Docker image is not working on Apple M1/M2 chips

[mvacher]

Summary

The version provided through the docker image is not working on Apple M1 and M2 chips

Description

minimal POC to trigger the bug

On a M1 or M2 machine:

# Install
docker pull zamafhe/concrete-ml:latest

# Run the with:
docker run --rm -it -p 8888:8888 -v ~/host_folder:/data zamafhe/concrete-ml

Python code: ( example from the readme)

from sklearn.datasets import make_classification
from sklearn.model_selection import train_test_split
from concrete.ml.sklearn import LogisticRegression


# Lets create a synthetic data-set
x, y = make_classification(n_samples=100, class_sep=2, n_features=30, random_state=42)

# Split the data-set into a train and test set
X_train, X_test, y_train, y_test = train_test_split(
    x, y, test_size=0.2, random_state=42
)
# Now we train in the clear and quantize the weights
model = LogisticRegression(n_bits=8)
model.fit(X_train, y_train)

# We then compile on a representative set
model.compile(X_train) # --> This line kills the kernel and restart the notebook

[RomanBredehoft]

We indeed don't handle docker with M1 / M2 chips currently but we'll try our best to patch it soon (I'm on M2 as well 😉). Thanks for the issue ! In the mean time, please use the native installation as Concrete-ML properly works with Apple Silicon !

[bcm-at-zama]

Hello @mvacher. We're having a look, sorry for the delay. Were you able to use Concrete with native packages?

[bcm-at-zama]

Hello @berhankarabay . Did you try to install natively, ie with

pip install -U pip wheel setuptools
pip install concrete-ml

ideally in a virtual env?

[fd0r]

Hello @berhankarabay , are you using a M1 system or intel based system?
Also what snippet of code gave you this error?
The code from the KDNuggets blog is deprecated as we changed the API when updating from 0.x to 1.x
I would recommend you take a look at either https://docs.zama.ai/concrete-ml or the use-case examples that we have in the repository for more up-to-date examples.
Also which version of python are you using in your conda env? I would recommend trying out with python 3.9
Best regards

[bcm-at-zama]

Yes, you might want to start with the very small example which is given in https://github.com/zama-ai/concrete-ml#a-simple-concrete-ml-example-with-scikit-learn to test your config

[bcm-at-zama]

(I don't see the screenshots)

[bcm-at-zama]

Sorry we don't see attachments, maybe it doesn't work by email? Maybe just copy-paste the text of the stdio, if you can't make it work

[fd0r]

Alright so 3.8 should work okay.
Could you please replace execute_in_fhe=True by fhe="execute" ?

[fd0r]

Which version do you have installed on your system?
Could you please make sure that you are using the latest version 1.3.0 of Concrete ML.

[RomanBredehoft]

Hello @berhankarabay ,
Running pip install concrete-ml==1.3.0 in your terminal should update Concrete ML properly !

Also, we recommend you to use virtual environments (ex: https://docs.python.org/3/library/venv.html) if you haven't done so already, it will prevent you having further dependency conflicts in your projects !

[berhankarabay]

Hello again,

I solved the issue, it turns out that my python architecture was incorrect. I fixed it to executable arm64 and installed 1.3.0. The simple example given in your website works now, thank you.