Fixed report breaking

neulab · Sep 27, 2018 · d72529b · d72529b
1 parent b2db445
commit d72529b
Show file tree

Hide file tree

Showing 2 changed files with 37 additions and 6 deletions.
diff --git a/test/config/seg_report.yaml b/test/config/seg_report.yaml
@@ -21,7 +21,23 @@ report: !Experiment
     loss_calculator: !FeedbackLoss
       child_loss: !MLELoss {}
       repeat: 2
+    dev_tasks:
+      - !LossEvalTask
+        src_file: examples/data/head.ja
+        ref_file: examples/data/head.en
+      - !AccuracyEvalTask
+        eval_metrics: bleu,wer
+        src_file: examples/data/dev.ja
+        ref_file: examples/data/dev.en
+        hyp_file: test/tmp/{EXP}.test_hyp
+        inference: !AutoRegressiveInference
+          reporter: !SegmentationReporter
+            report_path: examples/output/test.segment
+
   evaluate:
+    - !LossEvalTask
+      src_file: examples/data/head.ja
+      ref_file: examples/data/head.en
     - !AccuracyEvalTask
       eval_metrics: bleu,wer
       src_file: examples/data/dev.ja
@@ -30,7 +46,13 @@ report: !Experiment
       inference: !AutoRegressiveInference
         reporter: !SegmentationReporter
           report_path: examples/output/test.segment
-    - !LossEvalTask
-      src_file: examples/data/head.ja
-      ref_file: examples/data/head.en
+    - !AccuracyEvalTask
+      eval_metrics: bleu,wer
+      src_file: examples/data/dev.ja
+      ref_file: examples/data/dev.en
+      hyp_file: test/tmp/{EXP}.test_hyp
+      inference: !AutoRegressiveInference
+        reporter: !SegmentationReporter
+          report_path: examples/output/test.segment
+
 
diff --git a/xnmt/specialized_encoders/segmenting_encoder/segmenting_encoder.py b/xnmt/specialized_encoders/segmenting_encoder/segmenting_encoder.py
@@ -68,6 +68,14 @@ def __init__(self, embed_encoder=bare(IdentitySeqTransducer),
     self.compute_report = compute_report
     self.reporter = reporter
     self.no_char_embed = issubclass(segment_composer.__class__, VocabBasedComposer)
+    # Others
+    self.segmenting_action = None
+    self.compose_output = None
+    self.segment_actions = None
+    self.seg_size_unpadded = None
+    self.src_sent = None
+    self.reward = None
+    self.train = None
 
   def shared_params(self):
     return [{".embed_encoder.hidden_dim",".policy_learning.policy_network.input_dim"},
@@ -106,7 +114,8 @@ def transduce(self, embed_sent: ExpressionSequence) -> List[ExpressionSequence]:
       self.compose_output = outputs
       self.segment_actions = actions
       if not self.train and self.compute_report:
-        self.report_sent_info({"segment_actions": actions})
+        if len(actions) == 1: # Support only AccuracyEvalTask
+          self.report_sent_info({"segment_actions": actions})
 
   @handle_xnmt_event
   def on_calc_additional_loss(self, trg, generator, generator_loss):
@@ -232,14 +241,14 @@ def sparse_to_dense(self, actions, length):
   def pad(self, outputs):
     # Padding
     max_col = max(len(xs) for xs in outputs)
-    P0 = dy.vecInput(outputs[0][0].dim()[0][0])
+    p0 = dy.vecInput(outputs[0][0].dim()[0][0])
     masks = np.zeros((len(outputs), max_col), dtype=int)
     modified = False
     ret = []
     for xs, mask in zip(outputs, masks):
       deficit = max_col - len(xs)
       if deficit > 0:
-        xs.extend([P0 for _ in range(deficit)])
+        xs.extend([p0 for _ in range(deficit)])
         mask[-deficit:] = 1
         modified = True
       ret.append(dy.concatenate_cols(xs))