suppress errors in `vision/fair/pytorch3d`

Differential Revision: D37172764

fbshipit-source-id: a2ec367e56de2781a17f5e708eb5832ec9d7e6b4

pytorch3d/ops/points_to_volumes.py

@@ -276,6 +276,7 @@ def add_pointclouds_to_volumes(
 
     # obtain the conversion mask
     n_per_pcl = pointclouds.num_points_per_cloud().type_as(pcl_feats)
+    # pyre-fixme[6]: For 1st param expected `Union[bool, float, int]` but got `Tensor`.
     mask = torch.arange(n_per_pcl.max(), dtype=pcl_feats.dtype, device=pcl_feats.device)
     mask = (mask[None, :] < n_per_pcl[:, None]).type_as(mask)
 
@@ -388,6 +389,7 @@ def add_points_features_to_volume_densities_features(
             mode=mode,
             min_weight=min_weight,
             mask=mask,
+            # pyre-fixme[6]: For 8th param expected `LongTensor` but got `Tensor`.
             grid_sizes=grid_sizes,
         )
 
@@ -595,7 +597,6 @@ def _splat_points_to_volumes(
     rX, rY, rZ = rXYZ.split(1, dim=2)
 
     # get random indices for the purpose of adding out-of-bounds values
-    # pyre-fixme[16]: `Tensor` has no attribute `new_zeros`.
     rand_idx = X.new_zeros(X.shape).random_(0, n_voxels)
 
     # iterate over the x, y, z indices of the 8-neighborhood (xdiff, ydiff, zdiff)
@@ -635,7 +636,6 @@ def _splat_points_to_volumes(
 
                 # scatter add casts the votes into the weight accumulator
                 # and the feature accumulator
-                # pyre-fixme[16]: `Tensor` has no attribute `scatter_add_`.
                 volume_densities.scatter_add_(1, idx_valid, w_valid)
 
                 # reshape idx_valid -> (minibatch, feature_dim, n_points)
@@ -719,6 +719,7 @@ def _round_points_to_volumes(
     X, Y, Z = XYZ.split(1, dim=2)
 
     # valid - binary indicators of votes that fall into the volume
+    # pyre-fixme[9]: grid_sizes has type `LongTensor`; used as `Tensor`.
     grid_sizes = grid_sizes.type_as(XYZ)
     valid = (
         (0 <= X)
@@ -743,7 +744,6 @@ def _round_points_to_volumes(
 
     # scatter add casts the votes into the weight accumulator
     # and the feature accumulator
-    # pyre-fixme[16]: `Tensor` has no attribute `scatter_add_`.
     volume_densities.scatter_add_(1, idx_valid, w_valid)
 
     # reshape idx_valid -> (minibatch, feature_dim, n_points)

pytorch3d/ops/sample_farthest_points.py

@@ -81,6 +81,7 @@ def sample_farthest_points(
     start_idxs = torch.zeros_like(lengths)
     if random_start_point:
         for n in range(N):
+            # pyre-fixme[6]: For 1st param expected `int` but got `Tensor`.
             start_idxs[n] = torch.randint(high=lengths[n], size=(1,)).item()
 
     with torch.no_grad():
@@ -128,14 +129,23 @@ def sample_farthest_points_naive(
     for n in range(N):
         # Initialize an array for the sampled indices, shape: (max_K,)
         sample_idx_batch = torch.full(
-            (max_K,), fill_value=-1, dtype=torch.int64, device=device
+            # pyre-fixme[6]: For 1st param expected `Union[List[int], Size,
+            #  typing.Tuple[int, ...]]` but got `Tuple[Tensor]`.
+            (max_K,),
+            fill_value=-1,
+            dtype=torch.int64,
+            device=device,
         )
 
         # Initialize closest distances to inf, shape: (P,)
         # This will be updated at each iteration to track the closest distance of the
         # remaining points to any of the selected points
         closest_dists = points.new_full(
-            (lengths[n],), float("inf"), dtype=torch.float32
+            # pyre-fixme[6]: For 1st param expected `Union[List[int], Size,
+            #  typing.Tuple[int, ...]]` but got `Tuple[Tensor]`.
+            (lengths[n],),
+            float("inf"),
+            dtype=torch.float32,
         )
 
