test=develop (#2314)

PaddleNLP/dialogue_model_toolkit/auto_dialogue_evaluation/evaluation.py

@@ -6,15 +6,16 @@ import sys
 import numpy as np
 import pandas as pd
 
+
 def get_p_at_n_in_m(data, n, m, ind):
     """
     Get n in m
     """
     pos_score = data[ind][0]
-    curr = data[ind : ind+m]
-    curr = sorted(curr, key = lambda x: x[0], reverse=True)
+    curr = data[ind:ind + m]
+    curr = sorted(curr, key=lambda x: x[0], reverse=True)
 
-    if curr[n-1][0] <= pos_score:
+    if curr[n - 1][0] <= pos_score:
         return 1
     return 0
 
@@ -28,9 +29,10 @@ def evaluate_Recall(data):
     p_at_2_in_10 = 0.0
     p_at_5_in_10 = 0.0
 
-    length = len(data) / 10
+    length = len(data) // 10
+    print('length=%s' % length)
 
-    for i in xrange(0, length):
+    for i in range(0, length):
         ind = i * 10
         assert data[ind][1] == 1
 
@@ -43,7 +45,8 @@ def evaluate_Recall(data):
         '1_in_2': p_at_1_in_2 / length,
         '1_in_10': p_at_1_in_10 / length,
         '2_in_10': p_at_2_in_10 / length,
-        '5_in_10': p_at_5_in_10 / length}
+        '5_in_10': p_at_5_in_10 / length
+    }
 
     return recall_dict
 

PaddleNLP/dialogue_model_toolkit/auto_dialogue_evaluation/init.py

@@ -26,9 +26,8 @@ import copy
 import numpy as np
 import paddle.fluid as fluid
 
-def init_pretraining_params(exe,
-                            pretraining_params_path,
-                            main_program):
+
+def init_pretraining_params(exe, pretraining_params_path, main_program):
     """
     Init pretraining params
     """
@@ -50,4 +49,3 @@ def init_pretraining_params(exe,
         predicate=existed_params)
     print("Load pretraining parameters from {}.".format(
         pretraining_params_path))
-

PaddleNLP/dialogue_model_toolkit/auto_dialogue_evaluation/main.py

@@ -23,6 +23,7 @@ sys.path.append('../../models/dialogue_model_toolkit/auto_dialogue_evaluation/')
 from net import Network
 import config
 
+
 def train(args):
     """Train
     """
@@ -45,8 +46,7 @@ def train(args):
             fluid.clip.set_gradient_clip(clip=fluid.clip.GradientClipByValue(
                 max=1.0, min=-1.0))
 
-            optimizer = fluid.optimizer.Adam(
-                learning_rate=args.learning_rate)
+            optimizer = fluid.optimizer.Adam(learning_rate=args.learning_rate)
             optimizer.minimize(loss)
             print("begin memory optimization ...")
             fluid.memory_optimize(train_program)
@@ -73,8 +73,7 @@ def train(args):
 
     print("device count %d" % dev_count)
     print("theoretical memory usage: ")
-    print(
-        fluid.contrib.memory_usage(
+    print(fluid.contrib.memory_usage(
         program=train_program, batch_size=args.batch_size))
 
     exe = fluid.Executor(place)
@@ -127,15 +126,15 @@ def train(args):
         """
         Evaluate to choose model
         """
-        val_batches = reader.batch_reader(
-            args.val_path, args.batch_size, place, args.max_len, 1)
+        val_batches = reader.batch_reader(args.val_path, args.batch_size, place,
+                                          args.max_len, 1)
         scores = []
         labels = []
         for batch in val_batches:
             scores.extend(test_with_feed(batch))
             labels.extend([x[0] for x in batch[2]])
 
-        return eva.evaluate_Recall(zip(scores, labels))
+        return eva.evaluate_Recall(list(zip(scores, labels)))
 
     def save_exe(step, best_recall):
         """
@@ -143,17 +142,21 @@ def train(args):
         """
         recall_dict = evaluate()
         print('evaluation recall result:')
-        print('1_in_2: %s\t1_in_10: %s\t2_in_10: %s\t5_in_10: %s' % (
-            recall_dict['1_in_2'], recall_dict['1_in_10'], 
+        print('1_in_2: %s\t1_in_10: %s\t2_in_10: %s\t5_in_10: %s' %
+              (recall_dict['1_in_2'], recall_dict['1_in_10'],
                recall_dict['2_in_10'], recall_dict['5_in_10']))
 
         if recall_dict['1_in_10'] > best_recall and step != 0:
-            fluid.io.save_inference_model(args.save_path, 
+            fluid.io.save_inference_model(
+                args.save_path,
                 net.get_feed_inference_names(),
-                logits, exe, main_program=train_program)
+                logits,
+                exe,
+                main_program=train_program)
 
             print("Save model at step %d ... " % step)
-            print(time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(time.time())))
+            print(time.strftime('%Y-%m-%d %H:%M:%S',
+                                time.localtime(time.time())))
             best_recall = recall_dict['1_in_10']
         return best_recall
 
@@ -161,15 +164,16 @@ def train(args):
     global_step, train_time = 0, 0.0
     best_recall = 0
     for epoch in six.moves.xrange(args.num_scan_data):
-        train_batches = reader.batch_reader(
-            args.train_path, args.batch_size, place, 
-            args.max_len, args.sample_pro)
+        train_batches = reader.batch_reader(args.train_path, args.batch_size,
+                                            place, args.max_len,
+                                            args.sample_pro)
 
         begin_time = time.time()
         sum_cost = 0
         ce_cost = 0
         for batch in train_batches:
-            if (args.save_path is not None) and (global_step % args.save_step == 0):
+            if (args.save_path is not None) and (
+                    global_step % args.save_step == 0):
                 best_recall = save_exe(global_step, best_recall)
 
             cost = train_with_feed(batch)
@@ -178,7 +182,8 @@ def train(args):
             ce_cost = cost.mean()
 
             if global_step % args.print_step == 0:
-                print('training step %s avg loss %s' % (global_step, sum_cost / args.print_step))
+                print('training step %s avg loss %s' %
+                      (global_step, sum_cost / args.print_step))
                 sum_cost = 0
 
         pass_time_cost = time.time() - begin_time
@@ -187,7 +192,8 @@ def train(args):
               .format(epoch, "%2.2f sec" % pass_time_cost))
         if "CE_MODE_X" in os.environ and epoch == args.num_scan_data - 1:
             card_num = get_cards()
-            print("kpis\ttrain_duration_card%s\t%s" % (card_num, pass_time_cost))
+            print("kpis\ttrain_duration_card%s\t%s" %
+                  (card_num, pass_time_cost))
             print("kpis\ttrain_loss_card%s\t%s" % (card_num, ce_cost))
 
 
@@ -246,8 +252,7 @@ def finetune(args):
 
     print("device count %d" % dev_count)
     print("theoretical memory usage: ")
-    print(
-        fluid.contrib.memory_usage(
+    print(fluid.contrib.memory_usage(
         program=train_program, batch_size=args.batch_size))
 
     exe = fluid.Executor(place)
@@ -264,9 +269,7 @@ def finetune(args):
 
     if args.init_model:
         init.init_pretraining_params(
-            exe,
-            args.init_model,
-            main_program=train_startup)
+            exe, args.init_model, main_program=train_startup)
         print('sccuess init %s' % args.init_model)
 
     print("start loading data ...")
@@ -294,8 +297,8 @@ def finetune(args):
         """
         Evaluate to choose model
         """
-        val_batches = reader.batch_reader(
-            args.val_path, args.batch_size, place, args.max_len, 1)
+        val_batches = reader.batch_reader(args.val_path, args.batch_size, place,
+                                          args.max_len, 1)
         scores = []
         labels = []
         for batch in val_batches:
@@ -311,11 +314,15 @@ def finetune(args):
         cor = evaluate()
         print('evaluation cor relevance %s' % cor)
         if cor > best_cor and step != 0:
-            fluid.io.save_inference_model(args.save_path, 
-                net.get_feed_inference_names(), logits, 
-                exe, main_program=train_program)
+            fluid.io.save_inference_model(
+                args.save_path,
+                net.get_feed_inference_names(),
+                logits,
+                exe,
+                main_program=train_program)
             print("Save model at step %d ... " % step)
-            print(time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(time.time())))
+            print(time.strftime('%Y-%m-%d %H:%M:%S',
+                                time.localtime(time.time())))
             best_cor = cor
         return best_cor
 
