Merge branch 'develop' into joe_model_1.8_update_1

PaddleNLP/language_model/run.sh

@@ -4,7 +4,7 @@ export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
 function run_train ()
 {
     echo "training"
-    python train.py \
+    python3 train.py \
         --data_path data/simple-examples/data/ \
         --model_type small \
         --use_gpu True \

PaddleNLP/pretrain_language_models/BERT/convert_params.py

@@ -183,7 +183,7 @@ def convert(args):
             param]).get_tensor().set(value, place)
         print(param, ' --> ', tf_fluid_param_name_map[param], '  ', value.shape)
 
-    fluid.save(model_path=args.fluid_params_dir, program=program)
+    fluid.save(program=program, model_path=args.fluid_params_dir)
 
 
 if __name__ == '__main__':

PaddleNLP/pretrain_language_models/BERT/train.sh

@@ -32,7 +32,7 @@ VALIDATION_DATA_DIR=data/validation
 CONFIG_PATH=data/demo_config/bert_config.json
 VOCAB_PATH=data/demo_config/vocab.txt
 # Change your train arguments:
-python -u ./train.py ${is_distributed}\
+python3 -u ./train.py ${is_distributed}\
         --use_cuda true\
         --weight_sharing true\
         --batch_size ${BATCH_SIZE} \

PaddleNLP/pretrain_language_models/XLNet/run_classifier.py

@@ -415,7 +415,7 @@ def main(args):
                 if steps % args.save_steps == 0:
                     save_path = os.path.join(args.checkpoints,
                                              "step_" + str(steps))
-                    fluid.save(model_path=save_path, program=train_program)
+                    fluid.save(program=train_program, model_path=save_path)
 
                 if steps % args.validation_steps == 0:
                     print("Average throughtput: %s" % (np.average(throughput)))
@@ -427,7 +427,7 @@ def main(args):
                                  args.eval_split, processor.get_num_examples(phase=args.eval_split))
             except fluid.core.EOFException:
                 save_path = os.path.join(args.checkpoints, "step_" + str(steps))
-                fluid.save(model_path=save_path, program=train_program)
+                fluid.save(program=train_program, model_path=save_path)
                 train_data_loader.reset()
                 break
         if args.enable_ce:

PaddleNLP/pretrain_language_models/XLNet/run_squad.py

@@ -583,11 +583,11 @@ def train(args):
                 if steps % args.save_steps == 0 or steps == args.train_steps:
                     save_path = os.path.join(args.checkpoints,
                                              "step_" + str(steps))
-                    fluid.save(model_path=save_path, program=train_program)
+                    fluid.save(program=train_program, model_path=save_path)
             except fluid.core.EOFException:
                 save_path = os.path.join(args.checkpoints,
                                          "step_" + str(steps) + "_final")
-                fluid.save(model_path=save_path, program=train_program)
+                fluid.save(program=train_program, model_path=save_path)
                 train_data_loader.reset()
                 break
         print("Finish model training ...")

PaddleNLP/pretrain_language_models/XLNet/utils/init.py

@@ -55,7 +55,7 @@ def init_checkpoint(exe, init_checkpoint_path, main_program, use_fp16=False):
             return False
 
     fluid.load(
-        model_path=init_checkpoint_path, program=main_program, executor=exe)
+        program=main_program, model_path=init_checkpoint_path, executor=exe)
 
     if use_fp16:
         cast_fp32_to_fp16(exe, main_program)