         # Select a random point index and save it as the starting point
@@ -143,6 +153,10 @@ def sample_farthest_points_naive(
         sample_idx_batch[0] = selected_idx
 
         # If the pointcloud has fewer than K points then only iterate over the min
+        # pyre-fixme[6]: For 1st param expected `SupportsRichComparisonT` but got
+        #  `Tensor`.
+        # pyre-fixme[6]: For 2nd param expected `SupportsRichComparisonT` but got
+        #  `Tensor`.
         k_n = min(lengths[n], K[n])
 
         # Iteratively select points for a maximum of k_n
@@ -151,6 +165,8 @@ def sample_farthest_points_naive(
             # and all the other points. If a point has already been selected
             # it's distance will be 0.0 so it will not be selected again as the max.
             dist = points[n, selected_idx, :] - points[n, : lengths[n], :]
+            # pyre-fixme[58]: `**` is not supported for operand types `Tensor` and
+            #  `int`.
             dist_to_last_selected = (dist**2).sum(-1)  # (P - i)
 
             # If closer than currently saved distance to one of the selected

pytorch3d/ops/sample_points_from_meshes.py

@@ -172,6 +172,4 @@ def _rand_barycentric_coords(
     w0 = 1.0 - u_sqrt
     w1 = u_sqrt * (1.0 - v)
     w2 = u_sqrt * v
-    # pyre-fixme[7]: Expected `Tuple[torch.Tensor, torch.Tensor, torch.Tensor]` but
-    #  got `Tuple[float, typing.Any, typing.Any]`.
     return w0, w1, w2

pytorch3d/ops/subdivide_meshes.py

@@ -441,6 +441,8 @@ def _create_faces_index(faces_per_mesh: torch.Tensor, device=None):
 
     switch123_offset = F - faces_per_mesh  # e.g. (8, 5, 7)
 
+    # pyre-fixme[6]: For 1st param expected `Union[List[int], Size,
+    #  typing.Tuple[int, ...]]` but got `Tensor`.
     idx_diffs = torch.ones(4 * F, device=device, dtype=torch.int64)
     idx_diffs[switch1_idx] += switch123_offset
     idx_diffs[switch2_idx] += switch123_offset

pytorch3d/ops/utils.py

@@ -89,6 +89,8 @@ def wmean(
     args = {"dim": dim, "keepdim": keepdim}
 
     if weight is None:
+        # pyre-fixme[6]: For 1st param expected `Optional[dtype]` but got
+        #  `Union[Tuple[int], int]`.
         return x.mean(**args)
 
     if any(
@@ -97,6 +99,8 @@ def wmean(
     ):
         raise ValueError("wmean: weights are not compatible with the tensor")
 
+    # pyre-fixme[6]: For 1st param expected `Optional[dtype]` but got
+    #  `Union[Tuple[int], int]`.
     return (x * weight[..., None]).sum(**args) / weight[..., None].sum(**args).clamp(
         eps
     )

pytorch3d/ops/vert_align.py

@@ -87,7 +87,6 @@ def vert_align(
             padding_mode=padding_mode,
             align_corners=align_corners,
         )  # (N, C, 1, V)
-        # pyre-fixme[28]: Unexpected keyword argument `dim`.
         feat_sampled = feat_sampled.squeeze(dim=2).transpose(1, 2)  # (N, V, C)
         feats_sampled.append(feat_sampled)
     feats_sampled = torch.cat(feats_sampled, dim=2)  # (N, V, sum(C))
@@ -101,7 +100,6 @@ def vert_align(
                 .view(-1, 1)
                 .expand(-1, feats_sampled.shape[-1])
             )
-            # pyre-fixme[16]: `Tensor` has no attribute `gather`.
             feats_sampled = feats_sampled.gather(0, idx)  # (sum(V), C)
 
     return feats_sampled

pytorch3d/renderer/camera_conversions.py

@@ -80,7 +80,6 @@ def _opencv_from_cameras_projection(
     scale = scale.expand(-1, 2)
     c0 = image_size_wh / 2.0
 
