Cannot install saber (broken library dependencies)

[albinafisher]

1. Installing saber in new conda virtual environment:

pip install git+https://github.com/BaderLab/saber.git

ERROR: thinc 6.10.3 has requirement wrapt<1.11.0,>=1.10.0, but you'll have wrapt 1.11.2 which is incompatible. ERROR: multiprocess 0.70.9 has requirement dill>=0.3.1, but you'll have dill 0.2.9 which is incompatible. ERROR: fastai 1.0.58 has requirement spacy>=2.0.18, but you'll have spacy 2.0.13 which is incompatible. ERROR: en-core-web-sm 2.1.0 has requirement spacy>=2.1.0, but you'll have spacy 2.0.13 which is incompatible.

2. Trying to fix dependency errors:

pip install wrapt==1.10.0

ERROR: tensorflow 1.15.0rc3 has requirement wrapt>=1.11.1, but you'll have wrapt 1.10.0 which is incompatible. ERROR: fastai 1.0.58 has requirement spacy>=2.0.18, but you'll have spacy 2.0.13 which is incompatible. ERROR: en-core-web-sm 2.1.0 has requirement spacy>=2.1.0, but you'll have spacy 2.0.13 which is incompatible.

Also doesn't fix the problem: pip install git+https://www.github.com/keras-team/keras-contrib.git pip install https://github.com/huggingface/neuralcoref-models/releases/download/en_coref_md-3.0.0/e n_coref_md-3.0.0.tar.gz

3. If I try to use saber anyway, this error is always displayed:

TypeError: Tensors in list passed to 'values' of 'ConcatV2' Op have types [bool, float32] that don't all match.

I tried to install saber on colab, Windows and Ubuntu

[JohnGiorgi]

Hi @AlbinaGalimzyanova,

Thank you for pointing this out!

I found that the issue was with the Keras dependency. Pinning it to keras==2.2.4 fixed the issue. Although I still get the warnings you mentioned, I tested that the colab now runs without error.

If you are still having issues, please let me know.

[albinafisher]

@JohnGiorgi Thanks for the quick reply, now everything works :)