Closed aimlnerd closed 2 years ago
I am a first-time user of zenml, who also encountered a similar issue. (Perhaps more related to M1 chipset in particular)
What I did?
I just created a brand new virtual environment, and run pip install zenml
, and pip failed complaining dependency conflicts.
Environment: Python 3.8, Apple M1 CPU
ERROR: Cannot install zenml==0.1.0, zenml==0.1.1, zenml==0.1.2, zenml==0.1.3, zenml==0.1.4, zenml==0.1.5, zenml==0.2.0, zenml==0.3.1, zenml==0.3.2, zenml==0.3.3, zenml==0.3.4, zenml==0.3.5, zenml==0.3.6, zenml==0.3.6.1, zenml==0.3.7 and zenml==0.3.8 because these package versions have conflicting dependencies.
The conflict is caused by:
zenml 0.3.8 depends on tensorflow==2.4.1
zenml 0.3.7 depends on tensorflow<2.4.0 and >=2.3.0
zenml 0.3.6.1 depends on tensorflow<2.4.0 and >=2.3.0
zenml 0.3.6 depends on tensorflow<2.4.0 and >=2.3.0
zenml 0.3.5 depends on tensorflow<2.4.0 and >=2.3.0
zenml 0.3.4 depends on tensorflow<2.4.0 and >=2.3.0
zenml 0.3.3 depends on tensorflow<2.4.0 and >=2.3.0
zenml 0.3.2 depends on tensorflow<2.4.0 and >=2.3.0
zenml 0.3.1 depends on tensorflow<2.4.0 and >=2.3.0
zenml 0.2.0 depends on tensorflow<2.4.0 and >=2.3.0
zenml 0.1.5 depends on tensorflow<2.4.0 and >=2.3.0
zenml 0.1.4 depends on tensorflow<2.4.0 and >=2.3.0
zenml 0.1.3 depends on tensorflow<2.4.0 and >=2.3.0
zenml 0.1.2 depends on tensorflow<2.4.0 and >=2.3.0
zenml 0.1.1 depends on tensorflow==2.3.0
zenml 0.1.0 depends on tensorflow==2.3.0
To fix this you could try to:
1. loosen the range of package versions you've specified
2. remove package versions to allow pip attempt to solve the dependency conflict
ERROR: ResolutionImpossible: for help visit https://pip.pypa.io/en/latest/topics/dependency-resolution/#dealing-with-dependency-conflicts
(zenml) ➜ zenml-playground pip list
Package Version
---------- -------
pip 22.0.4
setuptools 62.1.0
wheel 0.37.1
@deepakiim Your issue is perhaps related to pydantic and spacy conflicting wth some core dependencies of ZenML. One way to figure this out is to let us know what are the minimum possible dependencies in your environment that you need. After that we can try to loosen some of our internal dependencies to help out.
@AsiaCao Your issue is a bit strange. We are at Version 0.7.2 and your pip seems not to find any version above 0.3.8. I am wondering how that is possible. You said you are on a M1 chipset and that is for sure not supported yet by ZenML (We internally are trying to solve this but it makes honestly it is hard as even other bigger packages like Tensorflow dont even have support yet)
Contact Details [Optional]
No response
What happened?
This is not a bug. The production project's model microservice requirements.txt contains the below packages
I am unable to add zenml 0.7.1 to the above requirements.txt Because many of the existing package version conflicts with zenml. The main ones that conflict are spacy, pydantic, jinja etc.
In a simple example notebook, zenml works well but not able to add it to production due to this dependencies. When i install zenml==0.7.1 after running pip install -r requirements.txt Its uninstalling and installing new versions of the package.
I believe this is causing unexpected errors like below
Really liked the mlops features and ml pipeline features of zenml. Would be a disappointment to remove zenml related codes and go back to old way of coding without zenml pipelines for the sake of putting the model in production. Any solutions? Perhaps there is a better way to manage the dependencies.
Reproduction steps
1. 2. 3. ...
ZenML Version
0.7.1
Python Version
3.8
OS Type
No response
Relevant log output
No response
Code of Conduct