wassname/ray
1
0
Fork 0
mirror of https://github.com/wassname/ray.git synced 2026-07-21 12:50:45 +08:00
Code Issues Packages Projects Releases Wiki Activity

[RLlib] Make sure, DQN torch actions are of type=long before torch.nn.functional.one_hot() op. (#11800)

Browse Source
This commit is contained in:
Sven Mika
2020-11-04 18:04:03 +01:00
committed by GitHub
parent 9073e6507c
commit d6c7c7c675
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
rllib/agents/dqn/dqn_torch_policy.py
+1 -1
View File
@@ -253,7 +253,7 @@ def build_q_losses(policy: Policy, model, _,
is_training=True)
# Q scores for actions which we know were selected in the given state.
one_hot_selection = F.one_hot(train_batch[SampleBatch.ACTIONS],
one_hot_selection = F.one_hot(train_batch[SampleBatch.ACTIONS].long(),
policy.action_space.n)
q_t_selected = torch.sum(
torch.where(q_t > FLOAT_MIN, q_t,