Deserialization fails based on whether "nlp" object was used yet

When I try to deserialize using a fresh nlp object, spaCy crashes, though it manages to deserialize fine if it’s the same nlp object that was used to originally parse the text. I can’t tell if I’m using the library in a way it’s not intended to be used, or if this is a bug? (This is spaCy version 1.5.0 on Linux with Python 2.7)

from spacy.tokens.doc import Doc
import spacy
nlp = spacy.load("en")
text = u"Hello world."

# Parse it. Works.
doc = nlp(text)
print len(set([o for o in doc]))

# Save a serialized copy
with open("out",'w') as f:
    f.write(str(doc.to_bytes()))

# Deserialize: works
doc2 = Doc(nlp.vocab)
doc2.from_bytes(open("out").read())
print len(set([o for o in doc2]))

# Deserialize with a fresh nlp object: crashes
nlp = spacy.load("en")
doc3 = Doc(nlp.vocab)
doc3.from_bytes(open("out").read())
print len(set([o for o in doc3]))

[brendano]

Oh now it works! I reinstalled the model and now it has both these two directories in spacy/data:

en-1.1.0
en_glove_cc_300_1m_vectors-1.0.0

Before, it only had en-1.1.0 after I installed the model.

[lock[bot]]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.