fixed some more remnant issues of the last commit

Signed-off-by: Martin <[email protected]>

dgs/models/combine/dynamic.py

@@ -65,11 +65,11 @@ def forward(
         r"""The forward call of this module combines an arbitrary number of similarity matrices
         using an importance weight :math:`\alpha`.
 
-        :math:`\alpha_i` describes how important :math:`s_i` is.
+        :math:`\alpha_i` describes how important the similarity :math:`s_i` is.
         The sum of all :math:`\alpha_i` should be 1 by definition given the last layer is a softmax layer.
-        :math:`\alpha` is computed using this class' neural network and the given ``alpha_input`` tensor.
+        :math:`\alpha` is computed using the respective :class:`BaseAlphaModule` and the given :class:`State`.
 
-        All tensors should be on the same device and all :math:`s_i` should have the same shape.
+        All tensors should be on the same device and should have the same shape.
 
         Args:
             tensors: A tuple of tensors describing similarities between the detections and tracks.

dgs/models/engine/dgs_engine.py

@@ -237,19 +237,18 @@ def _track_step(self, detections: State, frame_idx: int, name: str, timers: Diff
 
         if len(track_states) == 0 and N > 0:
             # No Tracks yet - every detection will be a new track!
-            # Make sure to compute the embeddings for every detection, to ensure correct behavior of collate later on
+            # Make sure to compute the embeddings for every detection,
+            # this is done to ensure correct behavior of the collate function later on.
             time_sim_start = time.time()
-            _ = self.model.forward(ds=detections, target=detections, alpha_inputs=self.get_data(detections))
+            _ = self.model.forward(ds=detections, target=detections, s=detections)
             timers.add(name="similarity", prev_time=time_sim_start)
-            # There are no tracks yet, therefore every detection is a new state!
+            # There are no tracks yet, therefore, every detection is a new state!
             time_match_start = time.time()
             new_states += detections.split()
             timers.add(name="match", prev_time=time_match_start)
         elif N > 0:
             time_sim_start = time.time()
-            similarity = self.model.forward(
-                ds=detections, target=collate_states(track_states), alpha_inputs=self.get_data(detections)
-            )
+            similarity = self.model.forward(ds=detections, target=collate_states(track_states), s=detections)
             timers.add(name="similarity", prev_time=time_sim_start)
 
             # Solve Linear sum Assignment Problem (LAP/LSA).

tests/models/combine/test__combine.py

@@ -296,7 +296,7 @@ def test_state_not_set(self):
             self.model.forward(*self.dummy_t)
         self.assertTrue("The state should be given" in str(e.exception), msg=e.exception)
 
-    def test_runtime_error_alpha_input_device(self):
+    def test_runtime_error_state_device(self):
         if t.cuda.is_available():
             # tensor based
             with self.assertRaises(RuntimeError) as e:
@@ -305,7 +305,7 @@ def test_runtime_error_alpha_input_device(self):
                 )  # Different devices
             self.assertIn("s should be on the same device as tensors", str(e.exception))
 
-    def test_value_error_on_nof_alpha_inputs(self):
+    def test_value_error_on_batch_mismatch(self):
         # Mismatch in number of alpha inputs against number of alpha models
         with self.assertRaises(ValueError) as e:
             self.model.forward(*self.dummy_t, s=self.dummy_state_batched)