verl/workers/critic/dp_critic.py

• DataParallelPPOCritic · class · L39-L204 — class DataParallelPPOCritic(BasePPOCritic)
• init · method · L41-L51 — def init(self, config, critic_module: nn.Module, critic_optimizer: optim.Optimizer)
• _forward_micro_batch · method · L53-L101 — def _forward_micro_batch(self, micro_batch)
• _optimizer_step · method · L103-L111 — def _optimizer_step(self)
• compute_values · method · L113-L144 — def compute_values(self, data: DataProto) -> torch.Tensor
• update_critic · method · L146-L204 — def update_critic(self, data: DataProto): # make sure we are in training mode