transformers

FX tracing improvement

#14321

Merged

Updated the tests and modeling xlnet

Added the non recoding of leaf modules to avoid recording more values for the methods to record than what will be seen at tracing time (which would otherwise desynchronize the recorded values and the values that need to be given to the proxies during tracing, causing errors).

Comments and making tracing work for gpt-j and xlnet

Refactored things related to num_choices (and batch_size, sequence_length)

style fix

Updated fx to work on PyTorch 1.10

Postponed autowrap_function feature usage for later

style fix

fixed issue

implemented suggestions

Add copyrights

Remove unnecessary file

Merge branch 'master' into fx_tracing_enhancement

Fix copies

Fix issue with add_new_model_like

Fix issue with add_new_model_like

Apply suggestions

Apply suggestions