import onnxruntime import numpy as np from transformers import AutoTokenizer from sentence_transformers import SentenceTransformer from typing import List, Dict class EndpointHandler: def __init__(self, path="."): # Initialize ONNX session using the provided model path self.onnx_session = onnxruntime.InferenceSession(f"{path}/model.onnx") self.tokenizer = AutoTokenizer.from_pretrained("neuml/pubmedbert-base-embeddings") self.sbert_model = SentenceTransformer("neuml/pubmedbert-base-embeddings") def preprocess(self, inputs: Dict) -> List[str]: return inputs["inputs"] def inference(self, sentences: List[str]) -> np.ndarray: inputs = self.tokenizer(sentences, padding=True, truncation=True, return_tensors="np") onnx_inputs = { "input_ids": inputs["input_ids"], "attention_mask": inputs["attention_mask"] } onnx_outputs = self.onnx_session.run(["sentence_embedding"], onnx_inputs) return onnx_outputs[0] def postprocess(self, embeddings: np.ndarray) -> List[List[float]]: return embeddings.tolist() def __call__(self, inputs: Dict) -> List[List[float]]: sentences = self.preprocess(inputs) embeddings = self.inference(sentences) return self.postprocess(embeddings)