search

nateraw / Lda2vec-Tensorflow

Tensorflow 1.5 implementation of Chris Moody's Lda2vec, adapted from @meereeum
MIT License
107 stars 40 forks source link

TypeError: __init__() got an unexpected keyword argument 'load_embeds' #33

Closed gveni closed 5 years ago

gveni commented 5 years ago

Now I am trying to run the actual lda2vec process, I get the following error:

freqs=freqs) # Python list of shape (vocab_size,). Frequencies of each token, same order as embed matrix mappings. TypeError: init() got an unexpected keyword argument 'load_embeds'

Any help is appreciated. Thanks!

dbl001 commented 5 years ago

I got that problem also.

On Mar 7, 2019, at 5:28 PM, gveni notifications@github.com wrote:

Now I am trying to run the actual lda2vec process, I get the following error:

freqs=freqs) # Python list of shape (vocab_size,). Frequencies of each token, same order as embed matrix mappings. TypeError: init() got an unexpected keyword argument 'load_embeds'

Any help is appreciated. Thanks!

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

nateraw commented 5 years ago

Whoops, thought I replied yesterday! That's what I get for being on mobile. The arguments changed in that function 2 days ago (which is one of the reasons I added a warning in the readme). load_embeds is no longer a parameter of the init function.

There were 2 redundant parameters previously: load_embeds and pretrained_embeddings. These have now turned into a single parameter w_in which is None by default. If you pass your numpy array of pretrained embeddings to the w_in parameter, it will do the same thing as having load_embeds=True, pretrained_embeddings=embedding_matrix.

nateraw commented 5 years ago

Also, I'm open to changing the w_in parameter name to pretrained_embeddings if that's less confusing. What do you think @dbl001 ?

dbl001 commented 5 years ago

Sounds good!

On Mar 8, 2019, at 9:26 AM, Nathan Raw notifications@github.com wrote:

Also, I'm open to changing the w_in parameter name to pretrained_embeddings if that's less confusing. What do you think @dbl001 https://github.com/dbl001 ?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/nateraw/Lda2vec-Tensorflow/issues/33#issuecomment-471008916, or mute the thread https://github.com/notifications/unsubscribe-auth/AC9i21ujqKdODSeSn4pbLRdvr6nGE828ks5vUp1XgaJpZM4bkdNX.

nateraw commented 5 years ago

Cool! I'll be updating more stuff tonight. I'll try to get that readme updated too. I'll leave this issue open until then. Thanks guys!

gveni commented 5 years ago

Thanks @nateraw. For now, should I replace load_embeds and pretrained_embeddings parameters of model function with w_in = None to resolve this error?

dbl001 commented 5 years ago

Or this:

Load embedding matrix from file path

embedding_matrix = P.load_glove("glove.6B.300d.txt")

m = model(num_docs, vocab_size, num_topics=num_topics, embedding_size=embed_size, restore=False, w_in=embedding_matrix, freqs=freqs)

On Mar 8, 2019, at 9:50 AM, gveni notifications@github.com wrote:

Thanks @nateraw https://github.com/nateraw. For now, should I replace load_embeds and pretrained_embeddings parameters of model function with w_in = None?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/nateraw/Lda2vec-Tensorflow/issues/33#issuecomment-471016520, or mute the thread https://github.com/notifications/unsubscribe-auth/AC9i2wwVd2Liy1Viv_H826dqdYz90RyWks5vUqLvgaJpZM4bkdNX.

nateraw commented 5 years ago

@gveni what @dbl001 just commented is correct!

gveni commented 5 years ago

Let me download the embedding matrix and see how this goes. Thanks @dbl001 and @nateraw.

dbl001 commented 5 years ago

FYI

On Mar 8, 2019, at 10:03 AM, gveni notifications@github.com wrote:

Let me download the embedding matrix and see how this goes. Thanks @dbl001 https://github.com/dbl001 and @nateraw https://github.com/nateraw.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/nateraw/Lda2vec-Tensorflow/issues/33#issuecomment-471020640, or mute the thread https://github.com/notifications/unsubscribe-auth/AC9i25L80Tb6QehWzQ6GXqvHRHERKmKvks5vUqYGgaJpZM4bkdNX.

nateraw commented 5 years ago

Links to download embeds should be in readme (but they aren't). I'll add an issue.

gveni commented 5 years ago

FYI, you meant 'glove.6B.50d.txt' and not 'glove.6B.300d.txt'. Otherwise it throws ValueError. Thanks

nateraw commented 5 years ago

Actually it depends which glove embeddings you downloaded. I have the 300 dimension ones, you must be using the 50 dimension version.

nateraw commented 5 years ago

When you download the glove.6B zip file, you get the 50d, 100d, 200d, and 300d versions.