-    # pyre-fixme[29]: `Union[BoundMethod[typing.Callable(torch.Tensor.__neg__)[[Named...
     principal_point = -p0_pytorch3d * scale + c0
     focal_length = focal_pytorch3d * scale
 

pytorch3d/renderer/cameras.py

@@ -401,6 +401,7 @@ class CamerasBase(TensorProperties):
 
         kwargs = {}
 
+        # pyre-fixme[16]: Module `cuda` has no attribute `LongTensor`.
         if not isinstance(index, (int, list, torch.LongTensor, torch.cuda.LongTensor)):
             msg = "Invalid index type, expected int, List[int] or torch.LongTensor; got %r"
             raise ValueError(msg % type(index))
@@ -600,7 +601,9 @@ class FoVPerspectiveCameras(CamerasBase):
         # so the so the z sign is 1.0.
         z_sign = 1.0
 
+        # pyre-fixme[58]: `/` is not supported for operand types `float` and `Tensor`.
         K[:, 0, 0] = 2.0 * znear / (max_x - min_x)
+        # pyre-fixme[58]: `/` is not supported for operand types `float` and `Tensor`.
         K[:, 1, 1] = 2.0 * znear / (max_y - min_y)
         K[:, 0, 2] = (max_x + min_x) / (max_x - min_x)
         K[:, 1, 2] = (max_y + min_y) / (max_y - min_y)
@@ -1755,6 +1758,8 @@ def get_ndc_to_screen_transform(
     K = torch.zeros((cameras._N, 4, 4), device=cameras.device, dtype=torch.float32)
     if not torch.is_tensor(image_size):
         image_size = torch.tensor(image_size, device=cameras.device)
+    # pyre-fixme[16]: Item `List` of `Union[List[typing.Any], Tensor, Tuple[Any,
+    #  ...]]` has no attribute `view`.
     image_size = image_size.view(-1, 2)  # of shape (1 or B)x2
     height, width = image_size.unbind(1)
 

pytorch3d/renderer/implicit/raymarching.py

@@ -158,7 +158,6 @@ class AbsorptionOnlyRaymarcher(torch.nn.Module):
         _check_density_bounds(rays_densities)
         total_transmission = torch.prod(1 - rays_densities, dim=-1, keepdim=True)
         opacities = 1.0 - total_transmission
-        # pyre-fixme[7]: Expected `Optional[torch.Tensor]` but got `float`.
         return opacities
 
 

pytorch3d/renderer/implicit/raysampling.py

@@ -180,6 +180,8 @@ class MultinomialRaysampler(torch.nn.Module):
                 # is not batched and does not support partial permutation
                 _, width, height, _ = xy_grid.shape
                 weights = xy_grid.new_ones(batch_size, width * height)
+            # pyre-fixme[6]: For 2nd param expected `int` but got `Union[bool,
+            #  float, int]`.
             rays_idx = _safe_multinomial(weights, num_rays)[..., None].expand(-1, -1, 2)
 
             xy_grid = torch.gather(xy_grid.reshape(batch_size, -1, 2), 1, rays_idx)[
@@ -478,7 +480,6 @@ def _safe_multinomial(input: torch.Tensor, num_samples: int) -> torch.Tensor:
     # in some versions of Pytorch, zero probabilty samples can be drawn without an error
     # due to this bug: https://github.com/pytorch/pytorch/issues/50034. Handle this case:
     repl = (input > 0.0).sum(dim=-1) < num_samples
-    # pyre-fixme[16]: Undefined attribute `torch.ByteTensor` has no attribute `any`.
     if repl.any():
         res[repl] = torch.multinomial(input[repl], num_samples, replacement=True)
 
@@ -515,7 +516,7 @@ def _xy_to_ray_bundle(
     """
     batch_size = xy_grid.shape[0]
     spatial_size = xy_grid.shape[1:-1]
-    n_rays_per_image = spatial_size.numel()  # pyre-ignore
+    n_rays_per_image = spatial_size.numel()
 
