VedantDave/Curiosity
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Update models to give distribution as well in the output

Browse Source
This commit is contained in:
Vedant Dave 2023-03-27 19:22:17 +02:00
parent a351134f08
commit 38cc645253
1 changed files with 16 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

29
DPI/models.py
View File

@ -34,12 +34,22 @@ class ObservationEncoder(nn.Module):
std = torch.clamp(std, min=0.0, max=1e5)
# Normal Distribution
dist = self.get_dist(mean, std)
# Sampling via reparameterization Trick
x = self.reparameterize(mean, std)
return x
encoded_output = {"sample": x, "distribution": dist}
return encoded_output
def reparameterize(self, mu, std):
eps = torch.randn_like(std)
return mu + eps * std
def get_dist(self, mean, std):
distribution = torch.distributions.Normal(mean, std)
distribution = torch.distributions.independent.Independent(distribution, 1)
return distribution
class ObservationDecoder(nn.Module):
@ -114,8 +124,12 @@ class TransitionModel(nn.Module):
state_prior_mean, state_prior_std = torch.chunk(state_prior, 2, dim=-1)
state_prior_std = F.softplus(state_prior_std)
# Normal Distribution
state_prior_dist = self.get_dist(state_prior_mean, state_prior_std)
# Sampling via reparameterization Trick
sample_state_prior = self.reparemeterize(state_prior_mean, state_prior_std)
prior = {"mean": state_prior_mean, "std": state_prior_std, "sample": sample_state_prior, "history": history}
prior = {"mean": state_prior_mean, "std": state_prior_std, "sample": sample_state_prior, "history": history, "distribution": state_prior_dist}
return prior
def reparemeterize(self, mean, std):
@ -154,15 +168,4 @@ class CLUBSample(nn.Module): # Sampled version of the CLUB estimator
def forward(self, x_samples, y_samples):
mu, logvar = self.get_mu_logvar(x_samples)
sample_size = x_samples.shape[0]
#random_index = torch.randint(sample_size, (sample_size,)).long()
random_index = torch.randperm(sample_size).long()
positive = - (mu - y_samples)**2 / logvar.exp()
negative = - (mu - y_samples[random_index])**2 / logvar.exp()
upper_bound = (positive.sum(dim = -1) - negative.sum(dim = -1)).mean()
return upper_bound/2.
def learning_loss(self, x_samples, y_samples):
return - self.loglikeli(x_samples, y_samples)