[fully_async] fix DAPO postprocess_generator_output to accept metrics kwargs#1812

Merged

SumanthRH merged 2 commits into

NovaSky-AI:mainfrom

erictang000:fix/fully-async-dapo-postprocess-kwargs

Jun 23, 2026

Commits on Jun 19, 2026

[fully_async] fix DAPO postprocess_generator_output to accept metrics kwargs

erictang000
and
claude
committed
[fully_async] avoid double penalty from shared inner reward lists

erictang000
and
claude
committed