     # ray z-coords
     rays_zs = xy_grid.new_empty((0,))

pytorch3d/renderer/mesh/clip.py

@@ -254,7 +254,6 @@ def _find_verts_intersecting_clipping_plane(
 
     # p1, p2, p3 are (T, 3) tensors storing the corresponding (x, y, z) coordinates
     # of p1_face_ind, p2_face_ind, p3_face_ind
-    # pyre-ignore[16]
     p1 = face_verts.gather(1, p1_face_ind[:, None, None].expand(-1, -1, 3)).squeeze(1)
     p2 = face_verts.gather(1, p2_face_ind[:, None, None].expand(-1, -1, 3)).squeeze(1)
     p3 = face_verts.gather(1, p3_face_ind[:, None, None].expand(-1, -1, 3)).squeeze(1)
@@ -398,7 +397,6 @@ def clip_faces(
     #      into a smaller quadrilateral and split into two triangles)
     #####################################################################################
 
-    # pyre-ignore[16]:
     faces_unculled = ~faces_culled
     # Case 1:  no clipped verts or culled faces
     cases1_unclipped = (faces_num_clipped_verts == 0) & faces_unculled
@@ -434,7 +432,13 @@ def clip_faces(
     # These will then be filled in for each case.
     ###########################################
     F_clipped = (
-        F + faces_delta_cum[-1].item() + faces_delta[-1].item()
+        F
+        # pyre-fixme[58]: `+` is not supported for operand types `int` and
+        #  `Union[bool, float, int]`.
+        + faces_delta_cum[-1].item()
+        # pyre-fixme[58]: `+` is not supported for operand types `int` and
+        #  `Union[bool, float, int]`.
+        + faces_delta[-1].item()
     )  # Total number of faces in the new Meshes
     face_verts_clipped = torch.zeros(
         (F_clipped, 3, 3), dtype=face_verts_unclipped.dtype, device=device

pytorch3d/renderer/mesh/textures.py

@@ -66,8 +66,11 @@ def _list_to_padded_wrapper(
                 "list_to_padded requires tensors to have the same number of dimensions"
             )
             raise ValueError(msg)
+        # pyre-fixme[6]: For 2nd param expected `int` but got `Union[bool, float, int]`.
         x_reshaped.append(y.reshape(-1, D))
     x_padded = list_to_padded(x_reshaped, pad_size=pad_size, pad_value=pad_value)
+    # pyre-fixme[58]: `+` is not supported for operand types `Tuple[int, int]` and
+    #  `Size`.
     return x_padded.reshape((N, -1) + reshape_dims)
 
 
@@ -96,8 +99,11 @@ def _padded_to_list_wrapper(
     N, M = x.shape[:2]
     reshape_dims = x.shape[2:]
     D = torch.prod(torch.tensor(reshape_dims)).item()
+    # pyre-fixme[6]: For 3rd param expected `int` but got `Union[bool, float, int]`.
     x_reshaped = x.reshape(N, M, D)
     x_list = padded_to_list(x_reshaped, split_size=split_size)
+    # pyre-fixme[58]: `+` is not supported for operand types `Tuple[typing.Any]` and
+    #  `Size`.
     x_list = [xl.reshape((xl.shape[0],) + reshape_dims) for xl in x_list]
     return x_list
 
@@ -132,8 +138,6 @@ def _pad_texture_maps(
             image_BCHW = image.permute(2, 0, 1)[None]
             new_image_BCHW = interpolate(
                 image_BCHW,
-                # pyre-fixme[6]: Expected `Optional[int]` for 2nd param but got
-                #  `Tuple[int, int]`.
                 size=max_shape,
                 mode="bilinear",
                 align_corners=align_corners,

pytorch3d/renderer/splatter_blend.py

@@ -130,12 +130,15 @@ def _get_splat_kernel_normalization(
 
     epsilon = 0.05
     normalization_constant = torch.exp(
-        -(offsets**2).sum(dim=1) / (2 * sigma**2)
+        # pyre-fixme[58]: `**` is not supported for operand types `Tensor` and `int`.
+        -(offsets**2).sum(dim=1)
+        / (2 * sigma**2)
     ).sum()
 
     # We add an epsilon to the normalization constant to ensure the gradient will travel
     # through non-boundary pixels' normalization factor, see Sec. 3.3.1 in "Differentia-
     # ble Surface Rendering via Non-Differentiable Sampling", Cole et al.
+    # pyre-fixme[58]: `/` is not supported for operand types `float` and `Tensor`.
     return (1 + epsilon) / normalization_constant
 
