This is the official repo for UniDM: A Unified Framework for Data Manipulation with Large Language Models.
mkdir dataset
wget https://fm-data-tasks.s3.us-west-1.amazonaws.com/datasets.tar.gz -P dataset
tar xvf dataset/datasets.tar.gz -C dataset/
pip install -r requirements.txt
# Manifest
git clone git@github.com:HazyResearch/manifest.git
cd manifest
pip install -e .
To run inference, use
python inference.py --help
Some examples are as follows: (The API KEY can be obtained by registering with the LLM provider. For instance, if you want to run inference with the OpenAI API models, create an account here.)
To run data imputation task select 3 example to add to the prompt by the auto-retrieve module, using both the adaptive data parsing and prompt engineering modules,
python inference.py \
--api_key <YOUR API KEY> \
--data_dir <DATA DIR> \
--task data_imputation \
--instance_num 3 \
--metadata_wise \
--instance_wise \
--data_parsing \
--prompt_engineering
To run entity resolution task and select 3 examples to add to the prompt, using both the adaptive data parsing and prompt engineering modules,
python inference.py \
--api_key <YOUR API KEY> \
--data_dir <DATA DIR> \
--task entity_resolution \
--context_num 3 \
--metadata_wise \
--instance_wise \
--data_parsing \
--prompt_engineering
To run data transformation task and select 3 examples to add to the prompt, using both the adaptive data parsing and prompt engineering modules,
python inference.py \
--api_key <YOUR API KEY> \
--data_dir <DATA DIR> \
--task data_transformation \
--context_num 3 \
--data_parsing \
--prompt_engineering
The data retrieval may take time. When inference, we restore the retrieval scores in the ret_score folder intermediately. And we provide the retrieval scores of examples above for quick verification.