xgduan / wsdec Goto Github PK

Just pointing it out in repo requirements section, it should be pytorch 0.3.1, not pytorch 3.1

Thanks for sharing the code!

I ran the training code with CUDA-9.0 under Pytorch-0.3.1-cuda90. But, I still met the bug. Can you tell me which part of the code leads to the bug? I would like to try to address it.

Thanks.

thanks.

Hi,

Thank you for sharing your code in such an organized way and it is really helpful! I was trying to run your code and evaluate the result with the provided third-party scripts. However, I encountered an error during the evaluation and I do not know how to fix it. I am wondering did you encountered the same problem while running the evaluate.py provided by the third-party? And if so, how did you solve it?

Here is the error while I was trying to run the evaluate.py script in the densevid_eval.

Before this, I added 'shell=True' to the subprocess.Popen in both ptbtokenizer.py and meteor.py in order to solve the following two errors:

I am not so sure whether my changes to the two files cause the key error, so I am wondering did you also encountered this. Thanks so much in advance:)

Hello, thank you very much for your sharing. Could you please tell me where to download the translator.pkl that is captioning dict between words and indexes? Looking forward to your reply.

Hi!
First, Thank you for your awesome and kindness github!

I'm on research about caption-to-video generation and I need dataset included pair of video and captions.

So This github show me the hope. But I have some problem

I already ready C3D-500dims features via C3D and PCA(n_samples * 1 * 500). But I don't have any video_length and video_mask because i used the trimmed video.

So how can I run just captioning model with your pretrained model?

Thanks you for reading :)

Hi,
Thanks for providing code in github. I am able to run your code. But unfortunately could not able to generate you reported results. Even I tried using your pretrained model from Onedrive specially for CIDEr, the scores are far away from the reported results. I followed your approach described in readme. Can you please tell me how can I able to reproduce the results?

In the "training" of your readme, it says "python train_script/train_cg_pretrain.py" first. However, there is no file named "train_cg_pretrain.py" in the "train_script" folder. Which file should we run, "train_cg.py" or "train_captionmodel_pretrain.py"?

I have trained supervised video captioning using train_sl.py. For evaluating the resultant .ckp which script should I use?

git clone --recurssive https://github.com/XgDuan/WSDEC

-> git clone --recursive https://github.com/XgDuan/WSDEC

just a little! Thank you :)

Your work seems charming， but the paper has yet to be released.
Would it possible for you to make it available in advance？ Thx~

Hello,

cd data
sh download.sh

Have you released the download.sh?

我的天！这数据是要从外网下吗？

Hi! Thanks for releasing your excellent work.
While reproducing your experiments, I encountered several problems.
1）Is the default --translator_path should be changed from ./data/translator6000.pkl to ./data/translator.pkl to keep consistent with the dictionary generated in captioning_preprocessing.py? BTW, do I need to change the vocab_size accordingly?
2) I did not find a download.sh under the folder ./third_party/densevid_eval as stated in the ReadMe. Would u mind give me a hint about how to get that?
Thanks again for releasing. It's really interesting work.

Hi, thanks for your helpful code. How many hours does the training phase of this model take on your device?