wassname/DeepRL
1
0
Fork 0
mirror of https://github.com/wassname/DeepRL.git synced 2026-09-10 11:40:58 +08:00
Code Issues Packages Projects Releases Wiki Activity

add eps to avoid NaN

Browse Source 
I tracked some NaN's I was getting down to here. It happens when std is a small number, then var is even smaller, and log_density=inf. There are some problems where the agent will learn to use small standard deviations because jittering movements have a high cost, this makes those more stable.
This commit is contained in:
Mike Clark
2017-11-04 16:24:54 +08:00
committed by GitHub
parent 9ef309696f
commit deab7043fc
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
network/continuous_action_network.py
+1 -1
View File
@@ -154,7 +154,7 @@ class GaussianActorNet(nn.Module, BasicNet):
def log_density(self, x, mean, log_std, std):
var = std.pow(2)
log_density = -(x - mean).pow(2) / (2 * var) - 0.5 * torch.log(2 * Variable(torch.FloatTensor([np.pi])).expand_as(x)) - log_std
log_density = -(x - mean).pow(2) / (2 * var + 1e-5) - 0.5 * torch.log(2 * Variable(torch.FloatTensor([np.pi])).expand_as(x)) - log_std
return log_density.sum(1)
def entropy(self, std):