XGBoost classifier example

examples/xgboost_classfication/README.md

@@ -0,0 +1,46 @@
+# XGBoost Classifier for Iris dataset
+
+This example shows how to serve an XGBoost classifier model using TorchServe.
+Here we train a model to classify iris dataset
+
+## Pre-requisites
+
+Train an XGBoost classifier model for Iris dataset
+
+```
+pip install -r requirements.txt
+python xgboost_train.py
+```
+
+results in
+
+```
+Model accuracy is 1.0
+Saving trained model to iris_model.json
+```
+
+## Create model archive
+
+```
+mkdir model_store
+torch-model-archiver --model-name xgb_iris --version 1.0 --serialized-file iris_model.json --handler xgboost_iris_handler.py --export-path model_store --extra-files index_to_name.json --config-file model-config.yaml -f
+```
+
+## Start TorchServe
+
+```
+torchserve --start --ncs --model-store model_store --models xgb_iris=xgb_iris.mar
+```
+
+## Inference request
+
+We send a batch of 2 requests
+```
+curl -X POST http://127.0.0.1:8080/predictions/xgb_iris -T sample_input_2.txt & curl -X POST http://127.0.0.1:8080/predictions/xgb_iris -T sample_input_1.txt
+```
+
+results in
+
+```
+versicolor setosa
+```

examples/xgboost_classfication/index_to_name.json

@@ -0,0 +1 @@
+{"0": "setosa", "1": "versicolor", "2": "virginica"}

examples/xgboost_classfication/model-config.yaml

@@ -0,0 +1,3 @@
+minWorkers: 1
+maxWorkers: 1
+batchSize: 2

examples/xgboost_classfication/requirements.txt

@@ -0,0 +1,2 @@
+scikit-learn
+xgboost

examples/xgboost_classfication/sample_input_1.txt

@@ -0,0 +1 @@
+6.1,2.8,4.7,1.2

examples/xgboost_classfication/sample_input_2.txt

@@ -0,0 +1 @@
+5.7,3.8,1.7,0.3

examples/xgboost_classfication/xgboost_iris_handler.py

@@ -0,0 +1,54 @@
+import logging
+import os
+
+import numpy as np
+import torch
+from xgboost import XGBClassifier
+
+from ts.torch_handler.base_handler import BaseHandler
+from ts.utils.util import load_label_mapping
+
+logger = logging.getLogger(__name__)
+
+
+class XGBIrisHandler(BaseHandler):
+    def __init__(self):
+        super().__init__()
+
+    def initialize(self, context):
+        # Set device type
+        self.device = torch.device("cpu")
+
+        # Load the model
+        properties = context.system_properties
+        self.manifest = context.manifest
+        model_dir = properties.get("model_dir")
+        self.model = XGBClassifier()
+        if "serializedFile" in self.manifest["model"]:
+            serialized_file = self.manifest["model"]["serializedFile"]
+            model_weights = os.path.join(model_dir, serialized_file)
+            self.model.load_model(model_weights)
+
+        mapping_file_path = os.path.join(model_dir, "index_to_name.json")
+        self.mapping = load_label_mapping(mapping_file_path)
+
+        logger.info(
+            f"XGBoost Classifier for iris dataset with weights {model_weights} loaded successfully"
+        )
+        self.initialized = True
+
+    def preprocess(self, requests):
+        inputs = []
+        for row in requests:
+            input = row.get("data") or row.get("body")
+            if isinstance(input, (bytes, bytearray)):
+                input = [float(value) for value in input.decode("utf-8").split(",")]
+            inputs.append(input)
+        return np.array(inputs)
+
+    def inference(self, data):
+        return self.model.predict(data)
+
+    def postprocess(self, result):
+        output = [self.mapping[str(res)] for res in result.tolist()]
+        return output

examples/xgboost_classfication/xgboost_train.py

@@ -0,0 +1,29 @@
+# Importing dataset from sklearn
+from sklearn import datasets, metrics
+
+iris = datasets.load_iris()  # dataset loading
+X = iris.data  # Features stored in X
+y = iris.target  # Class variable
+
+# Splitting dataset into Training (80%) and testing data (20%) using train_test_split
+from sklearn.model_selection import train_test_split
+
+X_train, X_test, y_train, y_test = train_test_split(
+    X, y, test_size=0.2, random_state=42
+)
+
+# Create an XGB classifier and instance of the same
+from xgboost import XGBClassifier
+
+clf = XGBClassifier()
+
+clf.fit(X_train, y_train)
+
+y_pred = clf.predict(X_test)
+# classification accuracy
+from sklearn import metrics
+
+print(f"Model accuracy is {metrics.accuracy_score(y_test, y_pred)}")
+saved_model = "iris_model.json"
+print(f"Saving trained model to {saved_model}")
+clf.save_model(saved_model)

ts_scripts/spellcheck_conf/wordlist.txt

@@ -1218,6 +1218,7 @@ venv
 TorchInductor
 Pytests
 deviceType
+XGBoost
 Clamd
 Fickling
 TorchServer