update: star readme

star/README.md

@@ -1,6 +1,19 @@
-# STAR
+# 🌟 STAR: SQL Guided Pre-Training for Context-dependent Text-to-SQL Parsing
 
-This is the project containing source code for the EMNLP 2022 paper "**STAR: SQL Guided Pre-Training for Context-dependent Text-to-SQL Parsing***"*
+<p align="center">
+    <a href="./LICENSE"><img src="https://img.shields.io/badge/license-MIT-red.svg">
+    </a>
+  	<a href="https://github.com/huggingface/transformers/tree/main/examples/research_projects/tapex">
+      <img alt="🤗 transformers support" src="https://img.shields.io/badge/🤗 transformers-master-green" />
+    </a>
+    <a href="support os"><img src="https://img.shields.io/badge/os-linux%2C%20win%2C%20mac-pink.svg">
+    </a>
+    <a href=""><img src="https://img.shields.io/badge/python-3.7+-aff.svg">
+    </a>
+    <br />
+</p>
+
+This is the official project containing source code for the EMNLP 2022 paper "STAR: SQL Guided Pre-Training for Context-dependent Text-to-SQL Parsing"
 
 You can use our checkpoint to evaluation directly or train from scratch with our instructions.
 
@@ -10,8 +23,17 @@ You can use our checkpoint to evaluation directly or train from scratch with our
 
 The relevant models and data involved in the paper can be downloaded through [Baidu Netdisk](https://pan.baidu.com/s/1uA63h4zpwyDSqY5cprbeJQ?pwd=6666), or downloaded through Google Drive in the corresponding folder.
 
+## Citation
+```
+@article{cai2022star,
+  title={STAR: SQL Guided Pre-Training for Context-dependent Text-to-SQL Parsing},
+  author={Cai, Zefeng and Li, Xiangyu and Hui, Binyuan and Yang, Min and Li, Bowen and Li, Binhua and Cao, Zheng and Li, Weijie and Huang, Fei and Si, Luo and others},
+  journal={arXiv preprint arXiv:2210.11888},
+  year={2022}
+}
+```
 
-## Pretrain
+## 🪜 Pretrain
 
 
 ### Create conda environment
@@ -49,7 +71,7 @@ python save_model.py
 
 Then you can get the trained model and its configuration (at least containing `model.bin` and `config.json`) under `pretrained/sss` direction.
 
-## Finetuning
+## 🚪 Fine-tuning and Evaluation
 
 This section presents the results on CoSQL and SParC datasets with STAR fine-tuned with LGESQL.
 
@@ -70,44 +92,46 @@ Create conda environment `lgesql`:
     python -c "import stanza; stanza.download('en')"
     python -c "import nltk; nltk.download('stopwords')"
     ```
-## Using our checkpoint to evaluation:
+### Using our checkpoint to evaluation:
   - Download our processed datasets [CoSQL](https://drive.google.com/file/d/1suuQnHVPxZZKRiUBvsUIlw7BnY21Q_6u/view?usp=sharing) or [SParC](https://drive.google.com/file/d/1DrGBq7WGdieanq90TjkiO5JgZMwcDGUu/view?usp=sharing) and unzip them into the `cosql/data` and `sparc/data` respectively. Make sure the datasets are correctly located as:
-```
+    ```
-data
+    data
-├── database
+    ├── database
-├── dev_electra.json
+    ├── dev_electra.json
-├── dev_electra.bin
+    ├── dev_electra.bin
-├── dev_electra.lgesql.bin
+    ├── dev_electra.lgesql.bin
-├── dev_gold.txt
+    ├── dev_gold.txt
-├── label.json
+    ├── label.json
-├── tables_electra.bin
+    ├── tables_electra.bin
-├── tables.json
+    ├── tables.json
-├── train_electra.bin
+    ├── train_electra.bin
-├── train_electra.json
+    ├── train_electra.json
-└── train_electra.lgesql.bin
+    └── train_electra.lgesql.bin
-```
+    ```
   - Download our processed checkpoints [CoSQL](https://drive.google.com/file/d/1y4edJJ2xoA_JUGCoegEd8xLopAaUuvmp/view?usp=sharing) or [SParC](https://drive.google.com/file/d/1UDs956PgVlZT1hZ4pRm3Mox3Hs5u42sF/view?usp=sharing) and unzip them into the `cosql/checkpoints` and `sparc/checkpoints` respectively. Make sure the checkpoints are correctly located as:
-```
+    ```
-checkpoints
+    checkpoints
-├── model_IM.bin
+    ├── model_IM.bin
-└── params.json
+    └── params.json
-```
+    ```
   - Execute the following command and the results are recorded in result_XXX.txt(it will take 10 to 30 minutes on one Tesla V100-PCIE-32GB GPU):
+    ```
+    sh run/run_evaluation.sh
+    ```
 
-        sh run/run_evaluation.sh
+### Train from scratch
-
-## Train from scratch
   - You can train STAR yourself by following the process in the `pretrain` file or download our pre-trained [STAR](https://drive.google.com/file/d/1zfvNpofVzLixzzFyqLO0NP-WQSKKENIC/view?usp=sharing) and unzip it into the `pretrained_models/sss` directory. Make sure the STAR are correctly located as:
-```
+    ```
-pretrained_models
+    pretrained_models
-└── sss
+    └── sss
-      ├── config.json
+        ├── config.json
-      ├── pytorch_model.bin
+        ├── pytorch_model.bin
-      └── vocab.txt
+        └── vocab.txt
-```
+    ```
   - You can preprocess the data with the `process_data&&label.py` file and refer to the methods in LGESQL, or download our processed data as described above directly. 
   - Traning:
   (it will take 4 days on one Tesla V100-PCIE-32GB GPU)
     ```
     sh run/run_lgesql_plm.sh
-    ```
+    ```
+