#!/usr/bin/env python
# coding: utf-8

# In[46]:


from gensim.models import KeyedVectors


# We will load only first 1000 (top 1000) vectors from python fasttext (128) model

# In[47]:


model = KeyedVectors.load_word2vec_format("../model.vec", binary=False, limit=1000)


# In case you are loading GloVe embeddings, you need to convert it first

# In[ ]:


tmpfile=get_tmpfile("source2vec")
glove2word2vec(datapath("../glove_model.txt"), tmpfile)
model = KeyedVectors.load_word2vec_format(tmpfile, binary=False, limit=1000)


# Now we can do fancy staff

# In[48]:


model.most_similar("for")


# In[49]:


model.most_similar_cosmul("for")


# In[50]:


model.doesnt_match(["for", "i", "a"])


# In[51]:


model.similarity("for", "i")


# In[52]:


model.similar_by_word("for")


# Raw vector values

# In[53]:


for word in model.vocab:
    print(word, ":", model[word])
    break # only first one now