Fixed: TorchText install issue: The procedure entry point?? could not be located in the dynamic link library

Sometimes, even though you already installed torchtext, but can’t import it, or when you import it, you receive this error:

The procedure entry point ??0Error@c10@@QEAA@V?$basic_string@DU?$char_traits@D@ std@@V?$allocator@D@2@@std@@0PEBX@Z could not be located in the dynamic link library \.conda\envs\time\lib\site-packages\torchtextlib\libtorchtext.pyd.

and then if you try to verify if the installation was successful by using this prompt in anaconda

python -c "import torch; import torchtext; print(torch.__version__, torchtext.__version__)"

you’ll get the same error.

There could be several reasons for this:

Non-compatiable Python enviroment

The first is that the python of the current environment is not compatible with torchtext. At https://pypi.org/project/torchtext/, you can see this table

So, when you create the Python environment where you want to install torchtext, you need to ensure that it is supported by specifying the Python version. For example,

conda create --name textcv python=3.8

Mismatch between PyTorch and torchtext

Another reason could be due to a mismatch between the installed versions of torch (PyTorch) and torchtext. It happens when the libraries are not compatible with each other due to different versions of the underlying C++ libraries they depend on.

To fix this issue, you’ll need to ensure that both torch and torchtext are compatible versions. Here’s what you can do:

1. Uninstall PyTorch and torchtext: First, uninstall the existing versions of torch and torchtext to start with a clean slate:

   pip uninstall torch torchtext

or if you’re using conda:

   conda remove pytorch torchtext

2. Install Compatible Versions: You should install versions of torch and torchtext that are designed to work together. The easiest way to ensure compatibility is to install them both using the same command.

• If you’re using pip, you can install the CPU-only versions of both PyTorch and torchtext like this:
  pip install torch torchvision torchaudio torchtex
  This ensures the latest compatible versions are installed.
  However, for me, it still showed the same bugs. So, I think it’s best to specify a particular version, you can do it like this:

pip install torch==2.0.0 torchvision==0.15.0 torchaudio==2.0.0 torchtext==0.15.0

Of course, these are not the latest editions. So, it’s better to check the table of Version Compatibility at https://pypi.org/project/torchtext/. So, PyTorch 2.2.0 is compatiable with torchtext 0.17.0, which are newer editions of these packages.

2. Verify the Installation: After installation, verify that the versions are correctly installed and working by running the following in Python:

   import torch
   import torchtext
   print(torch.__version__)
   print(torchtext.__version__)

This should show versions that are compatible and without error messages.

Common Version Combinations

Here are some version combinations of torch and torchtext that are known to work well together:

• PyTorch 2.0.0 -> torchtext 0.15.0
• PyTorch 1.13.1 -> torchtext 0.14.1
• PyTorch 1.12.1 -> torchtext 0.13.1