Merge pull request #60 from wangxiao1021/api

fix bugs

examples/classification/run.py

@@ -90,10 +90,10 @@ if __name__ == '__main__':
     # step 5: build forward graph with backbone and task head
     trainer.build_predict_forward(pred_ernie, cls_pred_head)
  
-    # step 6: load pretrained model
+    # step 6: load checkpoint
     # model_path = './outputs/ckpt.step'+str(save_steps)
     model_path = './outputs/ckpt.step'+str(11980)
-    pred_ckpt = trainer.load_ckpt(model_path)
+    trainer.load_ckpt(model_path)
 
     # step 7: fit prepared reader and data
     trainer.fit_reader(predict_cls_reader, phase='predict')

examples/matching/README.md

@@ -5,7 +5,7 @@ This task is a sentence pair matching task. The following sections detail model
 
 #### Download Pre-trained Model
 
-The pre-training model of this mission is: [ernie-en-base](https://github.com/PaddlePaddle/PALM/tree/r0.3-api).
+The pre-training model of this mission is: [ERNIE-v2-en-base](https://github.com/PaddlePaddle/PALM/tree/r0.3-api).
 
 Make sure you have downloaded the required pre-training model in the current folder.
 

examples/matching/run.py

@@ -93,8 +93,8 @@ if __name__ == '__main__':
     # step 5: build forward graph with backbone and task head
     trainer.build_predict_forward(pred_ernie, match_pred_head)
 
-    # step 6: load pretrained model
-    pred_ckpt = trainer.load_ckpt(pred_model_path)
+    # step 6: load checkpoint
+    trainer.load_ckpt(pred_model_path)
 
     # step 7: fit prepared reader and data
     trainer.fit_reader(predict_match_reader, phase='predict')

examples/mrc/run.py

@@ -89,7 +89,7 @@ if __name__ == '__main__':
     # step 5: build forward graph with backbone and task head
     trainer.build_predict_forward(pred_ernie, mrc_pred_head)
 
-    # step 6: load pretrained model
+    # step 6: load checkpoint
     pred_model_path =  './outputs/ckpt.step'+str(12160)
     trainer.load_ckpt(pred_model_path)
     

examples/multi-task/README.md

@@ -5,7 +5,7 @@ This task is a slot filling task. During training, the task uses intent determin
 
 #### Pre-trianed Model
 
-The pre-training model of this mission is: [ernie-en-base](https://github.com/PaddlePaddle/PALM/tree/r0.3-api).
+The pre-training model of this mission is: [ERNIE-v2-en-base](https://github.com/PaddlePaddle/PALM/tree/r0.3-api).
 
 Make sure you have downloaded the required pre-training model in the current folder.
 

examples/multi-task/predict-intent.py

@@ -45,13 +45,13 @@ if __name__ == '__main__':
     # step 5-2: build forward graph with backbone and task head
     trainer.build_predict_forward(pred_ernie, cls_pred_head)
  
-    # step 6: load pretrained model
-    pred_model_path = './outputs/ckpt.step9282'
-    pred_ckpt = trainer.load_ckpt(pred_model_path)
+    # step 6: load checkpoint
+    pred_model_path = './outputs/ckpt.step4641'
+    trainer.load_ckpt(pred_model_path)
 
     # step 7: fit prepared reader and data
     trainer.fit_reader(predict_cls_reader, phase='predict')
 
     # step 8: predict
     print('predicting..')
-    trainer.predict(print_steps=print_steps, output_dir=pred_output)
\ No newline at end of file
+    trainer.predict(print_steps=print_steps, output_dir=pred_output)

examples/multi-task/predict-slot.py

@@ -46,13 +46,13 @@ if __name__ == '__main__':
     # step 5-2: build forward graph with backbone and task head
     trainer_seq_label.build_predict_forward(pred_ernie, seq_label_pred_head)
     
-    # step 6: load pretrained model
-    pred_model_path = './outputs/ckpt.step9282'
-    pred_ckpt = trainer_seq_label.load_ckpt(pred_model_path)
+    # step 6: load checkpoint
+    pred_model_path = './outputs/ckpt.step4641'
+    trainer_seq_label.load_ckpt(pred_model_path)
     
     # step 7: fit prepared reader and data
     trainer_seq_label.fit_reader(predict_seq_label_reader, phase='predict')
    
     # step 8: predict
     print('predicting..')
-    trainer_seq_label.predict(print_steps=print_steps, output_dir=pred_output)
\ No newline at end of file
+    trainer_seq_label.predict(print_steps=print_steps, output_dir=pred_output)

examples/predict/run.py

@@ -42,8 +42,8 @@ if __name__ == '__main__':
     # step 5-2: build forward graph with backbone and task head
     trainer.build_predict_forward(pred_ernie, cls_pred_head)
  
-    # step 6: load pretrained model
-    pred_model = trainer.load_predict_model(pre_params)
+    # step 6: load checkpoint
+    trainer.load_predict_model(pre_params)
 
     # step 7: fit prepared reader and data
     trainer.fit_reader(predict_cls_reader, phase='predict')

examples/tagging/run.py

@@ -93,7 +93,7 @@ if __name__ == '__main__':
     # step 5: build forward graph with backbone and task head
     trainer.build_predict_forward(pred_ernie, seq_label_pred_head)
     
-    # step 6: load pretrained model
+    # step 6: load checkpoint
     pred_model_path = './outputs/ckpt.step' + str(save_steps)
     trainer.load_ckpt(pred_model_path)
     

paddlepalm/head/mrc.py

@@ -22,6 +22,7 @@ import math
 import six
 import paddlepalm.tokenizer.ernie_tokenizer as tokenization
 import json
+import io
 