 
@@ -260,6 +263,7 @@ def _compute_splatting_colors_and_weights(
         torch.floor(pixel_coords_screen[..., :2]) - pixel_coords_screen[..., :2] + 0.5
     ).view((N, H, W, K, 1, 2))
 
+    # pyre-fixme[58]: `**` is not supported for operand types `Tensor` and `int`.
     dist2_p_q = torch.sum((q_to_px_center + offsets) ** 2, dim=5)  # (N, H, W, K, 9)
     splat_weights = torch.exp(-dist2_p_q / (2 * sigma**2))
     alpha = colors[..., 3:4]
@@ -413,6 +417,7 @@ def _normalize_and_compose_all_layers(
 
     # Normalize each of bg/surface/fg splat layers separately.
     normalization_scales = 1.0 / (
+        # pyre-fixme[58]: `/` is not supported for operand types `float` and `Tensor`.
         torch.maximum(
             splatted_weights_per_occlusion_layer,
             torch.tensor([1.0], device=device),

pytorch3d/renderer/utils.py

@@ -394,6 +394,7 @@ def ndc_grid_sample(
 
     grid_ndc_flat = grid_ndc.reshape(batch, -1, 1, 2)
 
+    # pyre-fixme[6]: For 2nd param expected `Tuple[int, int]` but got `Size`.
     grid_flat = ndc_to_grid_sample_coords(grid_ndc_flat, input.shape[2:])
 
     sampled_input_flat = torch.nn.functional.grid_sample(

pytorch3d/structures/meshes.py

@@ -890,7 +890,6 @@ class Meshes:
 
             # NOTE: this is already applying the area weighting as the magnitude
             # of the cross product is 2 x area of the triangle.
-            # pyre-fixme[16]: `Tensor` has no attribute `index_add`.
             verts_normals = verts_normals.index_add(
                 0,
                 faces_packed[:, 1],

pytorch3d/structures/utils.py

@@ -210,7 +210,6 @@ def padded_to_packed(
 
     # Convert to packed using pad value
     if pad_value is not None:
-        # pyre-fixme[16]: `ByteTensor` has no attribute `any`.
         mask = x_packed.ne(pad_value).any(-1)
         x_packed = x_packed[mask]
         return x_packed

pytorch3d/transforms/rotation_conversions.py

@@ -50,6 +50,7 @@ def quaternion_to_matrix(quaternions: torch.Tensor) -> torch.Tensor:
         Rotation matrices as tensor of shape (..., 3, 3).
     """
     r, i, j, k = torch.unbind(quaternions, -1)
+    # pyre-fixme[58]: `/` is not supported for operand types `float` and `Tensor`.
     two_s = 2.0 / (quaternions * quaternions).sum(-1)
 
     o = torch.stack(
@@ -131,9 +132,17 @@ def matrix_to_quaternion(matrix: torch.Tensor) -> torch.Tensor:
     # we produce the desired quaternion multiplied by each of r, i, j, k
     quat_by_rijk = torch.stack(
         [
+            # pyre-fixme[58]: `**` is not supported for operand types `Tensor` and
+            #  `int`.
             torch.stack([q_abs[..., 0] ** 2, m21 - m12, m02 - m20, m10 - m01], dim=-1),
+            # pyre-fixme[58]: `**` is not supported for operand types `Tensor` and
+            #  `int`.
             torch.stack([m21 - m12, q_abs[..., 1] ** 2, m10 + m01, m02 + m20], dim=-1),
+            # pyre-fixme[58]: `**` is not supported for operand types `Tensor` and
+            #  `int`.
             torch.stack([m02 - m20, m10 + m01, q_abs[..., 2] ** 2, m12 + m21], dim=-1),
+            # pyre-fixme[58]: `**` is not supported for operand types `Tensor` and
+            #  `int`.
             torch.stack([m10 - m01, m20 + m02, m21 + m12, q_abs[..., 3] ** 2], dim=-1),
         ],
         dim=-2,
@@ -148,7 +157,7 @@ def matrix_to_quaternion(matrix: torch.Tensor) -> torch.Tensor:
     # forall i; we pick the best-conditioned one (with the largest denominator)
 
     return quat_candidates[
-        F.one_hot(q_abs.argmax(dim=-1), num_classes=4) > 0.5, :  # pyre-ignore[16]
+        F.one_hot(q_abs.argmax(dim=-1), num_classes=4) > 0.5, :
     ].reshape(batch_dim + (4,))
 
