export onnx RuntimeError: Expected tensor for argument #1 'indices' to have one of the following scalar types: Long, Int;

Question

I'm trying to exporting a roberta model from PyTorch to ONNX and running it using ONNX Runtime.

When I run the export

torch.onnx.export(roberta_model,
                  (input_ids),
                  ONNX_path,
                  input_names=['input'],
                  output_names=['output'],
                  dynamic_axes={'input' :{0 : 'batch_size',
                                          1: 'sentence_length'},
                                'output': {0: 'batch_size'}})

input_ids is:

input_ids = torch.tensor(profiles_input).squeeze(0)

It gives me the runtime error:

RuntimeError: Expected tensor for argument #1 'indices' to have one of the following scalar types: Long, Int; but got torch.FloatTensor instead (while checking arguments for embedding)

I have searched this error and most answers say to convert to long.

I tried the following ways to convert to long

input_ids = input_ids.type(torch.LongTensor)

and

c_input_ids = input_ids.long()

But it just gives me an out of index error:

IndexError: index out of range in self 

Also tried matching the vocab size according to this StackOverflow answer While training BERT variant, getting IndexError: index out of range in self

tokenizer = RobertaTokenizer.from_pretrained(model_name)
config = transformers.RobertaConfig()
config.vocab_size = RobertaTokenizer.vocab_size

But I still see the same runtime error.

Can I get some guidance on how to fix this?

This export was working before but now that I have loaded a model that I trained on my data it stopped working. There's no errors when loading the model_state_dict from my trained model.