Add gpt-j-6b model into torchbench

[chuanqi129]

Add HF model gpt-j-6b into torchbench

Work for roadmap #1293

$ python run.py hf_GPTJ --torchdynamo inductor
Running eval method from hf_GPTJ on cpu in dynamo inductor mode with input batch size 1 and precision fp32.
CPU Total Wall Time: 3891.941 milliseconds
CPU Peak Memory:               47.6953 GB
Correctness:                         True
PT2 Compilation time:      99.554 seconds