I have a problem cloning a project from GitHub. The error occurs when I try to install the requirements.txt file:
ERROR: pyzmq has an invalid wheel, multiple .dist-info directories found: libsodium-1.0.17.dist-info, pyzmq-18.1.0.dist-info
I tried the below options but nothing worked. Any idea how can I solve the problem?
pip install –no-cache-dir -r requirements.txt
pip install –no-binary=:all: pyzmq==18.1.0
pip install wheel
I also opened the requirements.txt file and removed this package, but other packages didn’t install correctly!
Advertisement
Answer
EDIT: Solution based on specific repo mentioned by OP:
I was able to install the requirements.txt in this repo using the following steps. Tested on Windows 10 and Python 3.7.
- Download the scipy v1.3.1 wheel file from here. Only Python 3.5-3.7 are supported, so pick the right version (cp35-cp37). Also pick the right OS and 32/64-bits version.
- Run
pip install scipy-1.3.1-cp37-cp37m-win_amd64.whl(for example) from the folder where the .whl file is. - In requirements.txt: change
tensorflow==2.0.0b0totensorflow==2.0.0 - In requirements.txt: change
torch==1.3.0totorch==1.9.0 - Run
pip install -r requirements.txt
Original Answer:
I personally saw the pyzmq has an invalid wheel error while installing tensorflow==2.7.0 with tfx==1.3.3. According to the tfx repo, these versions are not compatible, and result in a dependency conflict. You might have two similar dependencies that are doing the same.
I solved this in 3 ways:
- Specify compatible versions. In my case, switching to
tensorflow==2.6.0worked. - Search for and remove redundant dependencies. In my case,
tensorflowis already included withtfx. Removing it fixed the problem (actually there was another bug, but that’s off-topic). - Add a specific version of
pyzmqto your requirements.txt to force the resolution. I used the latest version (pyzmq==22.3.0), which also fixed the issue.
To help debug the problem, you can view the pip dependency tree using pipdeptree. See this answer.