onnxruntime

Add hugging-face models loss curve and performance guards to ROCm CI pipeline. (#8915)

* test running hf bert-large

* try again

* try again

* include other models

* correct names

* disable deberta-v2-xxlarge

* avoid torch.distributed

* add compare json loss and perf for bert-large to test

* fix sed expression

* remove pytest

* add more models

* move unit tests u

* display samples/sec