feat: add lora fine tuning for llama 3.2 (#958)

.github/workflows/ci_timing.yaml

@@ -1,4 +1,4 @@
-# This workflow uses GitHub CLI to get timings of last 50 runs of Concrete-ML main CI 
+# This workflow uses GitHub CLI to get timings of last 50 runs of Concrete ML main CI 
 # and send it to slack and add it as an artifact on the workflow
 name: CML build time
 on:

.github/workflows/refresh-one-notebook.yaml

@@ -28,6 +28,7 @@ on:
         - KNearestNeighbors \n
         - LinearRegression \n
         - LinearSVR \n
+        - LLamaFineTuning \n
         - LogisticRegression \n
         - LogisticRegressionTraining \n
         - LoraMLP \n
@@ -76,6 +77,7 @@ env:
   KNearestNeighbors: "docs/advanced_examples/KNearestNeighbors.ipynb" 
   LinearRegression: "docs/advanced_examples/LinearRegression.ipynb" 
   LinearSVR: "docs/advanced_examples/LinearSVR.ipynb" 
+  LLamaFineTuning: "use_case_examples/lora_finetuning/LLamaFineTuning.ipynb" 
   LogisticRegression: "docs/advanced_examples/LogisticRegression.ipynb" 
   LogisticRegressionTraining: "docs/advanced_examples/LogisticRegressionTraining.ipynb" 
   LoraMLP: "docs/advanced_examples/LoraMLP.ipynb" 

.github/workflows/release.yaml

@@ -288,7 +288,7 @@ jobs:
           tags: true
 
   # This action creates docker and pypi images directly on the AWS EC2 instance
-  # The 'PRIVATE_RELEASE_IMAGE_BASE' variable is kept here in case Concrete-ML starts to publish
+  # The 'PRIVATE_RELEASE_IMAGE_BASE' variable is kept here in case Concrete ML starts to publish
   # private nightly releases one day. Currently, release candidates and actual releases are all
   # done through the 'PUBLIC_RELEASE_IMAGE_BASE' image. The private image is also used to list all
   # tags easily
@@ -471,7 +471,7 @@ jobs:
           echo "" >> "${SECRETS_FILE}"
           echo "SECRETS_FILE=${SECRETS_FILE}" >> "$GITHUB_ENV"
 
-      - name: Build Docker Concrete-ML Image
+      - name: Build Docker Concrete ML Image
         if: ${{ success() && !cancelled() }}
         uses: docker/build-push-action@48aba3b46d1b1fec4febb7c5d0c644b249a11355
         with:

