TysonYu / Laysumm

The code repository for the paper "Dimsum @LaySumm 20: BART-based Approach for Scientific Document Summarization".
24 stars 6 forks source link

preprocess.py not there in code #4

Open sandeep82945 opened 2 years ago

sandeep82945 commented 2 years ago

The code to create preprocess.py is not given in the repository.

make data loader

python preprocess.py -data_name=abs_intro -mode=train -batch_size=1 python preprocess.py -data_name=abs_intro -mode=valid -batch_size=1

Due to which i am getting error as follows :

"[2022-01-16 11:57:30,933 INFO] Namespace(accumulation_steps=10, beta1=0.9, beta2=0.998, clip=1.0, customiza_model=False, data_name='abs_intro', decay_method='noam', enc_hidden_size=768, epoch=1000, log_file='./logs/', lr=0.05, max_grad_norm=0, minor_data=False, optim='adam', percentage='100', pre_trained_lm='', pre_trained_src=False, process_num=4, random_seed=199744, saving_path='./save/', train_from='', visible_gpu='0', warmup_steps=1000) [2022-01-16 11:57:30,934 INFO] starting to read dataloader Traceback (most recent call last): File "train.py", line 103, in train_loader, valid_loader = load_dataloader(args) File "train.py", line 27, in load_dataloader train_loader = load(train_file_name) File "/content/Laysumm/BART/others/utils.py", line 27, in load loaded = pickle.load(file) ModuleNotFoundError: No module named 'transformers.tokenization_bart'"