Ftest_emb# seq2sql
-
Use WikiSQL/annotate.py to annotate files in WikiSQL/data. The usage of annotate.py could be found at https://github.com/salesforce/WikiSQL
-
Use WikiSQL/convert.py to extract input/outputs for seq2seq task from annotated files.
Vocabs are generated automatically when you run the training script. If you want to create the vocabe your self, you can use code/create_vocab.py
-
The embeddings is a concatenate of Glove embeddings(http://nlp.stanford.edu/projects/glove) and Char embeddings(http://www.logos.t.u-tokyo.ac.jp/~hassy/publications/arxiv2016jmt/jmt_pre-trained_embeddings.tar.gz).
-
You can generate your own embedding file related to your vocab dict using code/test_emb.py
Preprocessing procedures are already done for train and dev sets.
python translate.py --from_train_data ./wikisql_in_nmt/train.seq --to_train_data ./wikisql_in_nmt/train.sql --from_dev_data ./wikisql_in_nmt/dev.seq --to_dev_data ./wikisql_in_nmt/dev.sql --train /your/path/to/save/model --steps_per_checkpoint 1000 --drop_out 0.3 --gpu_id 1 --optim_method sgd --train_emb False --pretrain_embs ./data/emb_50000.pkl
You may modifed the code in decode() function in translate.py to get your goals.
python translate.py --from_train_data ./wikisql_in_nmt/train.seq --to_train_data ./wikisql_in_nmt/train.sql --from_dev_data ./wikisql_in_nmt/dev.seq --to_dev_data ./wikisql_in_nmt/dev.sql --train /your/path/to/save/model --steps_per_checkpoint 1000 --drop_out 0.3 --gpu_id 1 --optim_method sgd --train_emb False --pretrain_embs ./data/emb_50000.pkl --decode True
Fisrt convert the output to json file.
python3 convert_to_json_split_py3.py --din tmp.eval.ids.true --dout out_split.json --dsource /users1/ybsun/seq2sql/WikiSQL/annotated/dev.jsonl --dtable ../WikiSQL/data/dev.tables.jsonl
or
python3 convert_to_json_split_py3.py --din tmp.eval.ids.true --dout out_split.json --dsource /users1/ybsun/seq2sql/WikiSQL/annotated/dev.jsonl --dtable ../WikiSQL/data/dev.tables.jsonl --strong True
Then use the evaluating scripts.
python3 evaluate_split_py3.py ../WikiSQL/data/dev.jsonl ../WikiSQL/data/dev.db ./out_split.json ../WikiSQL/annotated/dev.jsonl ../WikiSQL/data/dev.tables.jsonl ./wikisql_in_nmt/dev.sql
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent