Merge pull request #36 from nicolas-chaulet/mac

Mac

.github/workflows/tests.yaml

@@ -7,30 +7,39 @@ on:
       - master
 
 jobs:
-    unittests:
-        runs-on: ubuntu-latest
-        strategy:
-          matrix:
-            python-version: [3.6, 3.7, 3.8]
-        steps:
-            - uses: actions/checkout@v2
-            - name: Set up Python ${{ matrix.python-version }}
-              uses: actions/setup-python@v2
-              with:
-                  python-version: ${{ matrix.python-version }}
-            - name: Install dependencies
-              run: |
-                  python -m pip install --upgrade pip
-                  pip install torch numpy scikit-learn flake8 setuptools
-            - name: Build package
-              run: |
-                  python setup.py build_ext --inplace
-            - name: Lint with flake8
-              run: |
-                  # stop the build if there are Python syntax errors or undefined names
-                  flake8 . --count --select=E9,F63,F7,F82 --show-source --statistics
-                  # exit-zero treats all errors as warnings. The GitHub editor is 127 chars wide
-                  flake8 . --count --exit-zero --max-complexity=10 --max-line-length=127 --statistics
-            - name: Test with unittest
-              run: |
-                  python -m unittest -v
+  unittests:
+    strategy:
+      matrix:
+        os: [ubuntu-latest, macos-latest, windows-latest]
+        python-version: [3.6, 3.7]
+    runs-on: ${{ matrix.os }}
+    steps:
+        - uses: actions/checkout@v2
+        - name: Set up Python ${{ matrix.python-version }}
+          uses: actions/setup-python@v2
+          with:
+              python-version: ${{ matrix.python-version }}
+        - name: Install dependencies
+          run: |
+              python -m pip install --upgrade pip
+              pip install numpy scikit-learn flake8 setuptools
+              
+        - name: Install torch windows + linux
+          if: ${{matrix.os != 'macos-latest'}}
+          run: pip install torch==1.5.0+cpu torchvision==0.6.0+cpu -f https://download.pytorch.org/whl/torch_stable.html
+        - name: Install torch macos
+          if: ${{matrix.os == 'macos-latest'}}
+          run: pip install torch
+
+        - name: Build package
+          run: |
+              python setup.py build_ext --inplace
+        - name: Lint with flake8
+          run: |
+              # stop the build if there are Python syntax errors or undefined names
+              flake8 . --count --select=E9,F63,F7,F82 --show-source --statistics
+              # exit-zero treats all errors as warnings. The GitHub editor is 127 chars wide
+              flake8 . --count --exit-zero --max-complexity=10 --max-line-length=127 --statistics
+        - name: Test with unittest
+          run: |
+              python -m unittest -v

CHANGELOG.md

@@ -1,3 +1,8 @@
+# UNRELEASED
+
+## Bug fix
+- CPU version works for MacOS
+
 # 0.6.2
 
 ## Bug fix

cpu/include/neighbors.h

@@ -9,15 +9,15 @@ using namespace std;
 
 template <typename scalar_t>
 int nanoflann_neighbors(vector<scalar_t>& queries, vector<scalar_t>& supports,
-                        vector<long>& neighbors_indices, vector<float>& dists, float radius,
+                        vector<int64_t>& neighbors_indices, vector<float>& dists, float radius,
                         int max_num, int mode, bool sorted);
 
 template <typename scalar_t>
 int batch_nanoflann_neighbors(vector<scalar_t>& queries, vector<scalar_t>& supports,
-                              vector<long>& q_batches, vector<long>& s_batches,
-                              vector<long>& neighbors_indices, vector<float>& dists, float radius,
-                              int max_num, int mode, bool sorted);
+                              vector<int64_t>& q_batches, vector<int64_t>& s_batches,
+                              vector<int64_t>& neighbors_indices, vector<float>& dists,
+                              float radius, int max_num, int mode, bool sorted);
 
 template <typename scalar_t>
 void nanoflann_knn_neighbors(vector<scalar_t>& queries, vector<scalar_t>& supports,
-                             vector<long>& neighbors_indices, vector<float>& dists, int k);
+                             vector<int64_t>& neighbors_indices, vector<float>& dists, int k);

