Add log and check predicted scores. test=develop (#23506)

python/paddle/fluid/tests/unittests/dygraph_to_static/test_transformer.py

@@ -29,6 +29,7 @@ trainer_count = 1
 place = fluid.CUDAPlace(0) if fluid.is_compiled_with_cuda() else fluid.CPUPlace(
 )
 SEED = 10
+step_num = 10
 
 
 def train_static(args, batch_generator):
@@ -117,7 +118,7 @@ def train_static(args, batch_generator):
             batch_id += 1
             step_idx += 1
             total_batch_num = total_batch_num + 1
-            if step_idx == 10:
+            if step_idx == step_num:
                 if args.save_dygraph_model_path:
                     model_path = os.path.join(args.save_static_model_path,
                                               "transformer")
@@ -201,7 +202,7 @@ def train_dygraph(args, batch_generator):
                         avg_batch_time = time.time()
                 batch_id += 1
                 step_idx += 1
-                if step_idx == 10:
+                if step_idx == step_num:
                     if args.save_dygraph_model_path:
                         model_dir = os.path.join(args.save_dygraph_model_path)
                         if not os.path.exists(model_dir):
@@ -250,10 +251,11 @@ def predict_dygraph(args, batch_generator):
         transformer.eval()
 
         step_idx = 0
+        speed_list = []
         for input_data in test_loader():
             (src_word, src_pos, src_slf_attn_bias, trg_word,
              trg_src_attn_bias) = input_data
-            finished_seq, finished_scores = transformer.beam_search(
+            seq_ids, seq_scores = transformer.beam_search(
                 src_word,
                 src_pos,
                 src_slf_attn_bias,
@@ -263,12 +265,28 @@ def predict_dygraph(args, batch_generator):
                 eos_id=args.eos_idx,
                 beam_size=args.beam_size,
                 max_len=args.max_out_len)
-            finished_seq = finished_seq.numpy()
+            seq_ids = seq_ids.numpy()
-            finished_scores = finished_scores.numpy()
+            seq_scores = seq_scores.numpy()
+            if step_idx % args.print_step == 0:
+                if step_idx == 0:
+                    logging.info(
+                        "Dygraph Predict: step_idx: %d, 1st seq_id: %d, 1st seq_score: %.2f"
+                        % (step_idx, seq_ids[0][0][0], seq_scores[0][0]))
+                    avg_batch_time = time.time()
+                else:
+                    speed = args.print_step / (time.time() - avg_batch_time)
+                    speed_list.append(speed)
+                    logging.info(
+                        "Dygraph Predict: step_idx: %d, 1st seq_id: %d, 1st seq_score: %.2f, speed: %.3f step/s"
+                        % (step_idx, seq_ids[0][0][0], seq_scores[0][0], speed))
+                    avg_batch_time = time.time()
+
             step_idx += 1
-            if step_idx == 10:
+            if step_idx == step_num:
                 break
-        return finished_seq
+        logging.info("Dygraph Predict:  avg_speed: %.4f step/s" %
+                     (np.mean(speed_list)))
+        return seq_ids, seq_scores
 
 
 def predict_static(args, batch_generator):
@@ -318,16 +336,34 @@ def predict_static(args, batch_generator):
     loader.set_batch_generator(batch_generator, places=place)
 
     step_idx = 0
+    speed_list = []
     for feed_dict in loader:
         seq_ids, seq_scores = exe.run(
             test_prog,
             feed=feed_dict,
             fetch_list=[out_ids.name, out_scores.name],
             return_numpy=True)
+        if step_idx % args.print_step == 0:
+            if step_idx == 0:
+                logging.info(
+                    "Static Predict: step_idx: %d, 1st seq_id: %d, 1st seq_score: %.2f,"
+                    % (step_idx, seq_ids[0][0][0], seq_scores[0][0]))
+                avg_batch_time = time.time()
+            else:
+                speed = args.print_step / (time.time() - avg_batch_time)
+                speed_list.append(speed)
+                logging.info(
+                    "Static Predict: step_idx: %d, 1st seq_id: %d, 1st seq_score: %.2f, speed: %.3f step/s"
+                    % (step_idx, seq_ids[0][0][0], seq_scores[0][0], speed))
+                avg_batch_time = time.time()
+
         step_idx += 1
-        if step_idx == 10:
+        if step_idx == step_num:
             break
-    return seq_ids
+    logging.info("Static Predict:  avg_speed: %.4f step/s" %
+                 (np.mean(speed_list)))
+
+    return seq_ids, seq_scores
 
 
 class TestTransformer(unittest.TestCase):
@@ -344,12 +380,17 @@ class TestTransformer(unittest.TestCase):
 
     def _test_predict(self):
         args, batch_generator = self.prepare(mode='test')
-        static_res = predict_static(args, batch_generator)
+        static_seq_ids, static_scores = predict_static(args, batch_generator)
-        dygraph_res = predict_dygraph(args, batch_generator)
+        dygraph_seq_ids, dygraph_scores = predict_dygraph(args, batch_generator)
+
+        self.assertTrue(
+            np.allclose(static_seq_ids, static_seq_ids),
+            msg="static_seq_ids: {} \n dygraph_seq_ids: {}".format(
+                static_seq_ids, dygraph_seq_ids))
         self.assertTrue(
-            np.allclose(static_res, dygraph_res),
+            np.allclose(static_scores, dygraph_scores),
-            msg="static_res: {} \n dygraph_res: {}".format(static_res,
+            msg="static_scores: {} \n dygraph_scores: {}".format(
-                                                           dygraph_res))
+                static_scores, dygraph_scores))
 
     def test_check_result(self):
         self._test_train()