'StableBaselines-3 DDPG + HER Multiprocessing
I was reading documentation about HER and also about Multiprocessing in stable-baselines3 website However when i try to train it throws a error! Is there any example anywhere with multiprocessing with replay_buffer??
num_cpu = 4
env = make_vec_env(env_name, n_envs=num_cpu, seed=0, vec_env_cls=SubprocVecEnv)
model = DDPG(
'MultiInputPolicy',
env,
verbose = 1,
tensorboard_log = log_path,
replay_buffer_class = HerReplayBuffer,
# Parameters for HER
replay_buffer_kwargs=dict(
n_sampled_goal = 4,
goal_selection_strategy = GoalSelectionStrategy.FUTURE,
online_sampling = True,
),
)
model.learn(total_timesteps = 1000000)
ValueError: could not broadcast input array from shape (6,4) into shape (4,)
Solution 1:[1]
Stable baselines3 HER does not support multiprocessing.
Sources
This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.
Source: Stack Overflow
| Solution | Source |
|---|---|
| Solution 1 | Dharman |
