search

zorazrw / filco

[Preprint] Learning to Filter Context for Retrieval-Augmented Generaton
https://arxiv.org/pdf/2311.08377.pdf
Creative Commons Attribution Share Alike 4.0 International
183 stars 21 forks source link

get_inputs.py on generation task #6

Closed XiaoXiaoYi123 closed 4 months ago

XiaoXiaoYi123 commented 8 months ago

After I have trained the ctx model, I want to get the input to generate the model, so I execute the following script: python get_inputs.py \ --dataset_path "./datasets/nq/scored/train.json" \ --output_path "./datasets/nq/mgen/em/train_em_top1.json" \ --input_list question filtered --output_list answer \ --n_examples 0 --n_contexts 1 \ --filter_criteria strinc --print_example

But I found a problem, I did not use any output results of my trained ctx model here, is this normal? Or am I doing something wrong?

zorazrw commented 7 months ago

Could you explain a bit more on "get the input to generate the model"? Are you trying to "get the input" to the (generation) model, or trying to "generate the model" itself (and which model is this)?