(self)
| 41 | print(in_features, out_features) |
| 42 | |
| 43 | def test_backward(self): |
| 44 | IC, OC = 3, 5 |
| 45 | coordinates = torch.IntTensor([[0, 1, -1], [0, 2, 1]]) |
| 46 | in_features = torch.rand(len(coordinates), IC) |
| 47 | |
| 48 | manager = _C.CoordinateMapManager() |
| 49 | in_key, unique_inverse_map = manager.insert_and_map(coordinates, [1, 1], "") |
| 50 | kernel_size = [3, 3] |
| 51 | kernel_stride = [2, 2] |
| 52 | kernel_dilation = [1, 1] |
| 53 | out_key = _C.CoordinateMapKey(3) |
| 54 | |
| 55 | # size, in, out |
| 56 | kernel = torch.rand(9, IC, OC) |
| 57 | |
| 58 | out_features = _C.ConvolutionForwardCPU( |
| 59 | in_features, |
| 60 | kernel, |
| 61 | kernel_size, |
| 62 | kernel_stride, |
| 63 | kernel_dilation, |
| 64 | _C.RegionType.HYPER_CUBE, |
| 65 | torch.IntTensor(), |
| 66 | in_key, |
| 67 | out_key, |
| 68 | manager, |
| 69 | ) |
| 70 | |
| 71 | out_feat_grad = torch.rand_like(out_features) |
| 72 | in_feat_grad, kernel_grad = _C.ConvolutionBackwardCPU( |
| 73 | in_features, |
| 74 | out_feat_grad, |
| 75 | kernel, |
| 76 | kernel_size, |
| 77 | kernel_stride, |
| 78 | kernel_dilation, |
| 79 | _C.RegionType.HYPER_CUBE, |
| 80 | torch.IntTensor(), |
| 81 | in_key, |
| 82 | out_key, |
| 83 | manager, |
| 84 | ) |
| 85 | |
| 86 | def test_pcd(self): |
| 87 | IC, OC = 3, 16 |
nothing calls this directly
no test coverage detected