kserve/docs/samples/v1beta1/torchserve/v2

Predict on a InferenceService using TorchServe

In this example, we use a trained pytorch mnist model to predict handwritten digits by running an inference service with TorchServe predictor.

Setup

1. Your ~/.kube/config should point to a cluster with KServe installed.
2. Your cluster's Istio Ingress gateway must be network accessible.

Create the InferenceService

Deploy the InferenceService with v2 protocol

kubectl apply -f mnist.yaml

For deploying it in gpu

kubectl apply -f mnist_gpu.yaml

Expected Output

$inferenceservice.serving.kserve.io/torchserve-mnist-v2 created

Inference with V2 REST Protocol

The first step is to determine the ingress IP and ports and set INGRESS_HOST and INGRESS_PORT

MODEL_NAME=mnist
SERVICE_HOSTNAME=$(kubectl get inferenceservice torchserve-mnist-v2 -o jsonpath='{.status.url}' | cut -d "/" -f 3)

Use the following image converters to create input request for mnist

Use image converter to create input request for mnist.

1. Tensor input
2. Bytes input

For other models please refer to input request

For tensor input

MODEL_NAME=mnist
curl -v -H "Host: ${SERVICE_HOSTNAME}" http://${INGRESS_HOST}:${INGRESS_PORT}/v2/models/${MODEL_NAME}/infer -d @./tensor_conv/mnist_v2.json

Expected Output

{"id": "d3b15cad-50a2-4eaf-80ce-8b0a428bd298", "model_name": "mnist", "model_version": "1.0", "outputs": [{"name": "predict", "shape": [], "datatype": "INT64", "data": [1]}]}

For bytes input

curl -v -H "Host: ${SERVICE_HOSTNAME}" http://${INGRESS_HOST}:${INGRESS_PORT}/v2/models/${MODEL_NAME}/infer -d @./bytes_conv/mnist_v2_bytes.json

Expected output

{"id": "2266ec1e-f600-40af-97b5-7429b8195a80", "model_name": "mnist", "model_version": "1.0", "outputs": [{"name": "predict", "shape": [], "datatype": "INT64", "data": [1]}]}

Explanation

Model interpretability is an important aspect which help to understand which of the input features were important for a particular classification. Captum is a model interpretability library, the Explain Endpoint uses Captum's state-of-the-art algorithm, including integrated gradients to provide user with an easy way to understand which features are contributing to the model output.

Your can refer to Captum Tutorial for more examples.

Explain Request

curl -v -H "Host: ${SERVICE_HOSTNAME}" http://${INGRESS_HOST}:${INGRESS_PORT}/v2/models/mnist/explain -d @./tensor_conv/mnist_v2.json

