gaspersic/baselines
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

rename entcoeff to ent_coef in trpo_mpi for compatibility with other algos (#581)

Browse Source
This commit is contained in:
pzhokhov
2018-09-18 14:13:05 -07:00
committed by GitHub
parent 68a29d0ab3
commit 59662fff78
1 changed files with 3 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
baselines/trpo_mpi/trpo_mpi.py
View File

@@ -92,7 +92,7 @@ def learn(*,
gamma=0.99,
lam=1.0, # advantage estimation
seed=None,
entcoeff=0.0,
ent_coef=0.0,
cg_damping=1e-2,
vf_stepsize=3e-4,
vf_iters =3,
@@ -117,7 +117,7 @@ def learn(*,
max_kl max KL divergence between old policy and new policy ( KL(pi_old || pi) )
entcoeff coefficient of policy entropy term in the optimization objective
ent_coef coefficient of policy entropy term in the optimization objective
cg_iters number of iterations of conjugate gradient algorithm
@@ -182,7 +182,7 @@ def learn(*,
ent = pi.pd.entropy()
meankl = tf.reduce_mean(kloldnew)
meanent = tf.reduce_mean(ent)
entbonus = entcoeff * meanent
entbonus = ent_coef * meanent
vferr = tf.reduce_mean(tf.square(pi.vf - ret))