[rllib] Don't use get_gpu_ids() in ppo

This lets the num_gpus config work properly even when not using tune, since the gpu ids won't be set by ray in that case.
2026-08-04 13:14:14 +08:00 · 2018-08-01 16:25:11 -07:00
parent d9a36c4e39
commit a630e332f3
2 changed files with 5 additions and 4 deletions
@@ -91,8 +91,9 @@ class PPOAgent(Agent):
                    "sgd_batch_size": self.config["sgd_batchsize"],
                    "sgd_stepsize": self.config["sgd_stepsize"],
                    "num_sgd_iter": self.config["num_sgd_iter"],
+                    "num_gpus": self.config["num_gpus"],
                    "timesteps_per_batch": self.config["timesteps_per_batch"],
-                    "standardize_fields": ["advantages"]
+                    "standardize_fields": ["advantages"],
                })

    def _train(self):