xinli2008
commented
2 years ago Maybe i have come up with a bad solution.when the length of prefix is large ,the program will go wrong.As shown below: 90 [295, 323, 496, 569, 12, 4830, 323, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290] and 79 [606, 1040, 485, 481, 6545, 488, 886, 622, 8353, 1011, 295, 323, 496, 569, 12, 4830, 323, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290, 290] so i change the code: if len(prefix)>17: xxxxxxxxx; Although this method is successful and the program is training, it is not a perfect solution.And why the number is 290? If you have better solution,try to answer me .At last , i am so sorry to bother you ,hope you have a nice day!
Thanks for your code and answers for the previous question.I am so sorry to bother you for my new question.I followed the instruction in the ReadME file.when i ran bash run_emorl.sh , and my rl_method is comet, i came across the next question :
Traceback (most recent call last): File "train_rl.py", line 846, in
tf.app.run()
File "/home/lixin/enter/envs/lx03/lib/python3.6/site-packages/tensorflow/python/platform/app.py", line 40, in run
_run(main=main, argv=argv, flags_parser=_parse_flags_tolerate_undef)
File "/home/lixin/enter/envs/lx03/lib/python3.6/site-packages/absl/app.py", line 312, in run
_run_main(main, args)
File "/home/lixin/enter/envs/lx03/lib/python3.6/site-packages/absl/app.py", line 258, in _run_main
sys.exit(main(argv))
File "train_rl.py", line 835, in main
_train_epoch(sess, epoch==0)
File "train_rl.py", line 562, in _train_epoch
reward_base = get_reward(predictor, story_base, rets_data['batch']['unique_id'], train_arc_file, method=FLAGS.rl_method)
File "/home/lixin/Emo-Aware-Storytelling-master/Reinforcement/rewards_v2.py", line 281, in get_reward
comet_prediction = get_comet_prediction(gen_result)
File "/home/lixin/Emo-Aware-Storytelling-master/comet_generate.py", line 61, in get_comet_prediction
input_event, model, sampler, data_loader, text_encoder, category)
File "comet-commonsense/src/interactive/functions.py", line 124, in get_atomic_sequence
input_event, category, data_loader, text_encoder)
File "comet-commonsense/src/interactive/functions.py", line 158, in set_atomic_inputs
XMB[:, :len(prefix)] = torch.LongTensor(prefix)
RuntimeError: The expanded size of the tensor (18) must match the existing size (82) at non-singleton dimension 1. Target sizes: [1, 18]. Tensor sizes: [82]
I try to solve the question by using try,except,else,but it does not work well.So i guess whether you change the commonsense transformer in some place , if not ,have you ever encountered the above question?If you are busy now,do you own things,if you have free time ,try to answer me .Meantime,I will try to work it out.
Tips:Sometimes this problem appears early, sometimes it appears late。