Fix some issues with pdf

ml-agents/mlagents/trainers/distributions_torch.py

     def sample(self):
         return self.mean + torch.randn_like(self.mean) * self.std
 
-    def pdf(self, value):
+    def log_prob(self, value):
         var = self.std ** 2
         log_scale = self.std.log()
         return (
         )
 
-    def log_prob(self, value):
-        return torch.log(self.pdf(value))
+    def pdf(self, value):
+        log_prob = self.log_prob(value)
+        return torch.exp(log_prob)
 
     def entropy(self):
         return torch.log(2 * math.pi * math.e * self.std)

ml-agents/mlagents/trainers/models_torch.py

 import torch
 from torch import nn
 
-from mlagents.trainers.distributions_torch import GaussianDistribution, CategoricalDistInstance
+from mlagents.trainers.distributions_torch import (
+    GaussianDistribution,
+    MultiCategoricalDistribution,
+)
+from mlagents.trainers.exception import UnityTrainerException
 
 ActivationFunction = Callable[[torch.Tensor], torch.Tensor]
 EncoderFunction = Callable[
             hidden = encoder(vis_input)
             vis_embeds.append(hidden)
 
+        #embedding = vec_embeds[0]
         if len(vec_embeds) > 0:
             vec_embeds = torch.stack(vec_embeds, dim=-1).sum(dim=-1)
         if len(vis_embeds) > 0:
         super(ActorCritic, self).__init__()
         self.act_type = ActionType.from_str(act_type)
         self.act_size = act_size
+        self.version_number = torch.nn.Parameter(torch.Tensor([2.0]))
+        self.memory_size = torch.nn.Parameter(torch.Tensor([0]))
+        self.is_continuous_int = torch.nn.Parameter(torch.Tensor([1]))
+        self.act_size_vector = torch.nn.Parameter(torch.Tensor(act_size))
         self.separate_critic = separate_critic
         self.network_body = NetworkBody(
             vector_sizes,
             vec_inputs, vis_inputs, masks, memories, sequence_length
         )
         sampled_actions = self.sample_action(dists)
-        return sampled_actions, dists[0].pdf(sampled_actions)
+        return sampled_actions, dists[0].pdf(sampled_actions), self.version_number, self.memory_size, self.is_continuous_int, self.act_size_vector
 
 
 class Critic(nn.Module):