Merge branch 'main' into musa_mmcv_main

Author: hanhaowen-mt

docs/zh_cn/mmcv-logo.png

@@ -0,0 +1 @@
+../docs/mmcv-logo.png

mmcv/ops/csrc/pytorch/npu/stack_group_points_npu.cpp

@@ -0,0 +1,25 @@
+#include "pytorch_npu_helper.hpp"
+
+using namespace NPU_NAME_SPACE;
+using namespace std;
+
+void stack_group_points_forward_npu(int b, int c, int n, int nsample,
+                                    const Tensor features_tensor,
+                                    const Tensor features_batch_cnt_tensor,
+                                    const Tensor idx_tensor,
+                                    const Tensor idx_batch_cnt_tensor,
+                                    Tensor out_tensor) {
+  EXEC_NPU_CMD(aclnnStackGroupPoints, features_tensor,
+               features_batch_cnt_tensor, idx_tensor, idx_batch_cnt_tensor,
+               out_tensor);
+}
+
+void stack_group_points_forward_impl(int b, int c, int n, int nsample,
+                                     const Tensor features_tensor,
+                                     const Tensor features_batch_cnt_tensor,
+                                     const Tensor idx_tensor,
+                                     const Tensor idx_batch_cnt_tensor,
+                                     Tensor out_tensor);
+
+REGISTER_NPU_IMPL(stack_group_points_forward_impl,
+                  stack_group_points_forward_npu);

tests/test_ops/test_group_points.py

@@ -89,27 +89,27 @@ def test_grouping_points(dtype, device):
     assert torch.allclose(output, expected_output)
 
 
-@pytest.mark.skipif(
-    not (torch.cuda.is_available() or is_musa_available()),
-    reason='requires CUDA/MUSA support')
-@pytest.mark.parametrize('dtype', [
+
+@pytest.mark.parametrize('device', [
     pytest.param(
-        torch.half,
+        'cuda',
         marks=pytest.mark.skipif(
-            is_musa_available(),
-            reason='TODO [email protected]: not supported yet')),
-    torch.float,
+            not IS_CUDA_AVAILABLE, reason='requires CUDA support')),
     pytest.param(
-        torch.double,
+        'npu',
         marks=pytest.mark.skipif(
-            is_musa_available(),
-            reason='TODO [email protected]: not supported yet'))
+            not IS_NPU_AVAILABLE, reason='requires NPU support')),
+    pytest.param(
+        'musa',
+        marks=pytest.mark.skipif(
+            not IS_MUSA_AVAILABLE, reason='requires MUSA support'))
 ])
-def test_stack_grouping_points(dtype):
-    if torch.cuda.is_available():
-        device = 'cuda'
-    elif is_musa_available():
-        device = 'musa'
+@pytest.mark.parametrize('dtype', [torch.half, torch.float, torch.double])
+def test_stack_grouping_points(dtype, device):
+    if device == 'npu' and dtype == torch.double:
+        return
+    if device == 'musa' and dtype == torch.double:
+        return
     idx = torch.tensor([[0, 0, 0], [3, 3, 3], [8, 8, 8], [1, 1, 1], [0, 0, 0],
                         [2, 2, 2], [0, 0, 0], [6, 6, 6], [9, 9, 9], [0, 0, 0],
                         [1, 1, 1], [0, 0, 0]]).int().to(device)