1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
|
# ----------------------------------------------------------------------------
# - Open3D: www.open3d.org -
# ----------------------------------------------------------------------------
# Copyright (c) 2018-2024 www.open3d.org
# SPDX-License-Identifier: MIT
# ----------------------------------------------------------------------------
import open3d as o3d
import open3d.core as o3c
import numpy as np
import pytest
import sys
import os
sys.path.append(os.path.dirname(os.path.realpath(__file__)) + "/..")
from open3d_test import list_devices_with_torch, torch_available
if torch_available():
import torch
import torch.utils.dlpack
@pytest.mark.parametrize("device", list_devices_with_torch())
def test_tensor_to_pytorch_scope(device):
if not torch_available():
return
src_t = np.array([[10, 11, 12.], [13., 14., 15.]])
def get_dst_t():
o3d_t = o3c.Tensor(src_t, device=device) # Copy
dst_t = torch.utils.dlpack.from_dlpack(o3d_t.to_dlpack())
return dst_t
dst_t = get_dst_t().cpu().numpy()
np.testing.assert_equal(dst_t, src_t)
@pytest.mark.parametrize("device", list_devices_with_torch())
def test_tensor_from_to_pytorch(device):
if not torch_available():
return
device_id = device.get_id()
device_type = device.get_type()
# a, b, c share memory.
a = torch.ones((2, 2))
if device_type == o3c.Device.DeviceType.CUDA:
a = a.cuda(device_id)
b = o3c.Tensor.from_dlpack(torch.utils.dlpack.to_dlpack(a))
c = torch.utils.dlpack.from_dlpack(b.to_dlpack())
a[0, 0] = 100
c[0, 1] = 200
r = np.array([[100., 200.], [1., 1.]])
np.testing.assert_equal(r, a.cpu().numpy())
np.testing.assert_equal(r, b.cpu().numpy())
np.testing.assert_equal(r, c.cpu().numpy())
# Special strides.
np_r = np.random.randint(10, size=(10, 10)).astype(np.int32)
th_r = torch.Tensor(np_r)
th_t = th_r[1:10:2, 1:10:3].T
if device_type == o3c.Device.DeviceType.CUDA:
th_t = th_t.cuda(device_id)
o3_t = o3c.Tensor.from_dlpack(torch.utils.dlpack.to_dlpack(th_t))
np.testing.assert_equal(th_t.cpu().numpy(), o3_t.cpu().numpy())
th_t[0, 0] = 100
np.testing.assert_equal(th_t.cpu().numpy(), o3_t.cpu().numpy())
# Zero-sized tensors.
for shape in [(), (0), (0, 0), (0, 3)]:
np_t = np.ones(shape)
th_t = torch.Tensor(np_t)
o3_t = o3c.Tensor.from_dlpack(torch.utils.dlpack.to_dlpack(th_t))
th_t_v2 = torch.utils.dlpack.from_dlpack(o3_t.to_dlpack())
np.testing.assert_equal(o3_t.cpu().numpy(), np_t)
np.testing.assert_equal(th_t_v2.cpu().numpy(), np_t)
def test_tensor_numpy_to_open3d_to_pytorch():
if not torch_available():
return
# Numpy -> Open3D -> PyTorch all share the same memory
a = np.ones((2, 2))
b = o3c.Tensor.from_numpy(a)
c = torch.utils.dlpack.from_dlpack(b.to_dlpack())
a[0, 0] = 100
c[0, 1] = 200
r = np.array([[100., 200.], [1., 1.]])
np.testing.assert_equal(r, a)
np.testing.assert_equal(r, b.cpu().numpy())
np.testing.assert_equal(r, c.cpu().numpy())
|
