small changes, publish v2.1.8 prebuilts

README.md

@@ -34,7 +34,7 @@
 [![Build Status](https://github.com/traveller59/spconv/workflows/build/badge.svg)](https://github.com/traveller59/spconv/actions?query=workflow%3Abuild) 
 
 
-|                | PyPi   | Install  |Downloads  |
+|                | PyPI   | Install  |Downloads  |
 | -------------- |:---------------------:| ---------------------:| ---------------------:| 
 | CPU (Linux Only) | [![PyPI Version][pypi-ver-cpu]][pypi-url-cpu] | ```pip install spconv``` | [![pypi monthly download][pypi-download-cpu]][pypi-url-cpu] | 
 | CUDA 10.2 | [![PyPI Version][pypi-ver-102]][pypi-url-102] | ```pip install spconv-cu102``` | [![pypi monthly download][pypi-download-102]][pypi-url-102] | 
@@ -66,6 +66,8 @@ Spconv 1.x users **NEED READ [THIS](docs/SPCONV_2_BREAKING_CHANGEs.md)** before
 
 ## Spconv 2.x Development and Roadmap
 
+Spconv 2.2 development has started. See [this issue](https://github.com/traveller59/spconv/issues/380) for more details.
+
 See [dev plan](docs/SPCONV_DEVELOP_PLAN.md). A complete guide of spconv development will be released soon.
 
 ## Usage

example/voxel_gen.py

@@ -21,6 +21,7 @@ import torch
 
 
 def main():
+    np.random.seed(50051)
     # voxel gen source code: spconv/csrc/sparse/pointops.py
     gen = Point2VoxelCPU3d(vsize_xyz=[0.1, 0.1, 0.1],
                            coors_range_xyz=[-80, -80, -2, 80, 80, 6],
@@ -36,7 +37,7 @@ def main():
     voxels_np = voxels_tv.numpy_view()
     indices_np = indices_tv.numpy_view()
     num_p_in_vx_np = num_p_in_vx_tv.numpy_view()
-    print("------Raw Voxels-------")
+    print(f"------Raw Voxels {voxels_np.shape[0]}-------")
     print(voxels_np[0])
     # run voxel gen and FILL MEAN VALUE to voxel remain
     voxels_tv, indices_tv, num_p_in_vx_tv = gen.point_to_voxel_empty_mean(
@@ -49,6 +50,7 @@ def main():
 
 
 def main_point_with_features():
+    np.random.seed(50051)
     # voxel gen source code: spconv/csrc/sparse/pointops.py
     gen = Point2VoxelCPU3d(
         vsize_xyz=[0.1, 0.1, 0.1],
@@ -68,7 +70,7 @@ def main_point_with_features():
     voxels_np = voxels_tv.numpy_view()
     indices_np = indices_tv.numpy_view()
     num_p_in_vx_np = num_p_in_vx_tv.numpy_view()
-    print("------Raw Voxels-------")
+    print(f"------Raw Voxels {voxels_np.shape[0]}-------")
     print(voxels_np[0])
     # run voxel gen and FILL MEAN VALUE to voxel remain
     voxels_tv, indices_tv, num_p_in_vx_tv = gen.point_to_voxel_empty_mean(
@@ -81,6 +83,7 @@ def main_point_with_features():
 
 
 def main_pytorch_voxel_gen():
+    np.random.seed(50051)
     # voxel gen source code: spconv/csrc/sparse/pointops.py
     gen = PointToVoxel(vsize_xyz=[0.1, 0.1, 0.1],
                        coors_range_xyz=[-80, -80, -2, 80, 80, 6],
@@ -94,7 +97,7 @@ def main_pytorch_voxel_gen():
     voxels_np = voxels_th.numpy()
     indices_np = indices_th.numpy()
     num_p_in_vx_np = num_p_in_vx_th.numpy()
-    print("------Raw Voxels-------")
+    print(f"------Raw Voxels {voxels_np.shape[0]}-------")
     print(voxels_np[0])
     # run voxel gen and FILL MEAN VALUE to voxel remain
     voxels_tv, indices_tv, num_p_in_vx_tv = gen(pc_th, empty_mean=True)
@@ -106,6 +109,7 @@ def main_pytorch_voxel_gen():
 
 
 def main_pytorch_voxel_gen_cuda():
+    np.random.seed(50051)
     # voxel gen source code: spconv/csrc/sparse/pointops.py
     device = torch.device("cuda:0")
     gen = PointToVoxel(vsize_xyz=[0.1, 0.1, 0.1],
@@ -121,7 +125,7 @@ def main_pytorch_voxel_gen_cuda():
     voxels_np = voxels_th.cpu().numpy()
     indices_np = indices_th.cpu().numpy()
     num_p_in_vx_np = num_p_in_vx_th.cpu().numpy()
-    print("------Raw Voxels-------")
+    print(f"------Raw Voxels {voxels_np.shape[0]}-------")
     print(voxels_np[0])
     # run voxel gen and FILL MEAN VALUE to voxel remain
     voxels_tv, indices_tv, num_p_in_vx_tv = gen(pc_th, empty_mean=True)

spconv/pytorch/__init__.py

@@ -3,17 +3,21 @@ from pathlib import Path
 
 import numpy as np
 import torch
-
-from spconv.pytorch import ops, functional
-from spconv.pytorch.conv import (SparseConv2d, SparseConv3d,
+from spconv.pytorch import functional, ops
+from spconv.pytorch.conv import (SparseConv1d, SparseConv2d, SparseConv3d,
+                                 SparseConv4d, SparseConvTranspose1d,
                                  SparseConvTranspose2d, SparseConvTranspose3d,
+                                 SparseConvTranspose4d, SparseInverseConv1d,
                                  SparseInverseConv2d, SparseInverseConv3d,
-                                 SubMConv2d, SubMConv3d)
+                                 SparseInverseConv4d, SubMConv1d, SubMConv2d,
+                                 SubMConv3d, SubMConv4d)
 from spconv.pytorch.core import SparseConvTensor
 from spconv.pytorch.identity import Identity
-from spconv.pytorch.modules import SparseModule, SparseSequential, assign_name_for_sparse_modules
+from spconv.pytorch.modules import (SparseModule, SparseSequential,
+                                    assign_name_for_sparse_modules)
 from spconv.pytorch.ops import ConvAlgo
-from spconv.pytorch.pool import SparseMaxPool2d, SparseMaxPool3d
+from spconv.pytorch.pool import (SparseMaxPool1d, SparseMaxPool2d,
+                                 SparseMaxPool3d, SparseMaxPool4d)
 from spconv.pytorch.tables import AddTable, ConcatTable, JoinTable
 
 

spconv/pytorch/core.py

@@ -140,10 +140,11 @@ class SparseConvTensor(metaclass=SpConvTensorMeta):
             self.thrust_allocator = ThrustSortAllocator(features.device)
         self._timer = CUDAKernelTimer(enable_timer)
 
-    def replace_feature(self, feature):
+    def replace_feature(self, feature: torch.Tensor):
         """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"
         new_spt = SparseConvTensor(feature, self.indices, self.spatial_shape,
                                    self.batch_size, self.grid, self.voxel_num,
                                    self.indice_dict)