Infer whether a model needs to be exported to ONNX or not

[echarlaix]

Infer whether the model should be exported or not by checking whether a .onnx file is present when loading the model, removing the need to specify the export argument

  from optimum.onnxruntime import ORTModelForCausalLM

  model_id = "optimum-internal-testing/tiny-random-llama"
- model = ORTModelForCausalLM.from_pretrained(model_id, export=True)
+ model = ORTModelForCausalLM.from_pretrained(model_id)

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[echarlaix]

Failing tests unrelated so merging