why are there multiple settings for actor_rollout_ref.model.enable_gradient_checkpointing? Is this a deliberate design choice?
#4576
This workflow is awaiting approval from a maintainer in #4263
Triggered via pull request
February 27, 2026 02:59
khazic
Status
Action required
Total duration
–
Artifacts
–
This workflow is awaiting approval from a maintainer in #4263
e2e_ppo_trainer_megatron_vllm_2.yml
on: pull_request
setup
e2e_ppo_trainer_fsdp-qwen2_5vl-3b
e2e_ppo_trainer_fsdp_vllm
e2e_ppo_trainer_megatron-moe-expert-parallel
cleanup