Upload 18 files

.gitattributes

@@ -30,3 +30,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 deployment_logit/client_dir/3427324229_encrypted_output filter=lfs diff=lfs merge=lfs -text
+deployment_logit/server_dir/3427324229_encrypted_output filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -1,3 +1,4 @@
+.fhe/
 .fhe_keys/
 client_dir/
 server_dir/

README.md

@@ -1,13 +1,7 @@
----
-title: Titanic Survival Prediction using Fully Homomorphic Encryption (FHE)
-sdk: gradio
-colorFrom: blue
----
-
-Titanic Survival Prediction using Fully Homomorphic Encryption (FHE)
+Titanic Survival Prediction using Fully Homomorphic Encryption (FHE) and Cloud Deployment of Encrypted ML
 
 Titanic survival code adapted from:
 https://huggingface.co/spaces/gradio/titanic_survival_main
 
 FHE Cloud Deployment code adapted from:
-https://github.com/zama-ai/concrete-ml/blob/release/1.1.x/use_case_examples/deployment/README.md
+https://github.com/zama-ai/concrete-ml/blob/release/1.1.x/use_case_examples/deployment/README.md

app.py

@@ -166,9 +166,9 @@ def concrete_predict_survival(input_dict):
     pred = concrete_clf.predict_proba(df)[0]
     return {"Perishes": float(pred[0]), "Survives": float(pred[1])}
 
-# print("\nclear_test    ", clear_predict_survival({'Pclass': [1], 'Sex': [0], 'Age': [25], 'Fare': [20.0], 'Embarked': [2], 'Company': [1]}))
+print("\nclear_test    ", clear_predict_survival({'Pclass': [1], 'Sex': [0], 'Age': [25], 'Fare': [20.0], 'Embarked': [2], 'Company': [1]}))
 
-# print("encrypted_test", concrete_predict_survival({'Pclass': [1], 'Sex': [0], 'Age': [25], 'Fare': [20.0], 'Embarked': [2], 'Company': [1]}),"\n")
+print("encrypted_test", concrete_predict_survival({'Pclass': [1], 'Sex': [0], 'Age': [25], 'Fare': [20.0], 'Embarked': [2], 'Company': [1]}),"\n")
 
 
 def key_gen_fn() -> Dict:
@@ -362,7 +362,7 @@ def run_fhe_fn(user_id: str) -> Dict:
                 fhe_execution_time_box: gr.update(visible=False),
             }
         else:
-            time.sleep(1)
+            time.sleep(10)
             print(f"response.ok: {response.ok}, {response.json()} - Computed")
 
     return {
@@ -661,4 +661,4 @@ with gr.Blocks() as demo:
 
     # ------------------------- End -------------------------
 
-demo.launch(share=True)
+demo.launch()

deployment_logit/client.zip

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e1f806d1aa6f67cafd07321486215d4e5822279fb9fe14a401e9a77f4d44093b
-size 6147
+oid sha256:058a33d9d78957b853112809fc202cd1ee546c9beca2550ed95f140569833657
+size 6437

deployment_logit/client_dir/3427324229_encrypted_output

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7daf64ac339601799c4db40703bd2c1b9d699288f06814ddb3a2f391eb20d934
+size 3932690

deployment_logit/server.zip

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1b6ab82039055e3b69889be441918546b1609010f69096c88ed5b1c12c18f7ac
-size 9629
+oid sha256:54c45f9b27bd5493a2e58e3ad9fa8365490bd65abefbe0befb24b1f1a20c433c
+size 14985

deployment_logit/server_dir/3427324229_encrypted_output

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7daf64ac339601799c4db40703bd2c1b9d699288f06814ddb3a2f391eb20d934
+size 3932690

deployment_logit/versions.json

@@ -1 +1 @@
-{"concrete-ml": "1.4.0", "concrete-python": "2.5", "python": "3.10.12"}
+{"concrete-python": "2.5.1", "concrete-ml": "1.4.1", "python": "3.10.12"}

requirements.txt

@@ -1,7 +1,7 @@
 scikit-learn
 numpy
 pandas
-concrete-ml==1.4.0
-gradio==3.35.2
-uvicorn>=0.21.0
-fastapi>=0.93.0
+concrete-ml
+gradio
+uvicorn
+fastapi

server.py

@@ -94,7 +94,7 @@ def get_output(user_id: str = Form()):
     with encrypted_output_path.open("rb") as f:
         encrypted_output = f.read()
 