 RawResult = collections.namedtuple("RawResult",
                                    ["unique_id", "start_logits", "end_logits"])
@@ -361,15 +362,15 @@ def _write_predictions(all_examples, all_features, all_results, n_best_size,
     
 
 
-    with open(output_prediction_file, "w") as writer:
+    with io.open(output_prediction_file, "w", encoding='utf-8') as writer:
         
         writer.write(json.dumps(all_predictions, indent=4, ensure_ascii=False) + "\n")
 
-    with open(output_nbest_file, "w") as writer:
+    with io.open(output_nbest_file, "w", encoding='utf-8') as writer:
         writer.write(json.dumps(all_nbest_json, indent=4, ensure_ascii=False) + "\n")
 
     if with_negative:
-        with open(output_null_log_odds_file, "w") as writer:
+        with io.open(output_null_log_odds_file, "w", encoding='utf-8') as writer:
             writer.write(json.dumps(scores_diff_json, indent=4, ensure_ascii=False) + "\n")
 
 

paddlepalm/multihead_trainer.py

@@ -6,6 +6,7 @@ from paddlepalm import Trainer
 from paddlepalm.utils import reader_helper
 import numpy as np
 import time
+import sys
 
 dev_count = 1 if gpu_dev_count <= 1 else gpu_dev_count
 VERBOSE=False
@@ -156,6 +157,7 @@ class MultiHeadTrainer(Trainer):
             max_train_steps = int(num_epochs * t.mix_ratio * base_steps_pur_epoch)
             if not t._as_auxilary:
                 print('{}: expected train steps {}.'.format(t.name, max_train_steps))
+                sys.stdout.flush()
                 self._finish_steps[t.name] = max_train_steps
                 self._finish[t.name] = False
             else:
@@ -175,6 +177,7 @@ class MultiHeadTrainer(Trainer):
             joint_shape_and_dtypes.append(t._shape_and_dtypes)
 
         print('Estimated overall train steps {}.'.format(global_steps))
+        sys.stdout.flush()
         self._overall_train_steps = global_steps
 
         iterator_fn = reader_helper.create_multihead_iterator_fn(iterators, prefixes, joint_shape_and_dtypes, \
@@ -198,6 +201,7 @@ class MultiHeadTrainer(Trainer):
         if trainers[task_name]._cur_train_step == self._finish_steps[task_name]:
             if not silent:
                 print(task_name+' train finish!')
+                sys.stdout.flush()
             self._finish[task_name]=True
         flags = list(set(self._finish.values()))
         return len(flags) == 1 and flags[0] == True
@@ -235,6 +239,7 @@ class MultiHeadTrainer(Trainer):
                        (self._trainers[task_id]._cur_train_step-1) % self._trainers[task_id]._steps_pur_epoch + 1, \
                        self._trainers[task_id]._steps_pur_epoch, self._trainers[task_id]._cur_train_epoch, \
                        loss, print_steps / time_cost))
+                sys.stdout.flush()
                 time_begin = time.time()
 
             self._check_save()

paddlepalm/trainer.py

@@ -18,6 +18,7 @@ import os
 import json
 from paddle import fluid
 import time
+import sys
 import numpy as np
 import paddlepalm.utils.basic_helper as helper
 from paddlepalm.utils import reader_helper, saver
@@ -546,9 +547,11 @@ class Trainer(object):
                 if self._save_predict:
                     self._save(save_path, suffix='pred.step'+str(self._cur_train_step))
                     print('predict model has been saved at '+os.path.join(save_path, 'pred.step'+str(self._cur_train_step)))
+                    sys.stdout.flush()
                 if self._save_ckpt:
                     fluid.io.save_persistables(self._exe, os.path.join(save_path, 'ckpt.step'+str(self._cur_train_step)), self._train_prog)
                     print('checkpoint has been saved at '+os.path.join(save_path, 'ckpt.step'+str(self._cur_train_step)))
+                    sys.stdout.flush()
                 return True
             else:
                 return False
@@ -608,6 +611,7 @@ class Trainer(object):
                 print("step {}/{} (epoch {}), loss: {:.3f}, speed: {:.2f} steps/s".format(
                        (self._cur_train_step-1) % self._steps_pur_epoch + 1 , self._steps_pur_epoch, self._cur_train_epoch,
                        loss, print_steps / time_cost))
+                sys.stdout.flush()
                 time_begin = time.time() 
                 # self._check_save()
             # if cur_task.train_finish and cur_task.cur_train_step + cur_task.cur_train_epoch * cur_task.steps_pur_epoch == cur_task.expected_train_steps:
@@ -653,6 +657,7 @@ class Trainer(object):
                 print("batch {}/{}, speed: {:.2f} steps/s".format(
                        cur_predict_step, self._pred_steps_pur_epoch,
                        print_steps / time_cost))
+                sys.stdout.flush()
                 time_begin = time.time()
 
         if self._pred_head.epoch_inputs_attrs:
@@ -816,6 +821,7 @@ class Trainer(object):
         with open(os.path.join(dirpath, '__conf__'), 'w') as writer:
             writer.write(json.dumps(conf, indent=1))
         print(self._name + ': predict model saved at ' + dirpath)
+        sys.stdout.flush()
 
     
     def _load(self, infer_model_path=None):
@@ -827,5 +833,6 @@ class Trainer(object):
         pred_prog, self._pred_input_varname_list, self._pred_fetch_var_list = \
             fluid.io.load_inference_model(infer_model_path, self._exe)
         print(self._name+': inference model loaded from ' + infer_model_path)
+        sys.stdout.flush()
         return pred_prog