Rooba
/
spaCy
mirror of https://github.com/explosion/spaCy.git
1
1
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Add mix weights on fine_tune

This commit is contained in:
Matthew Honnibal 2017-08-07 06:32:59 -05:00
parent 42bd26f6f3
commit 5d837c3776
1 changed files with 10 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
spacy/_ml.py
View File

@ -356,17 +356,24 @@ def fine_tune(embedding, combine=None):
lengths = model.ops.asarray([len(doc) for doc in docs], dtype='i') lengths = model.ops.asarray([len(doc) for doc in docs], dtype='i')
vecs, bp_vecs = embedding.begin_update(docs, drop=drop) vecs, bp_vecs = embedding.begin_update(docs, drop=drop)
flat_tokvecs = embedding.ops.flatten(tokvecs)
flat_vecs = embedding.ops.flatten(vecs)
output = embedding.ops.unflatten( output = embedding.ops.unflatten(
embedding.ops.flatten(tokvecs) (model.mix[0] * flat_vecs + model.mix[1] * flat_tokvecs),
+ embedding.ops.flatten(vecs),
lengths) lengths)
def fine_tune_bwd(d_output, sgd=None): def fine_tune_bwd(d_output, sgd=None):
bp_vecs(d_output, sgd=sgd) bp_vecs(d_output, sgd=sgd)
flat_grad = model.ops.flatten(d_output)
model.d_mix[1] += flat_tokvecs.dot(flat_grad.T).sum()
model.d_mix[0] += flat_vecs.dot(flat_grad.T).sum()
sgd(model._mem.weights, model._mem.gradient, key=model.id)
return d_output return d_output
return output, fine_tune_bwd return output, fine_tune_bwd
model = wrap(fine_tune_fwd, embedding) model = wrap(fine_tune_fwd, embedding)
model.mix = model._mem.add((model.id, 'mix'), (2,))
model.mix.fill(1.)
model.d_mix = model._mem.add_gradient((model.id, 'd_mix'), (model.id, 'mix'))
return model return model