Fine-Tuning use dbgpt_hub

The DB-GPT-Hub project has released a pip package to lower the threshold for Text2SQL training. In addition to fine-tuning through the scripts provided in the warehouse, you can alse use the Python package we provide for fine-tuning.

Install

pip install dbgpt_hub

Show Baseline

from dbgpt_hub.baseline import show_scores
show_scores()

Fine-tuning

from dbgpt_hub.data_process import preprocess_sft_data
from dbgpt_hub.train import train_sft
from dbgpt_hub.predict import start_predict
from dbgpt_hub.eval import start_evaluate

Preprocessing data into fine-tuned data format.

data_folder = "dbgpt_hub/data"
data_info = [
     {
        "data_source": "spider",
        "train_file": ["train_spider.json", "train_others.json"],
        "dev_file": ["dev.json"],
        "tables_file": "tables.json",
        "db_id_name": "db_id",
        "is_multiple_turn": False,
        "train_output": "spider_train.json",
        "dev_output": "spider_dev.json",
    }
]

preprocess_sft_data(
      data_folder = data_folder,
      data_info = data_info
)

Fine-tune the basic model and generate model weights

train_args = {
            "model_name_or_path": "codellama/CodeLlama-13b-Instruct-hf",
            "do_train": True,
            "dataset": "example_text2sql_train",
            "max_source_length": 2048,
            "max_target_length": 512,
            "finetuning_type": "lora",
            "lora_target": "q_proj,v_proj",
            "template": "llama2",
            "lora_rank": 64,
            "lora_alpha": 32,
            "output_dir": "dbgpt_hub/output/adapter/CodeLlama-13b-sql-lora",
            "overwrite_cache": True,
            "overwrite_output_dir": True,
            "per_device_train_batch_size": 1,
            "gradient_accumulation_steps": 16,
            "lr_scheduler_type": "cosine_with_restarts",
            "logging_steps": 50,
            "save_steps": 2000,
            "learning_rate": 2e-4,
            "num_train_epochs": 8,
            "plot_loss": True,
            "bf16": True,
}

start_sft(train_args)

Predictive model output results

predict_args = {
            "model_name_or_path": "codellama/CodeLlama-13b-Instruct-hf",
            "template": "llama2",
            "finetuning_type": "lora",
            "checkpoint_dir": "dbgpt_hub/output/adapter/CodeLlama-13b-sql-lora",
            "predict_file_path": "dbgpt_hub/data/eval_data/dev_sql.json",
            "predict_out_dir": "dbgpt_hub/output/",
            "predicted_out_filename": "pred_sql.sql",
}
start_predict(predict_args)

Evaluate the accuracy of the output results on the test datasets

evaluate_args =  {
            "input": "./dbgpt_hub/output/pred/pred_sql_dev_skeleton.sql",
            "gold": "./dbgpt_hub/data/eval_data/gold.txt",
            "gold_natsql": "./dbgpt_hub/data/eval_data/gold_natsql2sql.txt",
            "db": "./dbgpt_hub/data/spider/database",
            "table": "./dbgpt_hub/data/eval_data/tables.json",
            "table_natsql": "./dbgpt_hub/data/eval_data/tables_for_natsql2sql.json",
            "etype": "exec",
            "plug_value": True,
            "keep_distict": False,
            "progress_bar_for_each_datapoint": False,
            "natsql": False,
}
start_evaluate(evaluate_args)