search

snumprlab / cl-alfred

Official Implementation of CL-ALFRED (ICLR'24)
https://bhkim94.github.io/projects/CL-ALFRED/
GNU General Public License v3.0
12 stars 2 forks source link

Missing tests_seen and unseen folders in dataset #2

Closed JACK-Chen-2019 closed 4 months ago

JACK-Chen-2019 commented 5 months ago

I have met the same problem as seen in https://github.com/gistvision/moca/issues/7

JACK-Chen-2019 commented 5 months ago

FileNotFoundError: [Errno 2] No such file or directory: '/json_feat_2.1.0/train/look_at_obj_in_light-Box-None-FloorLamp-212/trial_T20190908_193427_340509/traj_data.json'

However, in the downloaded directory, there is no traj_data.json file, but three ann.json files instead: /json_feat_2.1.0/look_at_obj_in_light-Box-None-FloorLamp-212/trial_T20190908_193427_340509/pp/ ann_0.json ann_1.json ann_2.json

bhkim94 commented 4 months ago

Hi @JACK-Chen-2019,

Thanks for having an interest in our work!

Can you share your command to reproduce the error? This is quite unexpected because our code is expected to use ann_x.json, not traj_data.json.

Thanks

JACK-Chen-2019 commented 4 months ago

Thanks for your replay. I don't have the 'pp.vocab' file and it needs to run with --preprocess. And it will run dataset.preprocess_splits(splits) in models/train/train_seq2seq.py line 152. And then it will run json_path = os.path.join(self.args.data, k, task['task'], 'traj_data.json').

bhkim94 commented 4 months ago

I see. I've just added our pp.vocab in the Huggingface repo. Can you download it and put it in the path data/json_2.1.0? As we provide pp.vocab, the code should run without preprocess.