Closed bacalfa closed 2 years ago
Hi, @bacalfa! This is a fair note. Since FEDOT is still an early-stage project, we have tried to have the same environment for all team members. Because it imports most of the popular DS-stack libraries and some of them can conflict with each other. However, I agree that strict requirements are not an option. The team will discover best practices about dependencies in well-known libraries and we will do something with it.
Hi, @bacalfa!
We tried to simplify the dependency requirements in the new version 0.4.0. I hope it will help to resolve your issues. However, some versions are still restricted in some range (e.g. xgboost<1.3.0) due to compatibility issues.
Please take your feedback if some problems remain unresolvable. In this case, we will continue the investigation.
Hi, @bacalfa!
Is the new version helps with your problem?
Thanks! I think it's less strict now. I'm only getting one complaint.
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
fedot 0.4.0 requires xgboost<1.3.0,>=1.0.1, but you have xgboost 1.4.2 which is incompatible.
Ok, a bit later we will try to adapt our code to the newer xgboost version and update the requirements.
Be free to ask any questions if you will try to do it by yourself (with a fork and pull request)!
Any update on this issue? It is almost impossible to install this tool on Linux platforms.
The installation is already simplified in master branch - you can test it for your system. The 0.6.0 release with this feature is also coming.
Thanks! It worked. May seem obvious now but best if you include this [see below] in the installation docs.
!pip install git+https://github.com/nccr-itmo/FEDOT.git
Resolved and will be added to release.
I'm very excited to try out this package, but its strict dependency requirements are giving me headaches when setting up my conda environment.
Is it possible to relax the requirements?