docs/advanced_examples/DecisionTreeRegressor.ipynb

@@ -6,9 +6,9 @@
    "id": "5755bc04",
    "metadata": {},
    "source": [
-    "# Decision Tree Regression Using Concrete-ML\n",
+    "# Decision Tree Regression Using Concrete ML\n",
     "\n",
-    "In this tutorial, we show how to create, train and evaluate a decision tree regression model using Concrete-ML library.\n",
+    "In this tutorial, we show how to create, train and evaluate a decision tree regression model using Concrete ML library.\n",
     "\n"
    ]
   },
@@ -18,16 +18,16 @@
    "id": "2c256087-c16a-4249-9c90-3f4863938385",
    "metadata": {},
    "source": [
-    "### Introducing Concrete-ML\n",
+    "### Introducing Concrete ML\n",
     "\n",
-    "> Concrete-ML is an open-source, privacy-preserving, machine learning inference framework based on fully homomorphic encryption (FHE).\n",
+    "> Concrete ML is an open-source, privacy-preserving, machine learning inference framework based on fully homomorphic encryption (FHE).\n",
     "> It enables data scientists without any prior knowledge of cryptography to automatically turn machine learning models into their FHE equivalent,using familiar APIs from Scikit-learn and PyTorch.\n",
     "> <cite>&mdash; [Zama documentation](../README.md)</cite>\n",
     "\n",
     "This tutorial does not require a deep understanding of the technology behind concrete-ML.\n",
     "Nonetheless, newcomers might be interested in reading introductory sections of the official documentation such as:\n",
     "\n",
-    "- [What is Concrete-ML](../README.md)\n",
+    "- [What is Concrete ML](../README.md)\n",
     "- [Key Concepts](../getting-started/concepts.md)\n",
     "\n",
     "In the tutorial, we will be using the following terminology:\n",
@@ -233,10 +233,10 @@
    "source": [
     "## Training A Decision Tree\n",
     "\n",
-    "ConcreteDecisionTreeRegressor is the Concrete-ML equivalent of scikit-learn's DecisionTreeRegressor.\n",
+    "ConcreteDecisionTreeRegressor is the Concrete ML equivalent of scikit-learn's DecisionTreeRegressor.\n",
     "It supports the same parameters and a similar interface, with the extra capability of predicting directly on ciphertext without the need to decipher it, thus preservacy privacy.\n",
     "\n",
-    "Currently, Concrete-ML models must be trained on plaintext. To see how it works, we train a DecisionTreeRegressor with default parameters and estimate its accuracy on test data. Note here that predictions are done on plaintext too, but soon, we will predict on ciphertext."
+    "Currently, Concrete ML models must be trained on plaintext. To see how it works, we train a DecisionTreeRegressor with default parameters and estimate its accuracy on test data. Note here that predictions are done on plaintext too, but soon, we will predict on ciphertext."
    ]
   },
   {
@@ -479,7 +479,7 @@
    "source": [
     "## Predicting on Ciphertext\n",
     "If the predictions are similar although slightly less accurate, the real advantage of ConcreteML is privacy.\n",
-    "We now show how we can perform prediction on ciphertext with Concrete-ML, so that the model does not need to decipher the data at all to compute its estimate."
+    "We now show how we can perform prediction on ciphertext with Concrete ML, so that the model does not need to decipher the data at all to compute its estimate."
    ]
   },
   {
@@ -798,7 +798,7 @@
     "Once the model is carefully trained and quantized, it is ready to be deployed and used in production. Here are some useful links on the subject:\n",
     "    \n",
     "  - [Inference in the Cloud](../getting-started/cloud.md) summarize the steps for cloud deployment\n",
-    "  - [Production Deployment](../guides/client_server.md) offers a high-level view of how to deploy a Concrete-ML model in a client/server setting.\n",
+    "  - [Production Deployment](../guides/client_server.md) offers a high-level view of how to deploy a Concrete ML model in a client/server setting.\n",
     "  - [Client Server in Concrete ML](./ClientServer.ipynb) provides a more hands-on approach as another tutorial."
    ]
   }

docs/advanced_examples/LinearSVR.ipynb

@@ -88,7 +88,7 @@
     "\n",
     "\n",
     "def get_concrete_plot_config(mse_score=None):\n",
-    "    label = \"Concrete-ML\"\n",
+    "    label = \"Concrete ML\"\n",
     "    if mse_score is not None:\n",
     "        label += f\", {'$MSE$'}={mse_score:.4f}\"\n",
     "    return {\"c\": \"orange\", \"linewidth\": 2.5, \"label\": label}"
@@ -646,7 +646,7 @@
     "y_pred_sklearn = sklearn_rgs.predict(X_test)\n",
     "print(f\"Execution time: {(time.time() - time_begin) / len(X_test):.4f} seconds per sample\")\n",
     "\n",
-    "# Now predict using clear quantized Concrete-ML model on testing set\n",
+    "# Now predict using clear quantized Concrete ML model on testing set\n",
     "time_begin = time.time()\n",
     "y_preds_quantized = concrete_rgs.predict(X_test)\n",
     "print(f\"Execution time: {(time.time() - time_begin) / len(X_test):.4f} seconds per sample\")"