dhruvkhattar / mvae Goto Github PK

As in the requirements.txt, the program need to run in a enivrenment with tensorflow>=1.14.0 and tensorboard>=1.14.0.
But the program requires to from torch.utils.tensorboard import SummaryWriter, which requires tensorboard after 1.15.0.
Maybe we need to check and modify the requirements

Hi. Thank you for your great work. I am trying to reproduce the model for Twitter dataset but there is a missing file named "image_embed.pkl". without it, the reproduction wouldn't be ideal. could you please share this file with us?
Thanks

Can you kindly provide details on these two files which are required for the data handling of the Twitter dataset?

Can you please provide your Image embedding file
img_embed = pkl.load(open('../data/image_embed.pkl'))

On the Weibo dataset, I only got 71 accuracy socre.
I didn't change your model or training parameters.
Cause I don't have image_embed.pkl and xx_content_segmented.txt, I can only preprocess the data according to my own guess.

I am unable to get your dataset used in MVAE model in the correct format that used in your model.
I am also unable to get Weibo Dataset source.
please help me to get the dataset or guide me on how to run your code with Twitter Dataset.

TypeError Traceback (most recent call last)
in ()
1 if name == 'main':
----> 2 train(47, 4096, 64, 0.05, 0.3, 'models/vae_fnd_0.05_0.3')
3 # test(20, 4096, 64, 0.05, 0.3, 'models/vae_fnd_0.05_0.3')
4 # save_features(20, 4096, 64, 0.05, 0.3, '../models/vae_fnd_0.05_0.3')

21 frames
/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/ops.py in _as_graph_element_locked(self, obj, allow_tensor, allow_operation)
3757 # We give up!
3758 raise TypeError("Can not convert a %s into a %s." %
-> 3759 (type(obj).name, types_str))
3760
3761 def get_operations(self):

TypeError: Can not convert a NoneType into a Tensor or Operation.

I tried the code in my datasets and the results are worse than that without VAE. Have you tried the model without VAE?

Since you have mentioned that you have used mediaeval 2016, images of testset like airstrikes_1 url is not mentioned .
How you you split the images in dev and test set

Originally posted by @vidhu22 in #3 (comment)