cpu/src/ball_query.cpp

@@ -15,7 +15,7 @@ std::pair<at::Tensor, at::Tensor> ball_query(at::Tensor support, at::Tensor quer
 
     at::Tensor out;
     at::Tensor out_dists;
-    std::vector<long> neighbors_indices(query.size(0), 0);
+    std::vector<int64_t> neighbors_indices(query.size(0), 0);
     std::vector<float> neighbors_dists(query.size(0), -1);
 
     auto options = torch::TensorOptions().dtype(torch::kLong).device(torch::kCPU);
@@ -34,7 +34,7 @@ std::pair<at::Tensor, at::Tensor> ball_query(at::Tensor support, at::Tensor quer
                                                   neighbors_dists, radius, max_num, mode, sorted);
     });
     auto neighbors_dists_ptr = neighbors_dists.data();
-    long* neighbors_indices_ptr = neighbors_indices.data();
+    int64_t* neighbors_indices_ptr = neighbors_indices.data();
     if (mode == 0)
     {
         out =
@@ -73,7 +73,7 @@ std::pair<at::Tensor, at::Tensor> batch_ball_query(at::Tensor support, at::Tenso
     at::Tensor idx;
 
     at::Tensor dist;
-    std::vector<long> neighbors_indices;
+    std::vector<int64_t> neighbors_indices;
     std::vector<float> neighbors_dists;
 
     auto options = torch::TensorOptions().dtype(torch::kLong).device(torch::kCPU);
@@ -91,10 +91,11 @@ std::pair<at::Tensor, at::Tensor> batch_ball_query(at::Tensor support, at::Tenso
     query_batch = at::cat({at::zeros(1, query_batch.options()), query_batch.cumsum(0)}, 0);
     support_batch = degree(support_batch, batch_size);
     support_batch = at::cat({at::zeros(1, support_batch.options()), support_batch.cumsum(0)}, 0);
-    std::vector<long> query_batch_stl(query_batch.DATA_PTR<long>(),
-                                      query_batch.DATA_PTR<long>() + query_batch.numel());
-    std::vector<long> support_batch_stl(support_batch.DATA_PTR<long>(),
-                                        support_batch.DATA_PTR<long>() + support_batch.numel());
+    std::vector<int64_t> query_batch_stl(query_batch.DATA_PTR<int64_t>(),
+                                         query_batch.DATA_PTR<int64_t>() + query_batch.numel());
+    std::vector<int64_t> support_batch_stl(support_batch.DATA_PTR<int64_t>(),
+                                           support_batch.DATA_PTR<int64_t>() +
+                                               support_batch.numel());
 
     AT_DISPATCH_ALL_TYPES(query.scalar_type(), "batch_radius_search", [&] {
         std::vector<scalar_t> queries_stl(query.DATA_PTR<scalar_t>(),
@@ -107,7 +108,7 @@ std::pair<at::Tensor, at::Tensor> batch_ball_query(at::Tensor support, at::Tenso
             neighbors_dists, radius, max_num, mode, sorted);
     });
     auto neighbors_dists_ptr = neighbors_dists.data();
-    long* neighbors_indices_ptr = neighbors_indices.data();
+    int64_t* neighbors_indices_ptr = neighbors_indices.data();
 
     if (mode == 0)
     {

cpu/src/fps.cpp

@@ -15,7 +15,7 @@ at::Tensor fps(at::Tensor points, const int nsamples, bool random)
     auto out_options = torch::TensorOptions().dtype(torch::kLong).device(torch::kCPU);
     auto batch_size = points.size(0);
     auto out = torch::empty({batch_size, nsamples}, out_options);
-    auto out_a = out.accessor<long, 2>();
+    auto out_a = out.accessor<int64_t, 2>();
 
     for (ptrdiff_t b = 0; b < batch_size; b++)
     {

cpu/src/interpolate.cpp

@@ -19,7 +19,7 @@ at::Tensor knn_interpolate(at::Tensor features, at::Tensor idx, at::Tensor weigh
         auto output_a = output.accessor<scalar_t, 3>();
         auto features_a = features.accessor<scalar_t, 3>();
         auto weight_a = weight.accessor<scalar_t, 3>();
-        auto idx_a = idx.accessor<long, 3>();
+        auto idx_a = idx.accessor<int64_t, 3>();
 
