VL model, both the base version and the chat version use lightweight tuning methods like LORA or full-parameter fine-tuning on LLM?

[simonJJJ]

@tfka, we fully fine-tuned LLM.