Merge pull request #70 from ajboyd2/patch-2

Correct scaling of output intensities in FullyNN.

Author: iLampard

easy_tpp/model/torch_model/torch_fullynn.py

@@ -64,13 +64,13 @@ def forward(self, hidden_states, time_delta_seqs):
             derivative_integral_lambdas = []
             for i in range(integral_lambda.shape[-1]):  # iterate over marks
                 derivative_integral_lambdas.append(grad(
-                    integral_lambda[..., i].mean(),
+                    integral_lambda[..., i].sum(),
                     time_delta_seqs,
                     create_graph=True, retain_graph=True)[0])
             derivative_integral_lambda = torch.stack(derivative_integral_lambdas, dim=-1)  # TODO: Check that it is okay to iterate over marks like this
         else:
-            derivative_integral_lambda = grad(
-                integral_lambda.sum(dim=-1).mean(),
+            derivative_integral_lambda = grad( 
+                integral_lambda.sum(),
                 time_delta_seqs,
                 create_graph=True, retain_graph=True)[0]
             derivative_integral_lambda = derivative_integral_lambda.unsqueeze(-1).expand(*derivative_integral_lambda.shape, self.num_event_types) / self.num_event_types