junjiehu / reco-rl Goto Github PK

Could you provide your model's generated stories on the VIST testing dataset?

Thank you for your sharing, but the links of the preprocessed data and pre-trained models are not available. Could you please send me an available link? Thank you very much!

File "/home/ReCo-RL/src/trainer.py", line 58, in _forward
lprobs = self.model(sample['src_seq'], sample['src_lengths'], sample['trg_seq'])
In vist_dataloader.py, there are not 'def forward()' in 'class VistDataLoader'. Can you submit a new code that include the 'forward' method?

compute_relevance_reward
if w not in vocab.word2pos:
TypeError: argument of type 'NoneType' is not iterable

There are some errors because the word2lemma and the word2pos are not assigned. Can you provide a revised code？

Hi, I was looking to evaluate the model on some of the image in the VIST data and other images in my curated dataset as well. Would it be possible for you to share the image pre-processing scripts that were used to
a) Resize / Transform the raw image.
b) Get the Resnet 152 Features.
i.e steps to get the resnet image features from say this raw image -> https://ibb.co/qxjgTtZ

I had experimented with multiple transformation to (256,256) and then getting the features etc , I also tried the AREL image processing script -> https://github.com/eric-xw/AREL/blob/master/scripts/extract_features.py but wasn't able to get the same values. Would really appreciate any pointers.
PS :- Thanks a lot for making the code public.

When i run the script train_rl.sh, there is a error " can't open file 'vist/train_wx_coh.py': [Errno 2] No such file or directory".