         auto batch_size = idx.size(0);
         for (auto b = 0; b < batch_size; b++)
@@ -51,7 +51,7 @@ at::Tensor knn_interpolate_grad(at::Tensor grad_out, at::Tensor idx, at::Tensor
         auto output_a = output.accessor<scalar_t, 3>();
         auto grad_out_a = grad_out.accessor<scalar_t, 3>();
         auto weight_a = weight.accessor<scalar_t, 3>();
-        auto idx_a = idx.accessor<long, 3>();
+        auto idx_a = idx.accessor<int64_t, 3>();
 
         auto batch_size = idx.size(0);
         for (auto b = 0; b < batch_size; b++)

cpu/src/knn.cpp

@@ -12,7 +12,7 @@ std::pair<at::Tensor, at::Tensor> _single_batch_knn(at::Tensor support, at::Tens
     if (support.size(0) < k)
         TORCH_CHECK(false,
                     "Not enough points in support to find " + std::to_string(k) + " neighboors")
-    std::vector<long> neighbors_indices(query.size(0) * k, -1);
+    std::vector<int64_t> neighbors_indices(query.size(0) * k, -1);
     std::vector<float> neighbors_dists(query.size(0) * k, -1);
 
     auto options = torch::TensorOptions().dtype(torch::kLong).device(torch::kCPU);
@@ -29,7 +29,7 @@ std::pair<at::Tensor, at::Tensor> _single_batch_knn(at::Tensor support, at::Tens
                                           neighbors_dists, k);
     });
     auto neighbors_dists_ptr = neighbors_dists.data();
-    long* neighbors_indices_ptr = neighbors_indices.data();
+    int64_t* neighbors_indices_ptr = neighbors_indices.data();
     auto out = torch::from_blob(neighbors_indices_ptr, {query.size(0), k}, options = options);
     auto out_dists =
         torch::from_blob(neighbors_dists_ptr, {query.size(0), k}, options = options_dist);

cpu/src/neighbors.cpp

@@ -7,7 +7,7 @@
 
 template <typename scalar_t>
 int nanoflann_neighbors(vector<scalar_t>& queries, vector<scalar_t>& supports,
-                        vector<long>& neighbors_indices, vector<float>& dists, float radius,
+                        vector<int64_t>& neighbors_indices, vector<float>& dists, float radius,
                         int max_num, int mode, bool sorted)
 {
     // Initiate variables
@@ -138,9 +138,9 @@ int nanoflann_neighbors(vector<scalar_t>& queries, vector<scalar_t>& supports,
 
 template <typename scalar_t>
 int batch_nanoflann_neighbors(vector<scalar_t>& queries, vector<scalar_t>& supports,
-                              vector<long>& q_batches, vector<long>& s_batches,
-                              vector<long>& neighbors_indices, vector<float>& dists, float radius,
-                              int max_num, int mode, bool sorted)
+                              vector<int64_t>& q_batches, vector<int64_t>& s_batches,
+                              vector<int64_t>& neighbors_indices, vector<float>& dists,
+                              float radius, int max_num, int mode, bool sorted)
 {
     // Initiate variables
     // ******************
@@ -187,7 +187,6 @@ int batch_nanoflann_neighbors(vector<scalar_t>& queries, vector<scalar_t>& suppo
     // Search params
     nanoflann::SearchParams search_params;
     search_params.sorted = sorted;
-    std::chrono::microseconds duration_search(0);
     for (size_t i = 0; i < num_query_points; i++)
     {
         // Check if we changed batch
@@ -295,7 +294,7 @@ int batch_nanoflann_neighbors(vector<scalar_t>& queries, vector<scalar_t>& suppo
 
 template <typename scalar_t>
 void nanoflann_knn_neighbors(vector<scalar_t>& queries, vector<scalar_t>& supports,
-                             vector<long>& neighbors_indices, vector<float>& dists, int k)
+                             vector<int64_t>& neighbors_indices, vector<float>& dists, int k)
 {
     // Nanoflann related variables
     // ***************************

setup.py

@@ -67,7 +67,7 @@ def get_cmdclass():
 requirements = ["torch>=1.1.0"]
 
 url = "https://github.com/nicolas-chaulet/torch-points-kernels"
-__version__ = "0.6.3"
+__version__ = "0.6.4"
 setup(
     name="torch-points-kernels",
     version=__version__,