How to set minimum number of output tokens ?

[thangld201]

This param is equivalent to min_new_tokens in huggingface .generate(). Is this already supported in vLLM ? @WoosukKwon I only see the max_tokens option in SamplingParam ..

[simon-mo]

You can checkout this PR here #1945 but it has been stalled. If you can help getting this into a good shape that would be even better!