from flashrag.config import Config
from flashrag.pipeline import SequentialPipeline
from flashrag.dataset import Dataset, Item
config = Config('my_config.yaml')
pipeline = SequentialPipeline(config)
# use your own query
my_query = ['Who discover DNA?', 'Who is the president of USA?']
data = [Item({"id": idx, 'question': q, 'golden_answers': []}) for idx, q in enumerate(my_query)]
dataset = Dataset(config = config, data=data)
result = pipeline.run(dataset, do_eval=False)
print(result.pred) # expected output: [answer1, answer2]
目前我们的pipeline仅支持接受dataset,可以通过自己构建dataset的方式直接进行RAG回答。具体可以参考下面的代码:
如果有其他具体的需求,我们可以在后续版本中间进行改进。