mirror of
https://github.com/luxfi/torus-ml.git
synced 2026-10-30 15:15:23 +00:00
zeekay
e037e8f7ea
The distribution called itself torus-ml while shipping a package directory named concrete, so the name a user installs and the name they import disagreed — and most of the tree already said torus.ml, which could not resolve against src/concrete. Moved src/concrete to src/torus and pointed the include at it. The dependency was Zama's concrete-python from their index. It is torus-fhe from ours, which is what luxfhe/torus already ships: that frontend declares torus-fhe and ships a torus/fhe module, so every 'from torus.fhe' in this tree was already naming our compiler and only the dependency lagged. Two distributions share one namespace, which is the shape the upstream this forked from used and the reason the split works: torus-fhe gives torus.fhe, torus-ml gives torus.ml. Swept every other spelling — identifiers, enum members, workflow variables, filenames, notebooks. LICENSE and NOTICE keep the upstream attribution a fork owes; the figures carry the old name in JPEG metadata a text sweep cannot reach.
54 lines
2.1 KiB
Python
54 lines
2.1 KiB
Python
import pickle
|
|
from pathlib import Path
|
|
|
|
import federated_utils
|
|
import numpy
|
|
|
|
from torus.ml.sklearn import LogisticRegression
|
|
|
|
if __name__ == "__main__":
|
|
# Load scikit-learn model
|
|
path_to_model = Path("./model.pkl")
|
|
if not path_to_model.exists() or not path_to_model.is_file():
|
|
raise ValueError(
|
|
"Couldn't find file 'model.pkl'.\n"
|
|
"Please run 'run.sh' to train the model with federated learning."
|
|
)
|
|
with path_to_model.open("rb") as file:
|
|
sklearn_model = pickle.load(file)
|
|
|
|
# Compile model without data since the server doesn't have access to it
|
|
# Indeed in this scenario the users have the data but the server doesn't.
|
|
# We then have to compile the model using random input sampled with the same
|
|
# low and high bounds as the real data, in this context [0, 255].
|
|
number_of_compile_samples = 1000
|
|
compile_set = numpy.random.randint(0, 255, (number_of_compile_samples, 784)).astype(float)
|
|
sklearn_model.classes_ = sklearn_model.classes_.astype(int)
|
|
model = LogisticRegression.from_sklearn_model(sklearn_model, compile_set)
|
|
model.compile(compile_set)
|
|
|
|
# Evaluate the model
|
|
# Load MNIST dataset from https://www.openml.org/d/554
|
|
(_, _), (X_test, y_test) = federated_utils.load_mnist()
|
|
|
|
torus_predictions = model.predict(X_test)
|
|
sklearn_predictions = sklearn_model.predict(X_test)
|
|
simulation_predictions = model.predict(X_test, fhe="simulate")
|
|
|
|
print(
|
|
(torus_predictions == sklearn_predictions).sum() / len(y_test),
|
|
"quantized vs scikit-learn",
|
|
)
|
|
print(
|
|
(torus_predictions == simulation_predictions).sum() / len(y_test),
|
|
"quantized vs simulation",
|
|
)
|
|
print(
|
|
(simulation_predictions == sklearn_predictions).sum() / len(y_test),
|
|
"simulation vs scikit-learn",
|
|
)
|
|
|
|
print((torus_predictions == y_test).sum() / len(y_test), "quantized accuracy")
|
|
print((sklearn_predictions == y_test).sum() / len(y_test), "scikit-learn accuracy")
|
|
print((simulation_predictions == y_test).sum() / len(y_test), "simulation accuracy")
|