I noticed that in the Dockerfile, the version number had not been specified for many packages. This already leads to some ML models not being executable because APIs have changed. For example, with the automatically installed numpy version (for currently 1.24.3), the numpy.float data type no longer exists, making the code in models/vime.py not executable. This bug is easily fixed by simply replacing np.float with float.
Another error occurs, for example, for the ML model STG in models/stg_lib/utils.py. In the collections module used, the collection.Sequence attribute no longer exists, making STG unusable.
In the Dockerfile, the package versions should still be added because more errors will occur sooner or later due to changed interfaces. Also, I have not tested all provided ML models, so further errors can not be excluded.
Thank you for pointing to that, indeed package versions need to be introduced in the docker file. However, due to my workload I'm not able to do that. Therefore, we are welcome open-source contributors!
I noticed that in the Dockerfile, the version number had not been specified for many packages. This already leads to some ML models not being executable because APIs have changed. For example, with the automatically installed numpy version (for currently 1.24.3), the
numpy.float
data type no longer exists, making the code inmodels/vime.py
not executable. This bug is easily fixed by simply replacingnp.float
withfloat
.Another error occurs, for example, for the ML model STG in
models/stg_lib/utils.py
. In the collections module used, the collection.Sequence attribute no longer exists, making STG unusable.In the Dockerfile, the package versions should still be added because more errors will occur sooner or later due to changed interfaces. Also, I have not tested all provided ML models, so further errors can not be excluded.
Many greetings Sedir Mohammed