Test commit

ml-agents/mlagents/trainers/torch/distributions.py

 
     def sample(self):
         sample = self.mean + torch.randn_like(self.mean) * self.std
-        return torch.clamp(sample, -3, 3) / 3
+        return torch.clamp(sample, -3, 3)
 
     def log_prob(self, value):
         unscaled_val = value * 3  # Inverse of the clipping