README.md

metadata

license: apache-2.0
datasets:
  - wikisql
language:
  - en
pipeline_tag: text2text-generation
tags:
  - nl2sql
widget:
  - text: 'question: get people name with age less 25 table: id, name, age'
    example_title: less than
  - text: 'question: get people name with age equal 25 table: id, name, age'
    example_title: equal

new version: LarkAI/codet5p-770m_nl2sql_oig

use oig-sql dataset and support more complex sql parse

How to Use

from transformers import AutoTokenizer, BartForConditionalGeneration

device = torch.device('cuda:0')

tokenizer = AutoTokenizer.from_pretrained("LarkAI/bart_large_nl2sql")
model = BartForConditionalGeneration.from_pretrained("LarkAI/bart_large_nl2sql").to(device)

text = "question: get people name with age less 25 table: id, name, age"
inputs = tokenizer([text], max_length=1024, return_tensors="pt")
output_ids = model.generate(inputs["input_ids"].to(self.device), num_beams=self.beams, max_length=128, min_length=8)
response_text = tokenizer.batch_decode(output_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
# SELECT name FROM table WHERE age < 25

reference: juierror/flan-t5-text2sql-with-schema - fix this discussion

How to Train

Quick start: https://github.com/huggingface/transformers/blob/main/examples/pytorch/summarization/README.md