Support MPNet model in NormalizedConfig and ORTOptimizer

[fatma-mahfoudh]

Feature request

Support MLPNet in:

• NormalizedConfig
• ORTOptimizer

Motivation

I would like to optimize the inference of PatentSBERTa model using optimum.onnxruntime When I run the following code, I get a NonImplementedError:

from optimum.onnxruntime import ORTModelForFeatureExtraction
from optimum.onnxruntime import ORTOptimizer
model_id = "AI-Growth-Lab/PatentSBERTa"
model = ORTModelForFeatureExtraction.from_pretrained(model_id, from_transformers=True)
optimizer = ORTOptimizer.from_pretrained(model)

---

KeyError Traceback (most recent call last)
File ~\NLP\lib\site-packages\optimum\onnxruntime\optimization.py:67, in ORTOptimizer.init(self, onnx_model_path, config, from_ortmodel)
66 try:
---> 67 self.normalized_config = NormalizedConfigManager.get_normalized_config_class(self.model_type)(self.config)
68 except KeyError:

File ~\NLP\lib\site-packages\optimum\utils\normalized_config.py:269, in NormalizedConfigManager.get_normalized_config_class(cls, model_type)
267 @classmethod
268 def get_normalized_config_class(cls, model_type: str) -> Type:
--> 269 cls.check_supported_model(model_type)
270 return cls._conf[model_type]

File ~\NLP\lib\site-packages\optimum\utils\normalized_config.py:262, in NormalizedConfigManager.check_supported_model(cls, model_type)
261 model_types = ", ".join(cls._conf.keys())
--> 262 raise KeyError(
263 f"{model_type} model type is not supported yet in NormalizedConfig. Only {model_types} are supported. "
264 f"If you want to support {model_type} please propose a PR or open up an issue."
265 )

KeyError: 'mpnet model type is not supported yet in NormalizedConfig. Only albert, bart, bert, blenderbot, blenderbot_small, bloom, falcon, camembert, codegen, cvt, deberta, deberta-v2, deit, distilbert, donut-swin, electra, encoder-decoder, gpt2, gpt-bigcode, gpt_neo, gpt_neox, llama, gptj, imagegpt, longt5, marian, mbart, mt5, m2m_100, nystromformer, opt, pegasus, pix2struct, poolformer, regnet, resnet, roberta, speech_to_text, splinter, t5, trocr, whisper, vision-encoder-decoder, vit, xlm-roberta, yolos, mpt, gpt_bigcode are supported. If you want to support mpnet please propose a PR or open up an issue.'

During handling of the above exception, another exception occurred:

NotImplementedError Traceback (most recent call last)
Cell In[4], line 5
2 from optimum.onnxruntime.configuration import OptimizationConfig
4 # create ORTOptimizer and define optimization configuration
----> 5 optimizer = ORTOptimizer.from_pretrained(model)
6 optimization_config = OptimizationConfig(optimization_level=99) # enable all optimizations
8 # apply the optimization configuration to the model

File ~\NLP\lib\site-packages\optimum\onnxruntime\optimization.py:125, in ORTOptimizer.from_pretrained(cls, model_or_path, file_names)
123 else:
124 raise ValueError(f"Unable to load the model from {model_or_path}.")
--> 125 return cls(onnx_model_path, config=config, from_ortmodel=from_ortmodel)

File ~\NLP\lib\site-packages\optimum\onnxruntime\optimization.py:69, in ORTOptimizer.init(self, onnx_model_path, config, from_ortmodel)
67 self.normalized_config = NormalizedConfigManager.get_normalized_config_class(self.model_type)(self.config)
68 except KeyError:
---> 69 raise NotImplementedError(
70 f"Tried to use ORTOptimizer for the model type {self.model_type}, but it is not available yet. Please open an issue"
71 " or submit a PR at https://github.com/huggingface/optimum."
72 )

NotImplementedError: Tried to use ORTOptimizer for the model type mpnet, but it is not available yet. Please open an issue or submit a PR at https://github.com/huggingface/optimum.

Your contribution

no