About the word2vector

[kikyou123]

Can you tell me how you get the semantic embedding vector for the class label?

[vmichals]

The relevant code should be here.

The embedding matrix is loaded from the npz file and then the label embeddings are retrieved using tf.nn.embedding_lookup()

[Surflyan]

hi, where can I find the embedding matrix (.npz) file, and I what to know if you trained the w2v mdoel by yourself? If not , how do you solve the 'two words' noun，like ‘Tobacco shop‘.

[zhangyikaii]

Hi, @vmichals The .npz file you mentioned will be created here.
BUT I don't know what is the embedding file path here, so the problem is still unresolved.

[spectralpublic]

Hi, @zhangyikaii
I have the same question as you, what is the embedding file path
--emb_addr', type=str, default=os.path.join(os.sep, 'mnt', 'datasets', 'public', 'mini-imagenet', 'raw-data'),

A file like 'mini_imagenet_class_label_dict3.txt' should exist.
Have you solved it? Thank You！