Expected Output

{"id": "d3b15cad-50a2-4eaf-80ce-8b0a428bd298", "model_name": "mnist", "model_version": "1.0", "outputs": [{"name": "explain", "shape": [1, 28, 28], "datatype": "FP64", "data": [-0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, 0.0, -0.0, -0.0, 0.0, -0.0, 0.0, -0.0, -0.0, -0.0, -0.0, -0.0, 0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, 0.0, -0.0, 0.0, -0.0, -0.0, -0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.0, -0.0, 0.0, 0.0, -0.0, 0.0, -0.0, -0.0, -0.0, -0.0, -0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, 0.0, -0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, 0.0, -0.0, 0.0, -0.0, -0.0, -0.0, -0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.0, -0.0, 0.0, 0.0, -0.0, -0.0, -0.0, -0.0, -0.0, 0.0, 0.0, -0.0, -0.0, -0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, -0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.0, -0.0040547529196303285, -0.000226128774499257, -0.00012734138382422276, 0.005648369544853077, 0.0089047843954152, 0.002638536593970295, 0.002680245911942565, -0.0026578015819202173, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, 0.0, -0.0, -0.0, -0.0, -0.0, -0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.00024465771891337887, 0.0008218450954311162, 0.01528591767842519, 0.007512832335428859, 0.00709498458333515, 0.0034056686436576803, -0.002091925041823873, -0.0007800293875604465, 0.02299587827540853, 0.019004329367380418, -0.0012529559050418735, -0.0014666116646934577, -0.0, -0.0, -0.0, 0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, -0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.005298396405518712, -0.0007901605729004231, 0.0039060659926479398, 0.023174082126728335, 0.01723791770922474, 0.010867034167828598, 0.003001563229273835, 0.00622421771715703, 0.006120712207087491, 0.01673632965122119, 0.005674718948781803, 0.004344134599735745, -0.0012328422311881568, -0.0, -0.0, -0.0, 0.0, 0.0, -0.0, -0.0, -0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.0, 0.0006867353833785289, 0.009772899792600862, -0.0038754932221901437, 0.001798693579973005, 0.001307544047675232, -0.0024510981010352315, -0.0008806773488194292, -0.0, -0.0, -0.00014277890760828639, -0.009322313235257151, 0.020608317727589167, 0.004351394518148479, -0.0007875566214137449, -0.0009075897508410689, -0.0, -0.0, 0.0, 0.0, 0.0, -0.0, -0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.00022247238084657642, -0.0007829029819622099, 0.0026663695200516055, 0.0009733366691924418, 0.0, -0.0, 0.0, 0.0, 0.0, 0.0, -0.0, 0.0004323207980879993, 0.023657171939959983, 0.01069484496100618, -0.0023759529165659743, -0.0, -0.0, 0.0, 0.0, -0.0, -0.0, -0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.0, -0.002074797197335781, -0.002320101263777886, -0.001289920656543141, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, 0.007629679763806616, 0.01044862710854819, 0.00025032875474040415, -0.0, -0.0, 0.0, 0.0, 0.0, -0.0, -0.0, 0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0003770836745884539, -0.005156369309364184, 0.0012477582083019567, 0.0, 0.0, 0.0, 0.0, 0.0, -0.0, -0.0, 0.0, -0.0, -4.442513564501309e-05, 0.010248046436803096, 0.0009971133914441863, -0.0, -0.0, 0.0, 0.0, 0.0, -0.0, -0.0, 0.0, -0.0, 0.0, 0.0, -0.0, 0.0004501048922351147, -0.00196305355861066, -0.0006664792277975681, 0.0020157403871024866, 0.0, 0.0, -0.0, -0.0, -0.0, -0.0, -0.0, 0.0, -0.0, -0.002214456978582924, 0.008361583668963536, 0.0031401942747203444, -0.0, -0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0028943545250037983, -0.0031301382844878753, 0.002113252994616467, 0.0, 0.0, 0.0, 0.0, 0.0, -0.0, -0.0, 0.0, -0.0, -0.0, -0.0010321050071136991, 0.008905753948020954, 0.0028464383724280478, -0.0, -0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, 0.0, -0.0053052889804602885, -0.0019271100770928186, 0.0012090042664300153, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, 0.0, 0.0, 0.0, -0.0, -0.0011945155805738324, 0.005654442809865844, 0.0020132075147173286, -0.0, -0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, 0.0, -0.0014689358119857122, 0.0010743412654248086, 0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, -0.0, 0.0, -0.0, -0.0, -0.0017047980433136346, 0.0029066051664685937, -0.0007805868937027288, -0.0, -0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, 0.0, -0.0, -0.0, 5.541726090138969e-05, 0.0014516115182299915, 0.0002827700518397855, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, 0.0, 0.0, 0.0, 0.0, -0.0, -0.001440140782635336, 0.002381249982038837, 0.002146825452068144, -0.0, -0.0, 0.0, -0.0, 0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, 0.001150052970321427, 0.0002865015237050364, 0.0029798150346815985, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.001775029606380323, 0.000833985914685474, -0.003770739075457816, -0.0, -0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, -0.0, 0.0, 0.0, -0.0006093176893524411, -0.00046905781658387527, 0.0034053217440919658, 0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, 0.0, -0.0, -0.0, -0.0007450012183962096, 0.001298767353118675, -0.008499247802184222, -6.145165255574976e-05, -0.0, -0.0, -0.0, 0.0, 0.0, -0.0, -0.0, -0.0, 0.0, -0.0, 0.0, 0.0011809726462884672, -0.0018384763902449712, 0.005411106715800028, 0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, -0.0, 0.0, -0.0021392341817010304, 0.0003259163122540385, -0.005276118905978749, -0.0019509840184772497, -9.545685077687876e-07, 0.0, -0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, -0.0, 0.0, 0.0, 0.0007772404694664217, -0.0001517954537059768, 0.006481484678129392, -0.0, 0.0, 0.0, -0.0, -0.0, -0.0, -0.0, -0.0, 8.098064554131295e-05, -0.0024904264199929506, -0.0020718618328775897, -5.3411287747038166e-05, -0.0004556472202791715, 0.0, -0.0, -0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, -0.0, 0.0, 0.0, 0.0, 0.0022750984867578, 0.001716405971437602, 0.0003221344811922982, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0015560282437342534, 9.107229584202956e-05, 0.0008772841867241755, 0.0006502979194500701, -0.004128780661881036, 0.0006030386196211547, 0.0, -0.0, 0.0, -0.0, -0.0, 0.0, 0.0, -0.0, -0.0, 0.0, -0.0, -0.0, 0.0, 0.0, 0.0013959959731925453, 0.0026791526421029673, 0.002399500793142178, -0.00044960969955281656, 0.003101832495190209, 0.007494535809079955, 0.002864118744003058, -0.003052590549800204, 0.003420222341277871, 0.0014924017873988514, -0.0009357389226494119, 0.0007856229438140384, -0.001843397373255761, 1.6031851430693252e-05, 0.0, 0.0, -0.0, -0.0, 0.0, 0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, 0.0, -0.000699901824825285, 0.0043822508549258565, -0.003541931476855951, -0.0028896746311921715, -0.0004873454583246359, -0.006087345141728267, 0.000388224886755815, 0.002533641621974457, -0.004352836429303485, -0.0006079421449756437, -0.003810133409713042, -0.0008284413779488711, 0.0, -0.0, 0.0, 0.0, -0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, 0.0, -0.0, -0.0, -0.0, -0.0, -0.0, 0.0010901530854686326, -0.013135007707490608, 0.0004734520308098294, 0.0020504232707536456, -0.006609452262924153, 0.0023647861306777536, 0.004678920703192049, -0.0018122526857900652, 0.0021375383049022263, 0.0, -0.0, -0.0, 0.0, 0.0, -0.0, -0.0, -0.0, 0.0, 0.0, 0.0, -0.0, -0.0, 0.0, -0.0, -0.0, -0.0, -0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, 0.0, -0.0, -0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, 0.0, 0.0, -0.0, -0.0, 0.0, -0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.0, -0.0, 0.0, -0.0, -0.0, -0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, 0.0, 0.0, -0.0, -0.0, -0.0, 0.0, 0.0, 0.0, 0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, -0.0, 0.0, 0.0, -0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.0, -0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.0, 0.0, -0.0, -0.0, -0.0, -0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]}]}