update PyTorch Bert SquAD notebooks to use onnxruntim-tools and update usage of intra_op_num_threads.
rename python files according to coding style
Fix change_input_to_int32.
update keras notebook to copy script from rel-1.3.0 branch (Will update them later)
* Enable optimizer on models with external data (>2GB)
* Refactoring optimizer: move fusion to separate file
* Update benchmark: (1) output datatime to csv (2) Add option --onnx_dir to benchmark.py for onnx model directory path (3) add gpt2-large (4) loose thrsholds for fp16 validation
* update optimizer (1) Add attribute of ConstantOfShape in fp16 conversion (2) Use OnnxRuntime level 1 optimization
* update bert_perf_test.py: rename --input_ids to --input_ids_name