NLP-Demo-for-Benchmark-Tasks

Introduction

this repository is for NLP beginners to get started with state-of-the-art NLP models and benchmark tasks.

Start-up

seeing is believing. just do it before asking why.

• git clone this repository
• for Mac/Linux, run .sh file in Teminal

./sh_file_name.sh

• for Windows, run Python command in .sh file

To-do

so far benchmark tasks demos that have been added.

• bert_demo
  • run
    • run_myCola.sh
    • run_myWnli.sh
    • run_myRte.sh

.sh	pretraining-model	accuracy
run_myCola.sh	uncased_L-8_H-512_A-8	0.74
run_myWnli.sh	uncased_L-8_H-512_A-8	0.35
run_myRte.sh	uncased_L-8_H-512_A-8	0.61

Requirement

some of the following not included due to repository size restriction.

• Python 3
• TensorFlow 1.0
• dataset (excluded)
  • GLUE
    • CoLA
    • WNLI
    • RTE
• pretraining models (excluded)
  • BERT
    • uncased_L-8_H-512_A-8