@@ -324,15 +331,15 @@ def finetune(args):
     best_cor = 0.0
     pre_index = -1
     for epoch in six.moves.xrange(args.num_scan_data):
-        train_batches = reader.batch_reader(
-            args.train_path, 
-            args.batch_size, place, 
-            args.max_len, args.sample_pro)
+        train_batches = reader.batch_reader(args.train_path, args.batch_size,
+                                            place, args.max_len,
+                                            args.sample_pro)
 
         begin_time = time.time()
         sum_cost = 0
         for batch in train_batches:
-            if (args.save_path is not None) and (global_step % args.save_step == 0):
+            if (args.save_path is not None) and (
+                    global_step % args.save_step == 0):
                 best_cor = save_exe(global_step, best_cor)
 
             cost = train_with_feed(batch)
@@ -340,7 +347,8 @@ def finetune(args):
             sum_cost += cost.mean()
 
             if global_step % args.print_step == 0:
-                print('training step %s avg loss %s' % (global_step, sum_cost / args.print_step))
+                print('training step %s avg loss %s' %
+                      (global_step, sum_cost / args.print_step))
                 sum_cost = 0
 
         pass_time_cost = time.time() - begin_time
@@ -361,36 +369,38 @@ def evaluate(args):
     with fluid.scope_guard(fluid.Scope()):
         infer_program, feed_target_names, fetch_vars = fluid.io.load_inference_model(
             args.init_model, exe)
+        print('init model %s' % args.init_model)
 
         global_step, infer_time = 0, 0.0
-        test_batches = reader.batch_reader(
-            args.test_path, args.batch_size, place,
-            args.max_len, 1)
+        test_batches = reader.batch_reader(args.test_path, args.batch_size,
+                                           place, args.max_len, 1)
         scores = []
         labels = []
         for batch in test_batches:
-            logits = exe.run(
-                infer_program,
-                feed = {
+            logits = exe.run(infer_program,
+                             feed={
                                  'context_wordseq': batch[0],
-                    'response_wordseq': batch[1]},
-                fetch_list = fetch_vars)
+                                 'response_wordseq': batch[1]
+                             },
+                             fetch_list=fetch_vars)
             logits = [x[0] for x in logits[0]]
 
             scores.extend(logits)
             labels.extend([x[0] for x in batch[2]])
 
-        mean_score = sum(scores)/len(scores)
+        print('len scores: %s len labels: %s' % (len(scores), len(labels)))
+        mean_score = sum(scores) / len(scores)
         if args.loss_type == 'CLS':
-            recall_dict = eva.evaluate_Recall(zip(scores, labels))
+            recall_dict = eva.evaluate_Recall(list(zip(scores, labels)))
             print('mean score: %s' % mean_score)
             print('evaluation recall result:')
-            print('1_in_2: %s\t1_in_10: %s\t2_in_10: %s\t5_in_10: %s' % (
-                recall_dict['1_in_2'], recall_dict['1_in_10'], 
+            print('1_in_2: %s\t1_in_10: %s\t2_in_10: %s\t5_in_10: %s' %
+                  (recall_dict['1_in_2'], recall_dict['1_in_10'],
                    recall_dict['2_in_10'], recall_dict['5_in_10']))
         elif args.loss_type == 'L2':
             cor = eva.evaluate_cor(scores, labels)
-            print('mean score: %s\nevaluation cor resuls:%s' % (mean_score, cor))
+            print('mean score: %s\nevaluation cor resuls:%s' %
+                  (mean_score, cor))
         else:
             raise ValueError
 
@@ -413,17 +423,16 @@ def infer(args):
             args.init_model, exe)
 
         global_step, infer_time = 0, 0.0
-        test_batches = reader.batch_reader(
-            args.test_path, args.batch_size, place,
-            args.max_len, 1)
+        test_batches = reader.batch_reader(args.test_path, args.batch_size,
+                                           place, args.max_len, 1)
         scores = []
         for batch in test_batches:
-            logits = exe.run(
-                infer_program,
-                feed = {
+            logits = exe.run(infer_program,
+                             feed={
                                  'context_wordseq': batch[0],
-                    'response_wordseq': batch[1]},
-                fetch_list = fetch_vars)
+                                 'response_wordseq': batch[1]
+                             },
+                             fetch_list=fetch_vars)
             logits = [x[0] for x in logits[0]]
 
             scores.extend(logits)
@@ -471,5 +480,6 @@ def main():
     else:
         raise ValueError
 
+
 if __name__ == '__main__':
     main()