search

DeepLearnXMU / seq2seqset

implementaion of our ACL Findings 2023 paper "A Sequence-to-SequenceSet Model for Text-to-Table Generation"
1 stars 0 forks source link

Requriements for more details. #1

Open jiangpw41 opened 5 months ago

jiangpw41 commented 5 months ago

Hello, I am trying to reproduce your paper. I know this is a sequel to the paper from 2022, and I have basically completed the reproduction of that paper. But due to your project introduction being too simplistic, I tried to train but failed. I would greatly appreciate it if you could provide some more detailed project operation guidelines. How to prepare data? It was mentioned that your experiments used the processed datasets from (Wu et al., 2022) in your paper, but your training scripts use "data/rotowire_player_full" and "data/rotowire_team_full" which are not the same as the former paper which only has "data/rotowire". Besides, I find that in your "set/tasks/seq2set3.py" file there is a line of code tgt_dataset_sizes = torch.load(os.path.join(data_path, "{}.{}-{}.target_size".format(split, src, tgt))) if tgt_dataset is not None else None which refers to a file called "{}.{}-{}.target_size", which also not exist in the former paper projections. I would appreciate it if you could tell me how to preprocess the data to fit your projection!!

z20021009 commented 4 months ago

复现成功了吗

z20021009 commented 4 months ago

如果可以加个联系方式聊一下 复现遇到许多问题

jiangpw41 commented 4 months ago

如果可以加个联系方式聊一下 复现遇到许多问题

没有，但可以给我的邮箱发联系方式wayneroaming@sjtu.edu.cn