Description

This PR adds the ONNX weights exported with optimum-onnx as:

from optimum.onnxruntime import ORTModelForSequenceClassification

model = ORTModelForSequenceClassification.from_pretrained("Roblox/roblox-pii-classifier", export=True)
model.save_pretrained("onnx/")

To be later on used via Python as follows (amongst many other options leveraging ONNX):

from transformers import AutoTokenizer, pipeline
from optimum.onnxruntime import ORTModelForSequenceClassification

model = ORTModelForSequenceClassification.from_pretrained("Roblox/roblox-pii-classifier", revision="refs/pr/3", subfolder="onnx/")
tokenizer = AutoTokenizer.from_pretrained("Roblox/roblox-pii-classifier", revision="refs/pr/3")

pipe = pipeline("text-classification", model=model, tokenizer=tokenizer)
print(pipe("He never went out without a book under his arm"))

Thanks @jasonxie-rblx 🤗