After #1327, importing `torchtext` fails with error `Undefined symbol: _ZTVN5torch3jit6MethodE`

[imaginary-person]

🐛 Bug

After the latest commit (#1327), builds succeed with Python 3.8.5 on Ubuntu 18.04, but importing torchtext fails with error Undefined symbol: _ZTVN5torch3jit6MethodE, if torch has also been built locally from the master branch.

If the HEAD of the cloned repo is changed to the commit before #1327, then this issue doesn't reproduce.

Traceback:

import torchtext
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/external/text/torchtext/__init__.py", line 5, in <module>
    from . import vocab
  File "/external/text/torchtext/vocab/__init__.py", line 11, in <module>
    from .vocab_factory import (
  File "/external/text/torchtext/vocab/vocab_factory.py", line 4, in <module>
    from torchtext._torchtext import (
ImportError: /external/text/torchtext/_torchtext.so: undefined symbol: _ZTVN5torch3jit6MethodE

Environment

PyTorch version: 1.10.0a0+git6a059f0
Is debug build: False
CUDA used to build PyTorch: None
ROCM used to build PyTorch: N/A

OS: Ubuntu 18.04.1 LTS (x86_64)
GCC version: (Ubuntu 9.4.0-1ubuntu1~18.04) 9.4.0
Clang version: Could not collect
CMake version: version 3.10.2
Libc version: glibc-2.27

Python version: 3.8.0 (default, Feb 25 2021, 22:10:10)  [GCC 8.4.0] (64-bit runtime)
Python platform: Linux-4.15.0-147-generic-x86_64-with-glibc2.27
Is CUDA available: False
CUDA runtime version: No CUDA
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA
HIP runtime version: N/A
MIOpen runtime version: N/A

Versions of relevant libraries:
[pip3] bert-pytorch==0.0.1a4
[pip3] numpy==1.21.1
[pip3] torch==1.10.0a0+git6a059f0
[pip3] torch-struct==0.4
[pip3] torchfile==0.1.0
[pip3] torchtext==0.11.0a0+0d7b9e2
[pip3] torchvision==0.10.0
[pip3] torchviz==0.0.2
[conda] Could not collect

[imaginary-person]

Issue was due to a PyTorch version conflict. The default Ubuntu image available already had PyTorch pre-installed