Support RecMetrics new options in Pyper config (#2987)

Summary:

Fix the issue in tower QPS metric update issue in P1806989639 and discussed in [link](https://fb.workplace.com/groups/527654686243695/posts/989373686738457/?comment_id=990338486641977&reply_comment_id=990584749950684), and expose the new TorchMetric fusion [feature](https://fb.workplace.com/groups/527654686243695/permalink/989373686738457) to Pyper users （default is disabled）

Differential Revision: D75120928

Author: HugeEngine

torchrec/metrics/tower_qps.py

@@ -227,9 +227,14 @@ def update(
                 RecComputeMode.FUSED_TASKS_AND_STATES_COMPUTATION,
             ]:
                 if not isinstance(labels, torch.Tensor):
-                    raise RecMetricException(
-                        "Fused computation only support where 'labels' is a tensor"
-                    )
+                    try:
+                        labels = torch.stack(
+                            [labels[task.name] for task in self._tasks]
+                        )
+                    except Exception as e:
+                        raise RecMetricException(
+                            f"Failed to convert labels to tensor for fused computation: {e}"
+                        )
                 labels = labels.view(-1, self._batch_size)
                 if self._should_validate_update:
                     # Set the default value to be all True. When weights is None, it's considered
@@ -241,9 +246,14 @@ def update(
                     )
                     if weights is not None:
                         if not isinstance(weights, torch.Tensor):
-                            raise RecMetricException(
-                                "Fused computation only support where 'weights' is a tensor"
-                            )
+                            try:
+                                weights = torch.stack(
+                                    [weights[task.name] for task in self._tasks]
+                                )
+                            except Exception as e:
+                                raise RecMetricException(
+                                    f"Failed to convert weights to tensor for fused computation: {e}"
+                                )
                         has_valid_weights = torch.gt(
                             torch.count_nonzero(
                                 weights.view(-1, self._batch_size), dim=-1