Onnx Runtime Errors With LongT5

### System Info - `optimum` version: 1.2.3 (installed via Github installation) - `transformers` version: 4.20.1 - Platform: Linux-5.4.188+-x86_64-with-Ubuntu-18.04-bionic - Python version: 3.7.13 - Huggingface_hub version: 0.8.1 - PyTorch version (GPU?): 1.11.0+cu113 (False) - Tensorflow version (GPU?): 2.8.2 (False) - Flax version (CPU?/GPU?/TPU?): not installed (NA) - Jax version: not installed - JaxLib version: not installed - Using GPU in script?: no - Using distributed or parallel set-up in script?: no ### Who can help? @stancld @echarlaix @LysandreJik ### Information - [ ] The official example scripts - [X] My own modified scripts ### Tasks - [ ] An officially supported task in the `examples` folder (such as GLUE/SQuAD, ...) - [x] My own task or dataset (give details below) ### Reproduction LongT5 with TGlobal Attention isn't able to run sequences longer than **global_block_size * 2**. This is because during the model tracing [num_globals > 0](https://github.com/huggingface/transformers/blob/main/src/transformers/models/longt5/modeling_longt5.py#L191) is being converted to False. I originally posted the error in Optimum (https://github.com/huggingface/optimum/issues/285) but @echarlaix asked me to open an issue here because this error concerns the ONNX export. Code to reproduce is below: ``` !pip install transformers !pip install transformers[onnx] !python -m pip install git+https://github.com/huggingface/optimum.git !python -m pip install git+[https://github.com/huggingface/optimum.git#egg=optimum[onnxruntime]](https://github.com/huggingface/optimum.git#egg=optimum%5Bonnxruntime%5D) !pip install datasets ``` ```py from optimum.onnxruntime import ORTModelForSeq2SeqLM model = ORTModelForSeq2SeqLM.from_pretrained("longt5-tglobal-base", from_transformers=True) from transformers import AutoTokenizer, pipeline tokenizer = AutoTokenizer.from_pretrained('google/long-t5-tglobal-base') onnx_summarization = pipeline("summarization", model=model, tokenizer=tokenizer) tex…