Wrong values of a Beta distribution sampling if the last dimension of the shape is 1 #10118
Description
Issue description
Errors occur when sampling from a Beta distribution if the last dimension of the shape is 1.
Code example
Sampling with shape [64, 2] returns correct values an a correct shape:
>>> torch.distributions.Beta(2,2).sample(torch.Size([64,2]))
tensor([[0.5614, 0.6955],
[0.7505, 0.9468],
[0.3592, 0.3131],
[0.6340, 0.2711],
[0.5731, 0.7529],
[0.7051, 0.4502], ...
Sampling with shape [64, 1] returns wrong values (always 0.5 and some NaNs) an a correct shape:
>>> torch.distributions.beta.Beta(2,2).sample(torch.Size([64,1]))
tensor([[0.5000],
[0.5000],
[0.5000],
[0.5000],
[ nan],
[0.5000],
[0.5000], ...
System Info
PyTorch version: 0.4.1
Is debug build: No
CUDA used to build PyTorch: None
OS: Ubuntu 16.04.4 LTS
GCC version: (Ubuntu 5.4.0-6ubuntu1~16.04.10) 5.4.0 20160609
CMake version: Could not collect
Python version: 3.6
Is CUDA available: No
CUDA runtime version: No CUDA
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA
Versions of relevant libraries:
[pip] Could not collect
[conda] pytorch-cpu 0.4.1 py36_cpu_1 pytorch
[conda] torchvision-cpu 0.2.1 py36_1 pytorch