wassname/ray
1
0
Fork 0
mirror of https://github.com/wassname/ray.git synced 2026-09-09 11:32:43 +08:00
Code Issues Packages Projects Releases Wiki Activity

[rllib] Fix truncate episodes mode in central critic example (#8073)

Browse Source
This commit is contained in:
Eric Liang
2020-04-20 12:58:01 -07:00
committed by GitHub
parent 3812bfedda
commit 17e3c545d9
1 changed files with 8 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
rllib/examples/centralized_critic.py
+8 -4
View File
@@ -96,8 +96,6 @@ def centralized_critic_postprocessing(policy,
other_agent_batches=None, other_agent_batches=None,
episode=None): episode=None):
if policy.loss_initialized(): if policy.loss_initialized():
assert sample_batch["dones"][-1], \
"Not implemented for train_batch_mode=truncate_episodes"
assert other_agent_batches is not None assert other_agent_batches is not None
[(_, opponent_batch)] = list(other_agent_batches.values()) [(_, opponent_batch)] = list(other_agent_batches.values())
@@ -116,11 +114,17 @@ def centralized_critic_postprocessing(policy,
sample_batch[OPPONENT_ACTION] = np.zeros_like( sample_batch[OPPONENT_ACTION] = np.zeros_like(
sample_batch[SampleBatch.ACTIONS]) sample_batch[SampleBatch.ACTIONS])
sample_batch[SampleBatch.VF_PREDS] = np.zeros_like( sample_batch[SampleBatch.VF_PREDS] = np.zeros_like(
sample_batch[SampleBatch.ACTIONS], dtype=np.float32) sample_batch[SampleBatch.REWARDS], dtype=np.float32)
completed = sample_batch["dones"][-1]
if completed:
last_r = 0.0
else:
last_r = sample_batch[SampleBatch.VF_PREDS][-1]
train_batch = compute_advantages( train_batch = compute_advantages(
sample_batch, sample_batch,
0.0, last_r,
policy.config["gamma"], policy.config["gamma"],
policy.config["lambda"], policy.config["lambda"],
use_gae=policy.config["use_gae"]) use_gae=policy.config["use_gae"])