cfahlgren1/datasets-ai

use correct llama

make model parameters more dynamic w env variables

move duckdb out of gpu function, add prompt lookup decoding

add max duration of 2 minutes and verbose llamacpp

move model inside generate query

set gpu layers to 50

fix pickle issue by using dict instead of pydantic model

add examples and handle incorrect label / data key output better

add spaces gpu decorator

add links and image to readme

use smaller dataset, improve graph

update readme

add plotting capabilities

update readme

add llm for generating sql

create view and get ddl for parquet

add basic querying with duckdb

initial commit

initial commit