* optimize nltk

* upgrade paddlehub version (1.3.0)

* switch the dependent packages from tb-nightly to tensorboard>=1.15

* add flask version dependment

* add hub-serving

* modify py2 numpy&pandas version

* modify requirements

* Add Reading Comprehension Taskwq

* Add Reading Comprehension Task

* Add Reading Comprehension Task

* dd Reading Comprehension Task

* Add the reading comprehension task

* add reading comprehension task

* Add reading comprehension

* Add reading comprehension task

* add reading comprehension task

* Add reading comprehension task

* Add reading comprehension task

* Fix tokenization for bert chn

* Add GLUE and XNLI, modify text-classification demo test=develop (#94)

* Add GLUE and XNLI, modify text-classification demo

* add hub_server rw lock

* Support GLUE (#108)

* Support GLUE

* Support MNLI_m and MNLI_mm

* restore checkpoint.py

* Modify for review

* Fix the bug whether module is valid or not

* Enhancetask (#122)

* accelerate predict

* Add autoft (#127)

* add autofinetune

* update paddlehub required libs

* Add copyright

* Split task and add regression demo  (#130)

* split task

* add regression demo

* implement CombinedStrategy (#128)

* implement CombinedStrategy

* optimize some details (#131)

* optimize some details

* optimize some details (#133)

* support cv_reader

* use logger instead of print

* add warning

* remove todo

* add gradual_unfreeze frz_blocks

* replace the visualization toolkit from visualdl to tb-paddle (#139)

* replace the logging toolkit from visualdl to tb-paddle.

* modified:   requirements.txt

* modified:   paddlehub/finetune/task/basic_task.py

* update autofinetune  (#135)

* update autofinetune (add modelbased evaluator)

* support cpu count

* fix ci

* Fix the encoding format

* Add the elmo demo

* Fix the bug that coding type between py2 and py3 in lac-reader

* Add the lib 'chardet' for detecting the coding type.

* Modify the requirement.txt