load whole dataset in train loop

Summary: Loads the whole dataset and moves it to the device and sends it to for sampling to enable full dataset heterogeneous raysampling.

Reviewed By: bottler

Differential Revision: D39263009

fbshipit-source-id: c527537dfc5f50116849656c9e171e868f6845b1

Author: Darijan Gudelj

projects/implicitron_trainer/experiment.py

@@ -222,6 +222,7 @@ def run(self) -> None:
             train_loader=train_loader,
             val_loader=val_loader,
             test_loader=test_loader,
+            # pyre-ignore[6]
             train_dataset=datasets.train,
             model=model,
             optimizer=optimizer,

projects/implicitron_trainer/impl/training_loop.py

@@ -22,7 +22,7 @@
 )
 from pytorch3d.implicitron.tools.stats import Stats
 from pytorch3d.renderer.cameras import CamerasBase
-from torch.utils.data import DataLoader
+from torch.utils.data import DataLoader, Dataset
 
 from .utils import seed_all_random_engines
 
@@ -44,6 +44,7 @@ def run(
         train_loader: DataLoader,
         val_loader: Optional[DataLoader],
         test_loader: Optional[DataLoader],
+        train_dataset: Dataset,
         model: ImplicitronModelBase,
         optimizer: torch.optim.Optimizer,
         scheduler: Any,
@@ -116,6 +117,7 @@ def run(
         train_loader: DataLoader,
         val_loader: Optional[DataLoader],
         test_loader: Optional[DataLoader],
+        train_dataset: Dataset,
         model: ImplicitronModelBase,
         optimizer: torch.optim.Optimizer,
         scheduler: Any,

pytorch3d/implicitron/models/generic_model.py

@@ -389,7 +389,8 @@ def safe_slice_targets(
         )
 
         # (1) Sample rendering rays with the ray sampler.
-        ray_bundle: ImplicitronRayBundle = self.raysampler(  # pyre-fixme[29]
+        # pyre-ignore[29]
+        ray_bundle: ImplicitronRayBundle = self.raysampler(
             target_cameras,
             evaluation_mode,
             mask=mask_crop[:n_targets]