-    time.sleep(1)
+    time.sleep(10)
 
     # Send the encrypted output
     return Response(encrypted_output)

train.py

@@ -0,0 +1,120 @@
+import os
+import time
+import numpy
+import pandas as pd
+import torch
+
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.model_selection import train_test_split
+
+from sklearn.metrics import average_precision_score
+from sklearn.model_selection import GridSearchCV, train_test_split
+
+from concrete.ml.sklearn import RandomForestClassifier as ConcreteRandomForestClassifier
+
+def train(dev_folder="./dev"):
+    # Download the data-sets
+    if not os.path.isfile("./files/titanic.csv"):
+        raise ValueError(
+            "no dataset"
+        )
+
+    current_dir = os.path.dirname(os.path.realpath(__file__))
+    data = pd.read_csv(os.path.join(current_dir, "files/titanic.csv"))
+
+    def encode_age(df):
+        df.Age = df.Age.fillna(-0.5)
+        bins = (-1, 0, 5, 12, 18, 25, 35, 60, 120)
+        categories = pd.cut(df.Age, bins, labels=False)
+        df.Age = categories
+        return df
+
+
+    def encode_fare(df):
+        df.Fare = df.Fare.fillna(-0.5)
+        bins = (-1, 0, 8, 15, 31, 1000)
+        categories = pd.cut(df.Fare, bins, labels=False)
+        df.Fare = categories
+        return df
+
+
+    def encode_df(df):
+        df = encode_age(df)
+        df = encode_fare(df)
+        sex_mapping = {"male": 0, "female": 1}
+        df = df.replace({"Sex": sex_mapping})
+        embark_mapping = {"S": 1, "C": 2, "Q": 3}
+        df = df.replace({"Embarked": embark_mapping})
+        df.Embarked = df.Embarked.fillna(0)
+        df["Company"] = 0
+        df.loc[(df["SibSp"] > 0), "Company"] = 1
+        df.loc[(df["Parch"] > 0), "Company"] = 2
+        df.loc[(df["SibSp"] > 0) & (df["Parch"] > 0), "Company"] = 3
+        df = df[
+            [
+                "PassengerId",
+                "Pclass",
+                "Sex",
+                "Age",
+                "Fare",
+                "Embarked",
+                "Company",
+                "Survived",
+            ]
+        ]
+        return df
+
+    train = encode_df(data)
+
+    X_all = train.drop(["Survived", "PassengerId"], axis=1)
+    y_all = train["Survived"]
+
+    num_test = 0.20
+    X_train, X_test, y_train, y_test = train_test_split(
+        X_all, y_all, test_size=num_test, random_state=23
+    )
+
+    # n_estimators = 50
+    # max_depth = 4
+    # n_bits = 6
+    # n_jobs_xgb = 1
+    # n_jobs_gridsearch = -1
+
+    # A gridsearch to find the best parameters
+    parameters = {
+        "n_bits": [6],
+        "max_depth": [4, 8],
+        "n_estimators": [30, 50],
+        "n_jobs": [-1],
+    }
+ 
+    concrete_clf = ConcreteRandomForestClassifier()
+    # concrete_clf.fit(X_train, y_train)
+    # concrete_predictions = concrete_clf.predict(X_test)
+
+    grid_search = GridSearchCV(concrete_clf, parameters, cv=3, n_jobs=-1, scoring="accuracy")
+    grid_search.fit(X_train, y_train)
+
+    # Check the accuracy of the best model
+    print(f"Best score: {grid_search.best_score_}")
+
+    # Check best hyper-parameters
+    print(f"Best parameters: {grid_search.best_params_}")
+
+    # Extract best model
+    best_model = grid_search.best_estimator_
+    assert isinstance(best_model, ConcreteRandomForestClassifier)   
+
+    best_model.compile(X_train)
+
+    # Export the final model such that we can reuse it in a client/server environment
+
+    # Save the model to be pushed to a server later
+    from concrete.ml.deployment import FHEModelDev
+
+    fhe_api = FHEModelDev(dev_folder, best_model)
+    fhe_api.save()
+
+
+if __name__ == "__main__":
+    train()