mirror of
https://github.com/explosion/spaCy.git
synced 2024-12-25 17:36:30 +03:00
Fix redundant tokvecs updates during update
This commit is contained in:
parent
12de263813
commit
4363b4aa4a
|
@ -305,14 +305,17 @@ class Language(object):
|
|||
grads[key] = (W, dW)
|
||||
pipes = list(self.pipeline[1:])
|
||||
random.shuffle(pipes)
|
||||
tokvecses, bp_tokvecses = tok2vec.model.begin_update(feats, drop=drop)
|
||||
all_d_tokvecses = [tok2vec.model.ops.allocate(tv.shape) for tv in tokvecses]
|
||||
for proc in pipes:
|
||||
if not hasattr(proc, 'update'):
|
||||
continue
|
||||
tokvecses, bp_tokvecses = tok2vec.model.begin_update(feats, drop=drop)
|
||||
d_tokvecses = proc.update((docs, tokvecses), golds,
|
||||
drop=drop, sgd=get_grads, losses=losses)
|
||||
if update_tensors and d_tokvecses is not None:
|
||||
bp_tokvecses(d_tokvecses, sgd=sgd)
|
||||
for i, d_tv in enumerate(d_tokvecses):
|
||||
all_d_tokvecses[i] += d_tv
|
||||
bp_tokvecses(all_d_tokvecses, sgd=sgd)
|
||||
for key, (W, dW) in grads.items():
|
||||
sgd(W, dW, key=key)
|
||||
# Clear the tensor variable, to free GPU memory.
|
||||
|
|
Loading…
Reference in New Issue
Block a user