PyTorch Tests Failing: 1 Failure 2 Errors

I reinstalled PyTorch, and out of curiosity decided to run PyTorch tests from Github repo. I had 1 failure and 2 errors :disappointed_relieved: Does it mean my PyTorch installation is broken?

root@4ef8202959fd:~# git clone https://github.com/pytorch/pytorch.git
Cloning into 'pytorch'...
remote: Counting objects: 34197, done.
remote: Compressing objects: 100% (6/6), done.
remote: Total 34197 (delta 12), reused 12 (delta 12), pack-reused 34179
Receiving objects: 100% (34197/34197), 13.77 MiB | 6.33 MiB/s, done.
Resolving deltas: 100% (25739/25739), done.
Checking connectivity... done.
root@4ef8202959fd:~# cd pytorch/test/
root@4ef8202959fd:~/pytorch/test# bash run_test.sh
~/pytorch/test ~/pytorch/test
Running torch tests
......................................sss.............F........s.............E...........................................E...............................................................................ss.
======================================================================
ERROR: test_has_storage_numpy (__main__.TestTorch)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "test_torch.py", line 142, in test_has_storage_numpy
    self.assertIsNotNone(torch.DoubleTensor(arr).storage())
RuntimeError: tried to construct a tensor from a float sequence, but found an item of type numpy.float32 at index (0)

======================================================================
ERROR: test_neg (__main__.TestTorch)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "test_torch.py", line 397, in test_neg
    self._test_neg(self, lambda t: t)
  File "test_torch.py", line 385, in _test_neg
    res_neg.neg_()
AttributeError: 'torch.LongTensor' object has no attribute 'neg_'

======================================================================
FAIL: test_dim_reduction (__main__.TestTorch)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "test_torch.py", line 245, in test_dim_reduction
    self._test_dim_reduction(self, lambda t: t)
  File "test_torch.py", line 242, in _test_dim_reduction
    test_multidim(x, singleton_dim)
  File "test_torch.py", line 221, in test_multidim
    self.assertEqual(fn(x, dim).unsqueeze(dim), fn(x, dim, keepdim=True))
  File "/root/pytorch/test/common.py", line 207, in assertEqual
    assertTensorsEqual(x, y)
  File "/root/pytorch/test/common.py", line 187, in assertTensorsEqual
    super(TestCase, self).assertEqual(a.size(), b.size())
AssertionError: torch.Size([1, 1, 4, 5]) != torch.Size([1, 4, 5])

----------------------------------------------------------------------
Ran 204 tests in 23.488s

FAILED (failures=1, errors=2, skipped=6)

Ahhh I am an Idiot. Checking out tag v.0.2.0 and running tests passed all.

Fist bump for PyTorch devs.
:facepunch:

I am looking at the same error output. Would you tell me how to solve it?
What is the meaning of ‘checking out tag v.0.2.0’ ?

$ git clone --branch v0.2.0 https://github.com/pytorch/pytorch.git

Thanks @FuriouslyCurious

1 Like