diff --git a/pytorch3d/common/workaround/symeig3x3.py b/pytorch3d/common/workaround/symeig3x3.py
index 955aa400..937485cb 100644
--- a/pytorch3d/common/workaround/symeig3x3.py
+++ b/pytorch3d/common/workaround/symeig3x3.py
@@ -104,7 +104,6 @@ class _SymEig3x3(nn.Module):
         )
 
         p = torch.sqrt(p2 / 6.0)
-        # pyrefly: ignore [unsupported-operation]
         B = (inputs - q[..., None, None] * self._identity) / p[..., None, None]
 
         r = torch.det(B) / 2.0
@@ -189,10 +188,8 @@ class _SymEig3x3(nn.Module):
         """
 
         # Find the eigenvector corresponding to alpha0, its eigenvalue is distinct
-        # pyrefly: ignore [unsupported-operation]
         ev0 = self._get_ev0(inputs - alpha0[..., None, None] * self._identity)
         u, v = self._get_uv(ev0)
-        # pyrefly: ignore [unsupported-operation]
         ev1 = self._get_ev1(inputs - alpha1[..., None, None] * self._identity, u, v)
         # Third eigenvector is computed as the cross-product of the other two
         ev2 = torch.cross(ev0, ev1, dim=-1)
@@ -267,7 +264,6 @@ class _SymEig3x3(nn.Module):
         """
 
         min_idx = w.abs().argmin(dim=-1)
-        # pyrefly: ignore [bad-index]
         rotation_2d = self._rotations_3d[min_idx].to(w)
 
         u = F.normalize((rotation_2d @ w[..., None])[..., 0], dim=-1)
diff --git a/pytorch3d/io/obj_io.py b/pytorch3d/io/obj_io.py
index 36b6ff53..c1dac7f1 100644
--- a/pytorch3d/io/obj_io.py
+++ b/pytorch3d/io/obj_io.py
@@ -794,7 +794,7 @@ def save_obj(
             mtl_path = str(mtl_path)
 
         # Save texture map to output folder
-        # pyre-fixme[16] # undefined attribute cpu
+        # undefined attribute cpu
         texture_map = texture_map.detach().cpu() * 255.0
         image = Image.fromarray(texture_map.numpy().astype(np.uint8))
         with _open_file(image_path, path_manager, "wb") as im_f:
diff --git a/pytorch3d/renderer/implicit/harmonic_embedding.py b/pytorch3d/renderer/implicit/harmonic_embedding.py
index 47382a0c..3fc2d8a7 100644
--- a/pytorch3d/renderer/implicit/harmonic_embedding.py
+++ b/pytorch3d/renderer/implicit/harmonic_embedding.py
@@ -136,16 +136,13 @@ class HarmonicEmbedding(torch.nn.Module):
             [..., (n_harmonic_functions * 2 + int(append_input)) * num_points_per_ray]
         """
         # [..., dim, n_harmonic_functions]
-        # pyrefly: ignore [unsupported-operation]
         embed = x[..., None] * self._frequencies
         # [..., 1, dim, n_harmonic_functions] + [2, 1, 1] => [..., 2, dim, n_harmonic_functions]
-        # pyrefly: ignore [bad-index]
         embed = embed[..., None, :, :] + self._zero_half_pi[..., None, None]
         # Use the trig identity cos(x) = sin(x + pi/2)
         # and do one vectorized call to sin([x, x+pi/2]) instead of (sin(x), cos(x)).
         embed = embed.sin()
         if diag_cov is not None:
-            # pyrefly: ignore [no-matching-overload]
             x_var = diag_cov[..., None] * torch.pow(self._frequencies, 2)
             exp_var = torch.exp(-0.5 * x_var)
             # [..., 2, dim, n_harmonic_functions]
@@ -184,7 +181,6 @@ class HarmonicEmbedding(torch.nn.Module):
         """
         return self.get_output_dim_static(
             input_dims,
-            # pyrefly: ignore [bad-argument-type]
             len(self._frequencies),
             self.append_input,
         )
diff --git a/pytorch3d/renderer/mesh/rasterizer.py b/pytorch3d/renderer/mesh/rasterizer.py
index bad4e070..2f7dba9c 100644
--- a/pytorch3d/renderer/mesh/rasterizer.py
+++ b/pytorch3d/renderer/mesh/rasterizer.py
@@ -239,12 +239,10 @@ class MeshRasterizer(nn.Module):
         if raster_settings.perspective_correct is not None:
             perspective_correct = raster_settings.perspective_correct
         else:
-            # pyrefly: ignore [missing-attribute]
             perspective_correct = cameras.is_perspective()
         if raster_settings.z_clip_value is not None:
             z_clip = raster_settings.z_clip_value
         else:
-            # pyrefly: ignore [missing-attribute]
             znear = cameras.get_znear()
             if isinstance(znear, torch.Tensor):
                 znear = znear.min().item()