# actor_rollout_ref.ref: FSDP config same as actor. For models larger than 7B, it’s recommended to turn on offload for ref by default
strategy: ${actor_rollout_ref.actor.strategy}

# whether to enable torch.compile
# same as actor_rollout_ref.actor.use_torch_compile if it exists, otherwise 1
use_torch_compile: ${oc.select:actor_rollout_ref.actor.use_torch_compile,true}

# [Will be deprecated, use log_prob_micro_batch_size_per_gpu]
# The batch size for one forward pass in the computation of log_prob. Global batch size.
log_prob_micro_batch_size: null

# The batch size for one forward pass in the computation of log_prob. Local batch size per GPU.
log_prob_micro_batch_size_per_gpu: null

# enable dynamic batch size (sequence packing) for log_prob computation
# same as actor_rollout_ref.actor.use_dynamic_bsz if it exists, otherwise false
log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}

# the max token length per GPU
# same as actor_rollout_ref.actor.ppo_max_token_len_per_gpu if it exists, otherwise 16384
log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}