update steps_counter to int

examples/trainval.py

@@ -69,14 +69,15 @@ def render_image(radiance_field, rays, render_bkgd, render_step_size):
         )
         results.append(chunk_results)
     rgb, depth, acc, counter, compact_counter = [
-        torch.cat(r, dim=0) for r in zip(*results)
+        torch.cat(r, dim=0) if isinstance(r[0], torch.Tensor) else r
+        for r in zip(*results)
     ]
     return (
         rgb.view((*rays_shape[:-1], -1)),
         depth.view((*rays_shape[:-1], -1)),
         acc.view((*rays_shape[:-1], -1)),
-        counter.sum(),
-        compact_counter.sum(),
+        sum(counter),
+        sum(compact_counter),
     )
 
 
@@ -191,7 +192,7 @@ if __name__ == "__main__":
             )
             num_rays = len(pixels)
             num_rays = int(
-                num_rays * (TARGET_SAMPLE_BATCH_SIZE / float(compact_counter.item()))
+                num_rays * (TARGET_SAMPLE_BATCH_SIZE / float(compact_counter))
             )
             train_dataset.update_num_rays(num_rays)
             alive_ray_mask = acc.squeeze(-1) > 0
@@ -210,7 +211,7 @@ if __name__ == "__main__":
                     f"elapsed_time={elapsed_time:.2f}s (data={data_time:.2f}s) | {step=} | "
                     f"loss={loss:.5f} | "
                     f"alive_ray_mask={alive_ray_mask.long().sum():d} | "
-                    f"counter={counter.item():d} | compact_counter={compact_counter.item():d} | num_rays={len(pixels):d} "
+                    f"counter={counter:d} | compact_counter={compact_counter:d} | num_rays={len(pixels):d} "
                 )
 
             # if time.time() - tic > 300:

nerfacc/volumetric_rendering.py

@@ -23,6 +23,12 @@ def volumetric_rendering(
     """A *fast* version of differentiable volumetric rendering."""
     n_rays = rays_o.shape[0]
 
+    rays_o = rays_o.contiguous()
+    rays_d = rays_d.contiguous()
+    scene_aabb = scene_aabb.contiguous()
+    scene_occ_binary = scene_occ_binary.contiguous()
+    render_bkgd = render_bkgd.contiguous()
+
     # get packed samples from ray marching & occupancy check.
     with torch.no_grad():
         (
@@ -45,7 +51,7 @@ def volumetric_rendering(
         frustum_positions = (
             frustum_origins + frustum_dirs * (frustum_starts + frustum_ends) / 2.0
         )
-        steps_counter = packed_info[:, -1].sum(0, keepdim=True)
+        steps_counter = frustum_origins.shape[0]
 
     # compat the samples thru volumetric rendering
     with torch.no_grad():
@@ -64,7 +70,7 @@ def volumetric_rendering(
             frustum_positions,
             frustum_dirs,
         )
-        compact_steps_counter = compact_packed_info[:, -1].sum(0, keepdim=True)
+        compact_steps_counter = compact_frustum_positions.shape[0]
 
     # network
     compact_query_results = query_fn(compact_frustum_positions, compact_frustum_dirs)