Support for moving the renderer to a new device

nikhilaravi · facebook-github-bot · commit 956d3a010cf8 · 2020-09-23T17:13:07.000-07:00
Summary:
Support for moving all the tensors of the renderer to another device by calling `renderer.to(new_device)`

Currently the `MeshRenderer`, `MeshRasterizer` and `SoftPhongShader` (and other shaders) are all of type `nn.Module` which already supports easily moving tensors of submodules (defined as class attributes) to a different device. However the class attributes of the rasterizer and shader (e.g. cameras, lights, materials), are of type `TensorProperties`, not nn.Module so we need to explicity create a `to` method to move these tensors to device. Note that the `TensorProperties` class already has a `to` method so we only need to call `cameras.to(device)` and don't need to worry about the internal tensors.

The other option is of course making these other classes (cameras, lights etc) also of type nn.Module.

Reviewed By: gkioxari

Differential Revision: D23885107

fbshipit-source-id: d71565c442181f739de4d797076ed5d00fb67f8e
diff --git a/pytorch3d/renderer/blending.py b/pytorch3d/renderer/blending.py
@@ -46,7 +46,7 @@ def hard_rgb_blend(colors, fragments, blend_params) -> torch.Tensor:
     is_background = fragments.pix_to_face[..., 0] < 0  # (N, H, W)
 
     if torch.is_tensor(blend_params.background_color):
-        background_color = blend_params.background_color
+        background_color = blend_params.background_color.to(device)
     else:
         background_color = colors.new_tensor(blend_params.background_color)  # (3)
 
@@ -163,6 +163,8 @@ def softmax_rgb_blend(
     background = blend_params.background_color
     if not torch.is_tensor(background):
         background = torch.tensor(background, dtype=torch.float32, device=device)
+    else:
+        background = background.to(device)
 
     # Weight for background color
     eps = 1e-10
diff --git a/pytorch3d/renderer/mesh/rasterizer.py b/pytorch3d/renderer/mesh/rasterizer.py
@@ -76,6 +76,10 @@ def __init__(self, cameras=None, raster_settings=None):
         self.cameras = cameras
         self.raster_settings = raster_settings
 
+    def to(self, device):
+        # Manually move to device cameras as it is not a subclass of nn.Module
+        self.cameras = self.cameras.to(device)
+
     def transform(self, meshes_world, **kwargs) -> torch.Tensor:
         """
         Args:
diff --git a/pytorch3d/renderer/mesh/renderer.py b/pytorch3d/renderer/mesh/renderer.py
@@ -33,6 +33,11 @@ def __init__(self, rasterizer, shader):
         self.rasterizer = rasterizer
         self.shader = shader
 
+    def to(self, device):
+        # Rasterizer and shader have submodules which are not of type nn.Module
+        self.rasterizer.to(device)
+        self.shader.to(device)
+
     def forward(self, meshes_world, **kwargs) -> torch.Tensor:
         """
         Render a batch of images from a batch of meshes by rasterizing and then
@@ -44,6 +49,7 @@ def forward(self, meshes_world, **kwargs) -> torch.Tensor:
         face f, clipping is required before interpolating the texture uv
         coordinates and z buffer so that the colors and depths are limited to
         the range for the corresponding face.
+        For this set rasterizer.raster_settings.clip_barycentric_coords=True
         """
         fragments = self.rasterizer(meshes_world, **kwargs)
         images = self.shader(fragments, meshes_world, **kwargs)
diff --git a/pytorch3d/renderer/mesh/shader.py b/pytorch3d/renderer/mesh/shader.py
@@ -50,6 +50,12 @@ def __init__(
         self.cameras = cameras
         self.blend_params = blend_params if blend_params is not None else BlendParams()
 
+    def to(self, device):
+        # Manually move to device modules which are not subclasses of nn.Module
+        self.cameras = self.cameras.to(device)
+        self.materials = self.materials.to(device)
+        self.lights = self.lights.to(device)
+
     def forward(self, fragments, meshes, **kwargs) -> torch.Tensor:
         cameras = kwargs.get("cameras", self.cameras)
         if cameras is None:
@@ -98,6 +104,12 @@ def __init__(
         self.cameras = cameras
         self.blend_params = blend_params if blend_params is not None else BlendParams()
 
