nadbordrozd / blog_stuff Goto Github PK

Hi Nadbor Drozd,

First of all: thank you for sharing this code, it is helping me a lot!

I was wondering if you could clarify to following code in the tranform functions of the MeanEmbeddingVectorizer and TfidfEmbeddingVectorizer classes:

[self.word2vec[w] * self.word2weight[w] for w in words if w in self.word2vec] or [np.zeros(self.dim)]

More specifically, I don't understand what the or [np.zeros(self.dim)] does.

I realize this is probably a question regarding general python syntax, but I can't seem to find an explanation in the python docs or on Stackoverflow.

Thanks again for the code!

why using the glove_small when defining the vectorizers? Is here

    self.dim=len(word2vec)

not enough? I don't get the next/iter and glove_small part...

hi, when I running you code ,some errors shows in page,it looks like this:


so , can you tell me how to fix it? thank you.

Does benchmarking_python3.ipynb have a rewritten version of GPU？

Hi,

Thank you for your work on the PGB code and your "Hello World" blogpost, those helped me a lot!
I am embedding a small graph, but while programming the PGB method I read through the open source project of PGB by facebookresearch, where they state;
“PBG is not optimized for small graphs” and “PyTorch-BigGraph ... may not produce high-quality embeddings for small graphs without careful tuning.”. I managed to implement the code and retrieve embeddings for the nodes in my graph, but based on their comments on small graphs I am unsure whether I have produced high-quality embeddings.

Do you have any tips for choosing the parameters for small graphs, or should I just use the parameter settings you proposed in the blogpost?

Thanks in advance!

Is multi_multi_kernel_nb.py missing the transform method?

Hi, thanks for your code in classification_w2v/benchmarking_python3.ipynb, it helps me a lot.
Now, there is some advice for it because of the new edition of sklearn.
my python version is 3.5.6 and sklearn version is 0.20.0

the first one:
now in code

from sklearn.cross_validation import cross_val_score
from sklearn.cross_validation import StratifiedShuffleSplit

suggest

from sklearn.model_selection import cross_val_score
from sklearn.model_selection import StratifiedShuffleSplit

the second one:
now in code

for train, test in StratifiedShuffleSplit(y, n_iter=5, test_size=test_size):

suggest

for train, test in StratifiedShuffleSplit(n_splits=5, test_size=test_size).split(X, y):

besides, the following code might move to %%bash.

wget http://nlp.stanford.edu/data/glove.6B.zip
wget http://nlp.stanford.edu/data/glove.840B.300d.zip