Skip to content

QA Inference with BERT model using Triton Inference Server

Bidirectional Embedding Representations from Transformers (BERT), is a method of pre-training language representations which obtains state-of-the-art results on a wide array of Natural Language Processing (NLP) tasks.

This example demonstrates

  • Inference on Question Answering (QA) task with BERT Base/Large model
  • The use of fine-tuned NVIDIA BERT models
  • Deploy Transformer for preprocess using BERT tokenizer
  • Deploy BERT model on Triton Inference Server
  • Inference with V2 KServe protocol

We can run inference on a fine-tuned BERT model for tasks like Question Answering.

Here we use a BERT model fine-tuned on a SQuaD 2.0 Dataset which contains 100,000+ question-answer pairs on 500+ articles combined with over 50,000 new, unanswerable questions.


  1. Your cluster's Istio Ingress gateway must be network accessible.
  2. Skip tag resolution for which requires auth to resolve triton inference server image digest
    kubectl patch cm config-deployment --patch '{"data":{"registriesSkippingTagResolving":""}}' -n knative-serving
  3. Increase progress deadline since pulling triton image and big bert model may longer than default timeout for 120s, this setting requires knative 0.15.0+
    kubectl patch cm config-deployment --patch '{"data":{"progressDeadline": "600s"}}' -n knative-serving

Create Custom Transformer for BERT Tokenizer

Extend ModelServer base and Implement pre/postprocess

  • The preprocess handler converts the paragraph and the question to BERT input using BERT tokenizer
  • The predict handler calls Triton Inference Server using PYTHON REST API
  • The postprocess handler converts raw prediction to the answer with the probability

class BertTransformer(kserve.Model):
    def __init__(self, name: str, predictor_host: str):
        self.short_paragraph_text = "The Apollo program was the third United States human spaceflight program. First conceived as a three-man spacecraft to follow the one-man Project Mercury which put the first Americans in space, Apollo was dedicated to President John F. Kennedy's national goal of landing a man on the Moon. The first manned flight of Apollo was in 1968. Apollo ran from 1961 to 1972 followed by the Apollo-Soyuz Test Project a joint Earth orbit mission with the Soviet Union in 1975."

        self.predictor_host = predictor_host
        self.tokenizer = tokenization.FullTokenizer(vocab_file="/mnt/models/vocab.txt", do_lower_case=True)
        self.model_name = "bert_tf_v2_large_fp16_128_v2"
        self.triton_client = None

    def preprocess(self, inputs: Dict) -> Dict:
        self.doc_tokens = data_processing.convert_doc_tokens(self.short_paragraph_text)
        self.features = data_processing.convert_examples_to_features(self.doc_tokens, inputs["instances"][0], self.tokenizer, 128, 128, 64)
        return self.features

    def predict(self, features: Dict) -> Dict:
        if not self.triton_client:
            self.triton_client = httpclient.InferenceServerClient(
                url=self.predictor_host, verbose=True)

        unique_ids = np.zeros([1,1], dtype=np.int32)
        segment_ids = features["segment_ids"].reshape(1,128)
        input_ids = features["input_ids"].reshape(1,128)
        input_mask = features["input_mask"].reshape(1,128)

        inputs = []
        inputs.append(httpclient.InferInput('unique_ids', [1,1], "INT32"))
        inputs.append(httpclient.InferInput('segment_ids', [1, 128], "INT32"))
        inputs.append(httpclient.InferInput('input_ids', [1, 128], "INT32"))
        inputs.append(httpclient.InferInput('input_mask', [1, 128], "INT32"))

        outputs = []
        outputs.append(httpclient.InferRequestedOutput('start_logits', binary_data=False))
        outputs.append(httpclient.InferRequestedOutput('end_logits', binary_data=False))
        result = self.triton_client.infer(self.model_name, inputs, outputs=outputs)
        return result.get_response()

    def postprocess(self, result: Dict) -> Dict:
        end_logits = result['outputs'][0]['data']
        start_logits = result['outputs'][1]['data']
        n_best_size = 20

        # The maximum length of an answer that can be generated. This is needed
        #  because the start and end predictions are not conditioned on one another
        max_answer_length = 30

        (prediction, nbest_json, scores_diff_json) = \
           data_processing.get_predictions(self.doc_tokens, self.features, start_logits, end_logits, n_best_size, max_answer_length)
        return {"predictions": prediction, "prob": nbest_json[0]['probability'] * 100.0}
Please find the code example here.

Build Transformer docker image

Build the KServe Transformer image with above code

cd bert_tokenizer_v2
docker build -t $USER/bert_transformer-v2:latest . --rm
Or you can use the prebuild image kfserving/bert-transformer-v2:latest

Create the InferenceService

Add above custom KServe Transformer image and Triton Predictor to the InferenceService spec

apiVersion: ""
kind: "InferenceService"
  name: "bert-v2"
      - name: kserve-container      
        image: kfserving/bert-transformer-v2:latest
          - "python"
          - "-m"
          - "bert_transformer_v2"
          - name: STORAGE_URI
            value: "gs://kfserving-examples/models/triton/bert-transformer"
      runtimeVersion: 20.10-py3
          cpu: "1"
          memory: 8Gi
          cpu: "1"
          memory: 8Gi
      storageUri: "gs://kfserving-examples/models/triton/bert"

Apply the InferenceService yaml.

kubectl apply -f bert_v1beta1.yaml

Expected Output created

Check the InferenceService

kubectl get inferenceservice bert-v2
NAME      URL                                           READY   AGE
bert-v2   True    71s
you will see both transformer and predictor are created and in ready state
kubectl get revision -l
NAME                                CONFIG NAME                   K8S SERVICE NAME                    GENERATION   READY   REASON
bert-v2-predictor-default-plhgs     bert-v2-predictor-default     bert-v2-predictor-default-plhgs     1            True    
bert-v2-transformer-default-sd6nc   bert-v2-transformer-default   bert-v2-transformer-default-sd6nc   1            True  

Run a Prediction

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

Send a question request with following input, the transformer expects sending a list of instances or inputs and preprocess then converts the inputs to expected tensor sending to Triton Inference Server.

  "instances": [
    "What President is credited with the original notion of putting Americans in space?"

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

curl -v -H "Host: ${SERVICE_HOSTNAME}" -d $INPUT_PATH http://${INGRESS_HOST}:${INGRESS_PORT}/v1/models/$MODEL_NAME:predict

Expected output

{"predictions": "John F. Kennedy", "prob": 77.91848979818604}

Back to top