Exporting our checkpoint with the transformers.onnx. (AutoModelForSequenceClassification)

 refer to codeL

..

import torch from transformers import AutoModelForSequenceClassification, AutoTokenizer # load model and tokenizer model_id = "distilbert-base-uncased-finetuned-sst-2-english" model = AutoModelForSequenceClassification.from_pretrained(model_id) tokenizer = AutoTokenizer.from_pretrained(model_id) dummy_model_input = tokenizer("This is a sample", return_tensors="pt") # export torch.onnx.export( model, tuple(dummy_model_input.values()), f="torch-model.onnx", input_names=['input_ids', 'attention_mask'], output_names=['logits'], dynamic_axes={'input_ids': {0: 'batch_size', 1: 'sequence'}, 'attention_mask': {0: 'batch_size', 1: 'sequence'}, 'logits': {0: 'batch_size', 1: 'sequence'}}, do_constant_folding=True, opset_version=13, )

..

Thank you.