Benchmark code using torch.onnx.export

Describe the bug The benchmark code using torch.onnx.export uses argument example_outputs but this named argument was removed from the export function.

Urgency non

System information

OS Platform and Distribution (e.g., Linux Ubuntu 16.04): linux ubuntu 20.04
ONNX Runtime installed from (source or binary): binary
ONNX Runtime version: 1.11.1
Python version: 3.7
CUDA/cuDNN version: 11.6
GPU model and memory:

To Reproduce

Describe steps/code to reproduce the behavior.
Attach the ONNX model to the issue (where applicable) to expedite investigation. Run
```
python -m onnxruntime.transformers.benchmark -g -m bert-base-cased
```

Expected behavior Benchmark runs and reports results

Screenshots If applicable, add screenshots to help explain your problem.

Additional context output

Arguments: Namespace(batch_sizes=[1], cache_dir='./cache_models', detail_csv=None, disable_ort_io_binding=False, engines=['onnxruntime'], force_num_layers=None, fusion_csv=None, input_counts=[1], model_class=None, model_source='pt', models=['bert-base-cased'], num_threads=[64], onnx_dir='./onnx_models', optimizer_info=<OptimizerInfo.BYSCRIPT: 'by_script'>, overwrite=False, precision=<Precision.FLOAT32: 'fp32'>, provider=None, result_csv=None, sequence_lengths=[4, 8, 16, 32, 64, 128, 256], test_times=100, use_gpu=True, validate_onnx=False, verbose=False)
Model class name: AutoModel
Some weights of the model checkpoint at bert-base-cased were not used when initializing BertModel: ['cls.predictions.decoder.weight', 'cls.seq_relationship.weight', 'cls.predictions.transform.LayerNorm.weight', 'cls.seq_relationship.bias', 'cls.predictions.transform.dense.bias', 'cls.predictions.transform.dense.weight', 'cls.predictions.transform.LayerNorm.bias', 'cls.predictions.bias']
- This IS expected if you are initializing BertModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing BertModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
Exporting ONNX model to ./onnx_models/bert_base_cased_1.onnx
Exception
Traceback (most recent call last):
  File "/net/storage149/mnt/md0/ccyang/miniconda/envs/mini-nlp/lib/python3.7/site-packages/onnxruntime/transformers/benchmark.py", line 587, in main
    args.model_source)
  File "/net/storage149/mnt/md0/ccyang/miniconda/envs/mini-nlp/lib/python3.7/site-packages/onnxruntime/transformers/benchmark.py", line 108, in run_onnxruntime
    validate_onnx, use_raw_attention_mask, overwrite, model_fusion_statistics)
  File "/net/storage149/mnt/md0/ccyang/miniconda/envs/mini-nlp/lib/python3.7/site-packages/onnxruntime/transformers/onnx_exporter.py", line 396, in export_onnx_model_from_pt
    use_external_data_format=use_external_data_format)
TypeError: export() got an unexpected keyword argument 'example_outputs'
No any result avaiable.

microsoft / onnxruntime

Benchmark code using torch.onnx.export #11399