Architecture should be Qwen2ForCausalLM or Qwen2Model?

#19
by ultraxyz - opened

In config.json it says:

"architectures": [
    "Qwen2ForCausalLM"
  ],

But according to this post(https://github.com/vllm-project/vllm/issues/5181) the architecture of embedding model should not contain "CausalLM".

ultraxyz changed discussion title from Architecture should be Qwen2ForCausalLM or Qwen2? to Architecture should be Qwen2ForCausalLM or Qwen2Model?
Alibaba-NLP org

All is ok

Sign up or log in to comment