bacow/easy-rl
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

update rainbowdqn

Browse Source
This commit is contained in:
johnjim0816
2022-05-31 01:20:58 +08:00
parent cfc0f6492e
commit c7c94468c9
149 changed files with 1866 additions and 1549 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
codes/TD3/task0_train.py
View File

@@ -72,7 +72,7 @@ def train(cfg,env,agent):
else:
action = (
agent.choose_action(np.array(state))
+ np.random.normal(0, max_action * cfg.expl_noise, size=action_dim)
+ np.random.normal(0, max_action * cfg.expl_noise, size=n_actions)
).clip(-max_action, max_action)
# Perform action
next_state, reward, done, _ = env.step(action)
@@ -121,11 +121,11 @@ def train(cfg,env,agent):
# else:
# action = (
# agent.choose_action(np.array(state))
# + np.random.normal(0, max_action * cfg.expl_noise, size=action_dim)
# + np.random.normal(0, max_action * cfg.expl_noise, size=n_actions)
# ).clip(-max_action, max_action)
# # action = (
# # agent.choose_action(np.array(state))
# # + np.random.normal(0, max_action * cfg.expl_noise, size=action_dim)
# # + np.random.normal(0, max_action * cfg.expl_noise, size=n_actions)
# # ).clip(-max_action, max_action)
# # Perform action
# next_state, reward, done, _ = env.step(action)
@@ -157,10 +157,10 @@ if __name__ == "__main__":
env.seed(cfg.seed) # Set seeds
torch.manual_seed(cfg.seed)
np.random.seed(cfg.seed)
state_dim = env.observation_space.shape[0]
action_dim = env.action_space.shape[0]
n_states = env.observation_space.shape[0]
n_actions = env.action_space.shape[0]
max_action = float(env.action_space.high[0])
agent = TD3(state_dim,action_dim,max_action,cfg)
agent = TD3(n_states,n_actions,max_action,cfg)
rewards,ma_rewards = train(cfg,env,agent)
make_dir(cfg.result_path,cfg.model_path)
agent.save(path=cfg.model_path)