Make multisequence evaluator use only cameras within a batch (fixes OOM in v2)

Summary: In a multisequence (fewview) scenario, it does not make sense to use all cameras for evaluating the difficulty as they come from different scenes. Using only this batch’s source (known) cameras instead.

Reviewed By: bottler

Differential Revision: D38491070

fbshipit-source-id: d6312d8fbb125b28a33db9f53d4215bcd1ca28a8

pytorch3d/implicitron/evaluation/evaluator.py

@@ -115,7 +115,9 @@ class ImplicitronEvaluator(EvaluatorBase):
                         implicitron_render,
                         bg_color="black",
                         lpips_model=lpips_model,
-                        source_cameras=all_train_cameras,
+                        source_cameras=(  # None will make it use batch’s known cameras
+                            None if self.is_multisequence else all_train_cameras
+                        ),
                     )
                 )