I tried using bert-tokenizer only for tokenization but cannot vouch for it . Further tried below.

You need to use tensorflow 1.11.0. TF syntax has changed for 2.0 which is why you're getting your errors.
But the migration will be worthless as many things are deprecated in 2.0
On Tue, 7 Jan 2020, 09:57 adilapapaya, notifications@github.com wrote:
You need to use tensorflow 1.11.0. TF syntax has changed for 2.0 which is
why you're getting your errors.—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
https://github.com/google-research/bert/issues/977?email_source=notifications&email_token=AFIHA62FH3GBZHDOOAIPEFLQ4QAELA5CNFSM4KB2YJR2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEIHU2WY#issuecomment-571428187,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/AFIHA65NKJCU24NAKBTL46DQ4QAELANCNFSM4KB2YJRQ
.
A quick monkey patch of tensorflow is a quick and ugly fix for this. Put the following line after you import tensorflow:
tf.gfile = tf.io.gfile
Most helpful comment
A quick monkey patch of tensorflow is a quick and ugly fix for this. Put the following line after you import tensorflow:
tf.gfile = tf.io.gfile