ORTModel.save_pretrained doesn't save model_type key in its config.json

System Info

optimum==1.4.0
transformers==4.22.1
onnx==1.12.0
onnxruntime==1.12.1

Python 3.9.13

Who can help?

@lewtun @michaelbenayoun

Information

[X] The official example scripts
[ ] My own modified scripts

Tasks

[X] An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
[ ] My own task or dataset (give details below)

Reproduction

from optimum.onnxruntime import ORTModelForQuestionAnswering
from transformers import AutoTokenizer

# Load a model from transformers and export it to ONNX
ort_model = ORTModelForQuestionAnswering.from_pretrained(MODEL_PATH, from_transformers=True)
tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH)

ONNX_PATH = f"{MODEL_PATH}/onnx"

# Save the onnx model and tokenizer
ort_model.save_pretrained(ONNX_PATH)
tokenizer.save_pretrained(ONNX_PATH)

from optimum.onnxruntime import ORTOptimizer

optimizer = ORTOptimizer.from_pretrained(ONNX_PATH) # <- This fails

Expected behavior

Loading the ORT model for the optimizer with ORTOptimizer.from_pretrained(ONNX_PATH) fails with: ValueError: Unrecognized model in ... Should have a model_type key in its config.json, or contain one of the following strings in its name: ...

The config.json generated when converting to Onnx doesn't have a model_type key. Unlike the config.json of the 🤗 Transformers model: "model_type": "xlm-roberta"; There only seems to be a "name": "XLMRoberta" key.

But the optimizer and quantizer classes expect a model_type key when loading the ort_model.

I was expecting the .save_pretrained method of a ORTModel class to also save the model_type key from the 🤗 Transformers config.json model.

huggingface / optimum