+    def to(self, device):
+        # Manually move to device modules which are not subclasses of nn.Module
+        self.cameras = self.cameras.to(device)
+        self.materials = self.materials.to(device)
+        self.lights = self.lights.to(device)
+
     def forward(self, fragments, meshes, **kwargs) -> torch.Tensor:
         cameras = kwargs.get("cameras", self.cameras)
         if cameras is None:
@@ -151,6 +163,12 @@ def __init__(
         self.cameras = cameras
         self.blend_params = blend_params if blend_params is not None else BlendParams()
 
+    def to(self, device):
+        # Manually move to device modules which are not subclasses of nn.Module
+        self.cameras = self.cameras.to(device)
+        self.materials = self.materials.to(device)
+        self.lights = self.lights.to(device)
+
     def forward(self, fragments, meshes, **kwargs) -> torch.Tensor:
         cameras = kwargs.get("cameras", self.cameras)
         if cameras is None:
@@ -203,6 +221,12 @@ def __init__(
         self.cameras = cameras
         self.blend_params = blend_params if blend_params is not None else BlendParams()
 
+    def to(self, device):
+        # Manually move to device modules which are not subclasses of nn.Module
+        self.cameras = self.cameras.to(device)
+        self.materials = self.materials.to(device)
+        self.lights = self.lights.to(device)
+
     def forward(self, fragments, meshes, **kwargs) -> torch.Tensor:
         cameras = kwargs.get("cameras", self.cameras)
         if cameras is None:
@@ -272,6 +296,12 @@ def __init__(
         self.cameras = cameras
         self.blend_params = blend_params if blend_params is not None else BlendParams()
 
+    def to(self, device):
+        # Manually move to device modules which are not subclasses of nn.Module
+        self.cameras = self.cameras.to(device)
+        self.materials = self.materials.to(device)
+        self.lights = self.lights.to(device)
+
     def forward(self, fragments, meshes, **kwargs) -> torch.Tensor:
         cameras = kwargs.get("cameras", self.cameras)
         if cameras is None:
diff --git a/tests/test_render_meshes.py b/tests/test_render_meshes.py
@@ -1042,3 +1042,67 @@ def test_simple_sphere_outside_zfar(self):
                 )
 
             self.assertClose(rgb, image_ref, atol=0.05)
+
+    def test_to(self):
+        # Test moving all the tensors in the renderer to a new device
+        # to support multigpu rendering.
+        device1 = torch.device("cpu")
+
+        R, T = look_at_view_transform(1500, 0.0, 0.0)
+
+        # Init shader settings
+        materials = Materials(device=device1)
+        lights = PointLights(device=device1)
+        lights.location = torch.tensor([0.0, 0.0, +1000.0], device=device1)[None]
+
+        raster_settings = RasterizationSettings(
+            image_size=256, blur_radius=0.0, faces_per_pixel=1
+        )
+        cameras = FoVPerspectiveCameras(
+            device=device1, R=R, T=T, aspect_ratio=1.0, fov=60.0, zfar=100
+        )
+        rasterizer = MeshRasterizer(cameras=cameras, raster_settings=raster_settings)
+
+        blend_params = BlendParams(
+            1e-4,
+            1e-4,
+            background_color=torch.zeros(3, dtype=torch.float32, device=device1),
+        )
+
+        shader = SoftPhongShader(
+            lights=lights,
+            cameras=cameras,
+            materials=materials,
+            blend_params=blend_params,
+        )
+        renderer = MeshRenderer(rasterizer=rasterizer, shader=shader)
+
+        def _check_props_on_device(renderer, device):
+            self.assertEqual(renderer.rasterizer.cameras.device, device)
+            self.assertEqual(renderer.shader.cameras.device, device)
+            self.assertEqual(renderer.shader.lights.device, device)
+            self.assertEqual(renderer.shader.lights.ambient_color.device, device)
+            self.assertEqual(renderer.shader.materials.device, device)
+            self.assertEqual(renderer.shader.materials.ambient_color.device, device)
+
+        mesh = ico_sphere(2, device1)
+        verts_padded = mesh.verts_padded()
+        textures = TexturesVertex(
+            verts_features=torch.ones_like(verts_padded, device=device1)
+        )
+        mesh.textures = textures
+        _check_props_on_device(renderer, device1)
+
+        # Test rendering on cpu
+        output_images = renderer(mesh)
+        self.assertEqual(output_images.device, device1)
+
+        # Move renderer and mesh to another device and re render
+        # This also tests that background_color is correctly moved to
+        # the new device
+        device2 = torch.device("cuda:0")
+        renderer.to(device2)
+        mesh = mesh.to(device2)
+        _check_props_on_device(renderer, device2)
+        output_images = renderer(mesh)
+        self.assertEqual(output_images.device, device2)
