hwang1996 / acme Goto Github PK

There is a food_id which I can get ingredients in self.split ,but how can I get the corresponding instructions？

Hello,
I am trying to train the model again and I am not getting the same results mentioned in the paper

I have downloaded the Recipe1M dataset from the offical link.
I have downloaded the following files : train.tar , valid.tar , test.tar and I have extracted them in the folder data and trained the network and I am getting so low results..

Am I doing something wrong or this is the way I should train?

Thanks

could you use a "install.md" to tell us versions of packages(the environment you use)?

Hi, I just install the pytorch=0.4.1, and run python test.py, but find torchwordemb hasn't been installed, so I sunpip install torchwordemb but find Could not find suitable distribution for Requirement.parse('torch>=1.0.0'), can you tell me how to configure the environment?

The test.py can caculate medR and Recall ,but how can I get recipe with a input of image ,is there a demo for use?
THX

Hi, I'm the beginner of the Recipe1M.
And I have some question about Vocabulary.
I find the vocab.txt, vocab.bin, new_word_dict.pkl in arg.py and ingredient_vocab.pkl in build_vocab.py.

The question is :

1. For your code, I didn't see the vocab.txt using in anywhere.
   Also can't find this file in your GitHub.
   Is it same as the vocab.bin?

2. For the new_word_dict.pkl, I don't know how to generate this file.
   And for the ingredient_vocab.pkl, I didn't find where you use in the code.
   Are they the same?

3. I check the im2recipe-Pytorch.
   I find the ingredient of your [partition]_split.pkl and the ingredient of im2recipe-Pytorch is different.
   Also, vocab.bin are different -> the index of ingredient in vocab.bin are different.
   In your code, you use the LMDB file in im2recipe-Pytorch.
   Could the vector of the ingredient assign the wrong vector.

4. Foringr_cap in data_loader.py, it seem to only append the vocab('<start>').
   But in the same file , you use it to get the targets (the index in thevocab of the ingredient )
   Isn't it only have the index of the '<start>'?

Sorry for many problems.
And thank you for reading.

Hi, you mentioned using pytorch v0.5.0.
There doesn't seem to have pytorch v0.5.0: https://pytorch.org/get-started/previous-versions/
May you verify?
Thanks!