I am currently working with the Qwen2-VL-7B-Instruct model and wanted to know if there is an option to set the temperature parameter for generation, since I would like to experiment with it if available.
I searched both here in GitHub and in HuggingFace but couldn't find anything related to it.
I tested different values of the temperature in the vLLM implementation, but it seems to not influence the output.
Could you please clarify if this feature exists for Qwen2-VL, and if so, provide guidance on how to configure it?
vardaan123
commented
2 days ago
Hello,
I am currently working with the Qwen2-VL-7B-Instruct model and wanted to know if there is an option to set the temperature parameter for generation, since I would like to experiment with it if available. I searched both here in GitHub and in HuggingFace but couldn't find anything related to it.
I tested different values of the temperature in the vLLM implementation, but it seems to not influence the output.
Could you please clarify if this feature exists for Qwen2-VL, and if so, provide guidance on how to configure it?
Thank you!