hongshaorou/dreamerv3-torch
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix bug when resetting envs at different time

Browse Source
This commit is contained in:
NM512 2023-07-24 22:26:21 +09:00
parent 22e9f8de7b
commit 43e1b2ab88
1 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
tools.py
View File

@ -150,15 +150,15 @@ def simulate(
indices = [index for index, d in enumerate(done) if d] indices = [index for index, d in enumerate(done) if d]
results = [envs[i].reset() for i in indices] results = [envs[i].reset() for i in indices]
results = [r() for r in results] results = [r() for r in results]
for i in indices: for index, result in zip(indices, results):
t = results[i].copy() t = result.copy()
t = {k: convert(v) for k, v in t.items()} t = {k: convert(v) for k, v in t.items()}
# action will be added to transition in add_to_cache # action will be added to transition in add_to_cache
t["reward"] = 0.0 t["reward"] = 0.0
t["discount"] = 1.0 t["discount"] = 1.0
# initial state should be added to cache # initial state should be added to cache
add_to_cache(cache, envs[i].id, t) add_to_cache(cache, envs[index].id, t)
for index, result in zip(indices, results): # replace obs with done by initial state
obs[index] = result obs[index] = result
# step agents # step agents
obs = {k: np.stack([o[k] for o in obs]) for k in obs[0]} obs = {k: np.stack([o[k] for o in obs]) for k in obs[0]}