installation error about concrete-ml HOT 7 CLOSED

pip freeze | grep concrete
concrete-compiler==0.10.0
concrete-ml==0.3.0
concrete-numpy==0.7.0

from concrete-ml.

Hello @sbzgyxzm ,
Are you on a linux machine ? If so, could you run ldd --version and share me the output ?

Else, could you tell me which OS and architecture you are using ?

Thanks !

from concrete-ml.

Distributor ID: Ubuntu
Description: Ubuntu 18.04.6 LTS
Release: 18.04

ldd --version
ldd (GNU libc) 2.28
python :3.8.19

Hello @sbzgyxzm , Are you on a linux machine ? If so, could you run ldd --version and share me the output ?

Else, could you tell me which OS and architecture you are using ?

Thanks !

from concrete-ml.

I try to install concrete-python manually, but it is not supported. (In vmware)
ERROR: concrete_python-2.6.0-cp38-cp38-manylinux_2_28_x86_64.whl is not a supported wheel on this platform.

from concrete-ml.

So there might be an issue with the fact that you have glibc 2.28 (we don't support any lower versions) and Ubuntu 18 (which we don't test as it starts to be a bit old). I would suggest you to try one of the following :

• use docker
• update to ubuntu 20

Tell me how it goes !

from concrete-ml.

now i install it in ubuntu20, python3.10, glibc2.31. I'm wondering if i could train an encrypted model using encrypted mnist dataset.

from concrete-ml.

Hello again @sbzgyxzm , great to know that it works on Ubuntu 20 !

If I understand your question well, you are looking at a way of doing encrypted training (meaning the data-set is encrypted) using a model for the mnist data-set, am I right ?

If so, here's the documentation section for this feature. For now, we've only given an example for training an encrypted logistic regression model so you want might to check if this could work for you. Note that we are planning on going further and add new examples in the future.

If you are only interested in encrypted inference, here's an example we have with a CNN on MNIST.

Besides, if you have any other questions, please feel free to ask them in a new issue !

Hope that helps 😉

from concrete-ml.