Installation Error - Failed building wheel for tokenizers

[victorlongo]

🐛 Bug

Information

Model I am using (Bert, XLNet ...): N/A

Language I am using the model on (English, Chinese ...): N/A

The problem arises when using:

• the official example scripts: (give details below)

Problem arises in transformers installation on Microsoft Windows 10 Pro, version 10.0.17763

After creating and activating the virtual environment, installing transformers is not possible, because the following error occurs:

"error: can not find Rust Compiler"
"ERROR: Failed building wheel for tokenizers"
Failed to build tokenizers
ERROR: Could not build wheels for tokenizers which use PEP 517 and cannot be installed d

The tasks I am working on is:
[X ] transformers installation

To reproduce

Steps to reproduce the behavior:

1. From command line interface, create and activate a virtual environment by following the steps in this URL: https://packaging.python.org/guides/installing-using-pip-and-virtual-environments/
2. Install transformers from source, by following the example in the topic From Source on this URL: https://github.com/huggingface/transformers

-m pip --version
-m pip install --upgrade pip
-m pip install --user virtualenv
-m venv env
.\env\Scripts\activate
pip install transformers

ERROR: Command errored out with exit status 1:
   command: 'c:\users\vbrandao\env\scripts\python.exe' 'c:\users\vbrandao\env\lib\site-packages\pip\_vendor\pep517\_in_process.py' build_wheel 'C:\Users\vbrandao\AppData\Local\Temp\tmpj6evjmze'
       cwd: C:\Users\vbrandao\AppData\Local\Temp\pip-install-sza2_lmj\tokenizers
  Complete output (10 lines):
  running bdist_wheel
  running build
  running build_py
  creating build
  creating build\lib
  creating build\lib\tokenizers
  copying tokenizers\__init__.py -> build\lib\tokenizers
  running build_ext
  running build_rust
  error: Can not find Rust compiler
  ----------------------------------------
  ERROR: Failed building wheel for tokenizers
Failed to build tokenizers
ERROR: Could not build wheels for tokenizers which use PEP 517 and cannot be installed directly

Expected behavior

Installation of transformers should be complete.

Environment info

• transformers version: N/A - installation step
• Platform: Command Line Interface / Virtual Env
• Python version: python 3.8
• PyTorch version (GPU?): N/A
• Tensorflow version (GPU?): N/A
• Using GPU in script?: N/A
• Using distributed or parallel set-up in script?: N/A
  

[JohnGiorgi]

Having the exact same issue on a Linux machine!

[GDBSD]

Environment: macOS Mojave Ver 10.14.6
Tried installing both from pip and source. Same issue:

Successfully built transformers
Failed to build tokenizers

Result was that Transformers was not installed (not listed in pip freeze)

This however should work - seems like you just won't get the the new tokenizers:
pip install transformers==2.4.1

[h21k]

@GDBSD I had the same issue on the same OS version and also tried pip and source. Your version specification worked.

[youssefavx]

Had the same issue on MacOS Mojave when doing pip3 install. Tried pip2 install, it worked but I got another error when running my script telling me I should really be using python 3.

I tried @GDBSD 's answer, but I got this error:

ERROR: Exception:
Traceback (most recent call last):
  File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/py_compile.py", line 143, in compile
    _optimize=optimize)
  File "<frozen importlib._bootstrap_external>", line 791, in source_to_code
  File "<frozen importlib._bootstrap>", line 219, in _call_with_frames_removed
  File "/private/var/folders/g0/5zwy4mtx7579v5x6rxqb083r0000gn/T/pip-unpacked-wheel-k410h9s0/sacremoses/sent_tokenize.py", line 69
    if re.search(IS_EOS, token)
                              ^
SyntaxError: invalid syntax

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/compileall.py", line 159, in compile_file
    invalidation_mode=invalidation_mode)
  File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/py_compile.py", line 147, in compile
    raise py_exc
py_compile.PyCompileError:   File "/private/var/folders/g0/5zwy4mtx7579v5x6rxqb083r0000gn/T/pip-unpacked-wheel-k410h9s0/sacremoses/sent_tokenize.py", line 69
    if re.search(IS_EOS, token)
                              ^
SyntaxError: invalid syntax


During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/pip/_internal/cli/base_command.py", line 186, in _main
    status = self.run(options, args)
  File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/pip/_internal/commands/install.py", line 404, in run
    use_user_site=options.use_user_site,
  File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/pip/_internal/req/__init__.py", line 71, in install_given_reqs
    **kwargs
  File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/pip/_internal/req/req_install.py", line 815, in install
    warn_script_location=warn_script_location,
  File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/pip/_internal/operations/install/wheel.py", line 614, in install_wheel
    warn_script_location=warn_script_location,
  File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/pip/_internal/operations/install/wheel.py", line 338, in install_unpacked_wheel
    compileall.compile_dir(source, force=True, quiet=True)
  File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/compileall.py", line 97, in compile_dir
    legacy, optimize, invalidation_mode):
  File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/compileall.py", line 169, in compile_file
    msg = err.msg.encode(sys.stdout.encoding,
  File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/pip/_internal/utils/misc.py", line 554, in encoding
    return self.orig_stream.encoding
  File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/codecs.py", line 409, in __getattr__
    return getattr(self.stream, name)
AttributeError: '_io.BufferedWriter' object has no attribute 'encoding'

[tayeb83]

yes I had the same issue with pip3.6 install

[BramVanroy]

Can you all run python transformers-cli env and post the output here? It provides some useful information about your platform that might be helpful to debug.

[gardnerds]

Hi, I had the same problem and resolved it by installing rust.
"error: Can not find Rust compiler"

For MacOS, I used "curl https://sh.rustup.rs -sSf | sh". I also found that it needed a nightly version of rust, so you have to specify that in the install options.

[Wild3d]

Hi, I also had the same problem with my initial installation of the library. After some time, I realized that my anaconda version was on 32Bit. You can check your version with
python -c "import struct;print( 8 * struct.calcsize('P'))"
The output should be 64.
If it is 32 then you have to reinstall your IDE

[h21k]

@Wild3d I can confirm after running your snippet that I am on a 64bit version

[h21k]

@gardnerds after creating a new environment to try your solution that also worked for me. I didn't have rust installed before. It successfully built the wheel for tokenizers (PEP 517).

[cremebrule]

@gardnerds also worked for me. Using python 3.7 and built from source using a clean conda env

[huaiyukhaw]

Install Python 64-bit instead of 32-bit solved my same issue.