"multinomial_kernel_cuda" not implemented for 'Half'

[nottombrown]

It appears that even after the 1.3.1 fixes, torch.distributions.Categorical no longer works with half dtypes

To Reproduce

pip install -U --pre torch==1.4.0.dev20191104 torchvision -f https://download.pytorch.org/whl/nightly/cu101/torch_nightly.html

import torch


def test_fp16_categorical():
    logits_fp16 = torch.randn(20).cuda().half()

    # These are fine
    torch.argmax(logits_fp16)
    torch.max(logits_fp16)

    # This is also fine
    logits_fp32 = logits_fp16.float()
    sample = torch.distributions.Categorical(logits=logits_fp32).sample()
    print(sample)

    # This fails
    sample = torch.distributions.Categorical(logits=logits_fp16).sample()
    print(sample)


if __name__ == "__main__":
    test_fp16_categorical()

File "/opt/conda/lib/python3.7/site-packages/torch/distributions/categorical.py", line 107, in sample
    sample_2d = torch.multinomial(probs_2d, 1, True)
RuntimeError: "multinomial_kernel_cuda" not implemented for 'Half'

Environment

Collecting environment information...
PyTorch version: 1.4.0.dev20191104
Is debug build: No
CUDA used to build PyTorch: 10.1

OS: Ubuntu 16.04.5 LTS
GCC version: (Ubuntu 5.4.0-6ubuntu1~16.04.11) 5.4.0 20160609
CMake version: version 3.11.1

Python version: 3.7
Is CUDA available: Yes
CUDA runtime version: 10.0.130
GPU models and configuration:
GPU 0: Tesla V100-SXM2-16GB
GPU 1: Tesla V100-SXM2-16GB
GPU 2: Tesla V100-SXM2-16GB
GPU 3: Tesla V100-SXM2-16GB
GPU 4: Tesla V100-SXM2-16GB
GPU 5: Tesla V100-SXM2-16GB
GPU 6: Tesla V100-SXM2-16GB
GPU 7: Tesla V100-SXM2-16GB

Nvidia driver version: 418.87.01
cuDNN version: /usr/lib/x86_64-linux-gnu/libcudnn.so.7.4.2

Additional context

This is not blocking us because we can convert to fp32 before sampling