v2.1.9: fix a small bug (assert isn't correct)

spconv/pytorch/core.py

@@ -144,7 +144,7 @@ class SparseConvTensor(metaclass=SpConvTensorMeta):
         """we need to replace x.features = F.relu(x.features) with x = x.replace_feature(F.relu(x.features))
         due to limit of torch.fx
         """
-        assert feature.shape[0] == self.indices.shape[0], "replaced num of features not equal to indices"
+        # assert feature.shape[0] == self.indices.shape[0], "replaced num of features not equal to indices"
         new_spt = SparseConvTensor(feature, self.indices, self.spatial_shape,
                                    self.batch_size, self.grid, self.voxel_num,
                                    self.indice_dict)

version.txt

-2.1.8
+2.1.9