Merge pull request #1212 from kuke/fix_sequence_tagging_ner

Use metrics.chunk_evaluator for sequence_taggging_for_ner

fluid/sequence_tagging_for_ner/train.py

@@ -15,17 +15,23 @@ from utils import logger, load_dict
 from utils_extend import to_lodtensor, get_embedding
 
 
-def test(exe, chunk_evaluator, inference_program, test_data, place):
-    chunk_evaluator.reset(exe)
+def test(exe, chunk_evaluator, inference_program, test_data, test_fetch_list,
+         place):
+    chunk_evaluator.reset()
     for data in test_data():
         word = to_lodtensor([x[0] for x in data], place)
         mark = to_lodtensor([x[1] for x in data], place)
         target = to_lodtensor([x[2] for x in data], place)
-        acc = exe.run(inference_program,
-                      feed={"word": word,
-                            "mark": mark,
-                            "target": target})
-    return chunk_evaluator.eval(exe)
+        rets = exe.run(inference_program,
+                       feed={"word": word,
+                             "mark": mark,
+                             "target": target},
+                       fetch_list=test_fetch_list)
+        num_infer = np.array(rets[0])
+        num_label = np.array(rets[1])
+        num_correct = np.array(rets[2])
+        chunk_evaluator.update(num_infer[0], num_label[0], num_correct[0])
+    return chunk_evaluator.eval()
 
 
 def main(train_data_file,
@@ -58,16 +64,16 @@ def main(train_data_file,
     crf_decode = fluid.layers.crf_decoding(
         input=feature_out, param_attr=fluid.ParamAttr(name='crfw'))
 
-    chunk_evaluator = fluid.evaluator.ChunkEvaluator(
-        input=crf_decode,
-        label=target,
-        chunk_scheme="IOB",
-        num_chunk_types=int(math.ceil((label_dict_len - 1) / 2.0)))
+    (precision, recall, f1_score, num_infer_chunks, num_label_chunks,
+     num_correct_chunks) = fluid.layers.chunk_eval(
+         input=crf_decode,
+         label=target,
+         chunk_scheme="IOB",
+         num_chunk_types=int(math.ceil((label_dict_len - 1) / 2.0)))
+    chunk_evaluator = fluid.metrics.ChunkEvaluator()
 
     inference_program = fluid.default_main_program().clone(for_test=True)
-    with fluid.program_guard(inference_program):
-        test_target = chunk_evaluator.metrics + chunk_evaluator.states
-        inference_program = fluid.io.get_inference_program(test_target)
+    test_fetch_list = [num_infer_chunks, num_label_chunks, num_correct_chunks]
 
     if "CE_MODE_X" not in os.environ:
         train_reader = paddle.batch(
@@ -100,26 +106,29 @@ def main(train_data_file,
     embedding_param = fluid.global_scope().find_var(embedding_name).get_tensor()
     embedding_param.set(word_vector_values, place)
 
+    time_begin = time.time()
     for pass_id in six.moves.xrange(num_passes):
-        chunk_evaluator.reset(exe)
+        chunk_evaluator.reset()
         for batch_id, data in enumerate(train_reader()):
-            cost, batch_precision, batch_recall, batch_f1_score = exe.run(
+            cost_var, nums_infer, nums_label, nums_correct = exe.run(
                 fluid.default_main_program(),
                 feed=feeder.feed(data),
-                fetch_list=[avg_cost] + chunk_evaluator.metrics)
+                fetch_list=[
+                    avg_cost, num_infer_chunks, num_label_chunks,
+                    num_correct_chunks
+                ])
             if batch_id % 5 == 0:
-                print(cost)
-                print("Pass " + str(pass_id) + ", Batch " + str(
-                    batch_id) + ", Cost " + str(cost[0]) + ", Precision " + str(
-                        batch_precision[0]) + ", Recall " + str(batch_recall[0])
-                      + ", F1_score" + str(batch_f1_score[0]))
-
-        pass_precision, pass_recall, pass_f1_score = chunk_evaluator.eval(exe)
+                print("Pass " + str(pass_id) + ", Batch " + str(batch_id) +
+                      ", Cost " + str(cost_var[0]))
+            chunk_evaluator.update(nums_infer, nums_label, nums_correct)
+        pass_precision, pass_recall, pass_f1_score = chunk_evaluator.eval()
         print("[TrainSet] pass_id:" + str(pass_id) + " pass_precision:" + str(
             pass_precision) + " pass_recall:" + str(pass_recall) +
               " pass_f1_score:" + str(pass_f1_score))
+
         test_pass_precision, test_pass_recall, test_pass_f1_score = test(
-            exe, chunk_evaluator, inference_program, test_reader, place)
+            exe, chunk_evaluator, inference_program, test_reader,
+            test_fetch_list, place)
         print("[TestSet] pass_id:" + str(pass_id) + " pass_precision:" + str(
             test_pass_precision) + " pass_recall:" + str(test_pass_recall) +
               " pass_f1_score:" + str(test_pass_f1_score))
@@ -128,12 +137,10 @@ def main(train_data_file,
         fluid.io.save_inference_model(save_dirname, ['word', 'mark', 'target'],
                                       crf_decode, exe)
 
-        if ("CE_MODE_X" in os.environ) and (pass_id % 50 == 0):
-            if pass_id > 0:
-                print("kpis	train_precision	%f" % pass_precision)
-                print("kpis	test_precision	%f" % test_pass_precision)
-                print("kpis	train_duration	%f" % (time.time() - time_begin))
-            time_begin = time.time()
+    if "CE_MODE_X" in os.environ:
+        print("kpis	train_precision	%f" % pass_precision)
+        print("kpis	test_precision	%f" % test_pass_precision)
+        print("kpis	train_duration	%f" % (time.time() - time_begin))
 
 
 if __name__ == "__main__":