explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2025-07-19 20:52:23 +03:00
Code Issues Packages Projects Releases Wiki Activity

Move placeholder handling into model code

Browse Source
This commit is contained in:
Paul O'Leary McCann 2021-07-03 18:38:48 +09:00
parent 3f66e18592
commit f2e0e9dc28
2 changed files with 15 additions and 18 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
spacy/ml/models/coref.py
View File

@ -387,7 +387,13 @@ def ant_scorer_forward(
scores = pw_prod + pw_sum + mask scores = pw_prod + pw_sum + mask
top_scores, top_scores_idx = topk(xp, scores, min(ant_limit, len(scores))) top_limit = min(ant_limit, len(scores))
top_scores, top_scores_idx = topk(xp, scores, top_limit)
# now add the placeholder
placeholder = ops.alloc2f(scores.shape[0], 1)
top_scores = xp.concatenate( (placeholder, top_scores), 1)
top_scores = ops.softmax(top_scores, axis=1)
out.append((top_scores, top_scores_idx)) out.append((top_scores, top_scores_idx))
# In the full model these scores can be further refined. In the current # In the full model these scores can be further refined. In the current
@ -414,6 +420,8 @@ def ant_scorer_forward(
offset = 0 offset = 0
for dy, (prod_back, pw_sum_back), ll in zip(dYscores, backprops, veclens): for dy, (prod_back, pw_sum_back), ll in zip(dYscores, backprops, veclens):
dyscore, dyidx = dy dyscore, dyidx = dy
# remove the placeholder
dyscore = dyscore[:, 1:]
# the full score grid is square # the full score grid is square
fullscore = ops.alloc2f(ll, ll) fullscore = ops.alloc2f(ll, ll)

23
spacy/pipeline/coref.py
View File

@ -142,10 +142,6 @@ class CoreferenceResolver(TrainablePipe):
starts = idxs[offset:hi, 0] starts = idxs[offset:hi, 0]
ends = idxs[offset:hi, 1] ends = idxs[offset:hi, 1]
# need to add the placeholder
placeholder = self.model.ops.alloc2f(cscores.shape[0], 1)
cscores = xp.concatenate((placeholder, cscores), 1)
predicted = get_predicted_clusters(xp, starts, ends, ant_idxs, cscores) predicted = get_predicted_clusters(xp, starts, ends, ant_idxs, cscores)
clusters_by_doc.append(predicted) clusters_by_doc.append(predicted)
return clusters_by_doc return clusters_by_doc
@ -291,9 +287,8 @@ class CoreferenceResolver(TrainablePipe):
offset = 0 offset = 0
gradients = [] gradients = []
loss = 0 total_loss = 0
for example, (cscores, cidx) in zip(examples, score_matrix): for example, (cscores, cidx) in zip(examples, score_matrix):
# assume cids has absolute mention ids
ll = cscores.shape[0] ll = cscores.shape[0]
hi = offset + ll hi = offset + ll
@ -310,20 +305,14 @@ class CoreferenceResolver(TrainablePipe):
# boolean to float # boolean to float
top_gscores = ops.asarray2f(top_gscores) top_gscores = ops.asarray2f(top_gscores)
# add the placeholder to cscores grad, loss = self.loss(cscores.T, top_gscores.T)
placeholder = self.model.ops.alloc2f(ll, 1)
cscores = xp.concatenate((placeholder, cscores), 1)
# do softmax to cscores gradients.append((grad.T, cidx))
cscores = ops.softmax(cscores, axis=1) total_loss += float(loss)
diff = self.loss.get_grad(cscores.T, top_gscores.T).T offset = hi
diff = diff[:, 1:]
gradients.append((diff, cidx))
loss += float(self.loss.get_loss(cscores.T, top_gscores.T)) return total_loss, gradients
offset += ll
return loss, gradients
def initialize( def initialize(
self, self,