A module for reading, writing, and using trained word embeddings.
Install with pip!
pip install pyemblib
This package currently supports word embeddings trained by the following packages:
Both text-format and binary embedding files are supported.
The example below shows reading each format of embedding:
## import text embeddings
text_embs = pyemblib.read('/tmp/text_embeddings.txt', mode=pyemblib.Mode.Text)
## import binary embeddings
bin_embs = pyemblib.read('/tmp/bin_embeddings.bin', mode=pyemblib.Mode.Binary)
Embeddings are read as a pyemblib.Embeddings
object, which inherits from Python's dictionary class; keys are words, and values are the embedding arrays.
To get the word vector for "python", just use dictionary access:
vec = embs['python']
print(vec)
# [ 0.001 -0.237 ... ]
The same text and binary modes can be used for writing out embedding files as for reading.
embs = { 'a' : np.array([0.3 0.1 -0.2]), 'b' : np.array([-0.9, -0.2, -0.2]) }
## write as text
pyemblib.write(embs, '/tmp/text_embeddings.txt', mode=pyemblib.Mode.Text)
## write as binary
pyemblib.write(embs, '/tmp/bin_embeddings.bin', mode=pyemblib.Mode.Binary)
Please report any issues you encounter to the Github Issues page!