 
@@ -314,6 +323,7 @@ def random_quaternions(
     """
     if isinstance(device, str):
         device = torch.device(device)
+    # pyre-fixme[6]: For 2nd param expected `dtype` but got `Optional[dtype]`.
     o = torch.randn((n, 4), dtype=dtype, device=device)
     s = (o * o).sum(1)
     o = o / _copysign(torch.sqrt(s), o[:, 0])[:, None]

pytorch3d/transforms/se3.py

@@ -194,9 +194,12 @@ def _se3_V_matrix(
     V = (
         torch.eye(3, dtype=log_rotation.dtype, device=log_rotation.device)[None]
         + log_rotation_hat
+        # pyre-fixme[58]: `**` is not supported for operand types `Tensor` and `int`.
         * ((1 - torch.cos(rotation_angles)) / (rotation_angles**2))[:, None, None]
         + (
             log_rotation_hat_square
+            # pyre-fixme[58]: `**` is not supported for operand types `Tensor` and
+            #  `int`.
             * ((rotation_angles - torch.sin(rotation_angles)) / (rotation_angles**3))[
                 :, None, None
             ]
@@ -211,6 +214,7 @@ def _get_se3_V_input(log_rotation: torch.Tensor, eps: float = 1e-4):
     A helper function that computes the input variables to the `_se3_V_matrix`
     function.
     """
+    # pyre-fixme[58]: `**` is not supported for operand types `Tensor` and `int`.
     nrms = (log_rotation**2).sum(-1)
     rotation_angles = torch.clamp(nrms, eps).sqrt()
     log_rotation_hat = hat(log_rotation)

pytorch3d/transforms/so3.py

@@ -160,6 +160,7 @@ def _so3_exp_map(
     nrms = (log_rot * log_rot).sum(1)
     # phis ... rotation angles
     rot_angles = torch.clamp(nrms, eps).sqrt()
+    # pyre-fixme[58]: `/` is not supported for operand types `float` and `Tensor`.
     rot_angles_inv = 1.0 / rot_angles
     fac1 = rot_angles_inv * rot_angles.sin()
     fac2 = rot_angles_inv * rot_angles_inv * (1.0 - rot_angles.cos())
@@ -167,8 +168,8 @@ def _so3_exp_map(
     skews_square = torch.bmm(skews, skews)
 
     R = (
-        # pyre-fixme[16]: `float` has no attribute `__getitem__`.
         fac1[:, None, None] * skews
+        # pyre-fixme[16]: `float` has no attribute `__getitem__`.
         + fac2[:, None, None] * skews_square
         + torch.eye(3, dtype=log_rot.dtype, device=log_rot.device)[None]
     )
@@ -216,6 +217,7 @@ def so3_log_map(
     # 2nd order Taylor expansion: phi_factor = 0.5 + (1.0 / 12) * phi**2
     phi_factor = torch.empty_like(phi)
     ok_denom = phi_sin.abs() > (0.5 * eps)
+    # pyre-fixme[58]: `**` is not supported for operand types `Tensor` and `int`.
     phi_factor[~ok_denom] = 0.5 + (phi[~ok_denom] ** 2) * (1.0 / 12)
     phi_factor[ok_denom] = phi[ok_denom] / (2.0 * phi_sin[ok_denom])
 

pytorch3d/transforms/transform3d.py

@@ -556,7 +556,9 @@ class Scale(Transform3d):
         Return the inverse of self._matrix.
         """
         xyz = torch.stack([self._matrix[:, i, i] for i in range(4)], dim=1)
+        # pyre-fixme[58]: `/` is not supported for operand types `float` and `Tensor`.
         ixyz = 1.0 / xyz
+        # pyre-fixme[6]: For 1st param expected `Tensor` but got `float`.
         imat = torch.diag_embed(ixyz, dim1=1, dim2=2)
         return imat