Error while loading pre-trained model for PAWSX and XNLI

[wasiahmad]

When we try to load pre-trained model from model_name_or_path in this line for run_classify.py, it throws error (as expected).

I think we should instead use args.init_checkpoint as:

if args.init_checkpoint and os.path.exists(args.init_checkpoint):
    # set global_step to gobal_step of last saved checkpoint from model path
    global_step = int(args.init_checkpoint.split("-")[-1].split("/")[0])

Am I missing anything?

[melvinjosej]

Hi Wasi,

Sorry for the delay in getting back. We're looking into this now and will have a response in a few days.