from optimum.onnxruntime import ORTModelForTokenClassification
from transformers import AutoTokenizer

model = ORTModelForTokenClassification.from_pretrained("xa2go/xlm-roberta-large-lemma-eu-onnx-int8")
tokenizer = AutoTokenizer.from_pretrained("xa2go/xlm-roberta-large-lemma-eu-onnx-int8")

text = "Euskal Herriko Unibertsitatea Bilboko campusean dago."
inputs = tokenizer(text, return_tensors="np")
outputs = model(**inputs)

Características

Format: ONNX INT8 quantized
Optimizado: AVX2 (AMD EPYC, Intel Xeon)
Speedup: 2-3x vs FP32
Tamaño: 75% reducción