brozek/rltorch
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

d.sample returns a tensor, so we stack them to not lose the device

Browse source
This commit is contained in:
Brandon Rozek 2019-02-28 14:30:49 -05:00
parent 714443192d
commit 9740c40527
1 changed files with 1 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
rltorch/agents/QEPAgent.py
View file

@ -23,7 +23,7 @@ class QEPAgent:
def fitness(self, policy_net, value_net, state_batch):
action_probabilities = policy_net(state_batch)
distributions = list(map(Categorical, action_probabilities))
actions = torch.tensor([d.sample() for d in distributions])
actions = torch.stack([d.sample() for d in distributions])
with torch.no_grad():
state_values = value_net(state_batch)