File size: 884 Bytes
2aa7bf4 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 |
# rag/models/embedder.py
from typing import List
import numpy as np
import onnxruntime as ort
from fastapi import Request
def _l2_normalize(vec: np.ndarray) -> List[float]:
norm = np.linalg.norm(vec) or 1.0
return (vec / norm).tolist()
def get_embedding(request: Request, text: str) -> List[float]:
"""
request.app.state.embedder_sess : ONNX Runtime InferenceSession
request.app.state.embedder_tokenizer : 토크나이저
"""
tokenizer = request.app.state.embedder_tokenizer
sess: ort.InferenceSession = request.app.state.embedder_sess
inputs = tokenizer(text, return_tensors="np", padding=True, truncation=True, max_length=256)
ort_inputs = {k: v for k, v in inputs.items()}
ort_outs = sess.run(None, ort_inputs)
# 일반적으로 첫 번째 출력이 [batch, dim] 임베딩
vec = ort_outs[0][0]
return _l2_normalize(vec)
|