|
from typing import Dict, List, Any |
|
from optimum.onnxruntime import ORTModelForFeatureExtraction |
|
from transformers import pipeline, AutoTokenizer |
|
|
|
def cls_pooling(model_output): |
|
return model_output.last_hidden_state[:,0] |
|
class PreTrainedPipeline(): |
|
def __init__(self, path=""): |
|
|
|
self.model = ORTModelForFeatureExtraction.from_pretrained(path) |
|
self.tokenizer = AutoTokenizer.from_pretrained(path, model_max_length=128) |
|
|
|
|
|
def __call__(self, inputs: Any) -> Dict[str, Any]: |
|
""" |
|
Args: |
|
data (:obj:`str`): |
|
a string containing some text |
|
Return: |
|
A :obj:`list`:. The object returned should be a list of one list like [[{"label": 0.9939950108528137}]] containing : |
|
- "label": A string representing what the label/class is. There can be multiple labels. |
|
- "score": A score between 0 and 1 describing how confident the model is for this label/class. |
|
""" |
|
|
|
encoded_input = self.tokenizer(inputs, padding="longest", truncation=True, return_tensors='pt') |
|
|
|
model_output = self.model(**encoded_input, return_dict=True) |
|
embeddings = cls_pooling(model_output) |
|
|
|
return {"vectors": [float(vec) for vec in embeddings[0].tolist()]} |
