!10425 reformat code and remove useless code in textrcnn

From: @chenmai1102
Reviewed-by: @oacjiewen,@guoqi1024
Signed-off-by: @guoqi1024

model_zoo/research/nlp/textrcnn/data_helpers.py

@@ -16,6 +16,7 @@
 import argparse
 import os
 import numpy as np
+
 parser = argparse.ArgumentParser(description='textrcnn')
 parser.add_argument('--task', type=str, help='the data preprocess task, including dataset_split.')
 parser.add_argument('--data_dir', type=str, help='the source dataset directory.', default='./data_src')
@@ -24,6 +25,7 @@ parser.add_argument('--out_dir', type=str, help='the target dataset directory.',
 args = parser.parse_args()
 np.random.seed(2)
 
+
 def dataset_split(label):
     """dataset_split api"""
     # label can be 'pos' or 'neg'
@@ -33,7 +35,6 @@ def dataset_split(label):
     pos_samples += pfhand.readlines()
     pfhand.close()
     perm = np.random.permutation(len(pos_samples))
-    # print(perm[0:int(len(pos_samples)*0.8)])
     perm_train = perm[0:int(len(pos_samples) * 0.9)]
     perm_test = perm[int(len(pos_samples) * 0.9):]
     pos_samples_train = []
@@ -51,10 +52,7 @@ def dataset_split(label):
     f.close()
 
 
-
 if __name__ == '__main__':
     if args.task == "dataset_split":
         dataset_split('pos')
         dataset_split('neg')
-
-    # search(args.q)

model_zoo/research/nlp/textrcnn/eval.py

@@ -32,7 +32,6 @@ from src.textrcnn import textrcnn
 
 set_seed(1)
 
-
 if __name__ == '__main__':
     parser = argparse.ArgumentParser(description='textrcnn')
     parser.add_argument('--ckpt_path', type=str)
@@ -46,7 +45,7 @@ if __name__ == '__main__':
     context.set_context(device_id=device_id)
 
     embedding_table = np.loadtxt(os.path.join(cfg.preprocess_path, "weight.txt")).astype(np.float32)
-    network = textrcnn(weight=Tensor(embedding_table), vocab_size=embedding_table.shape[0], \
+    network = textrcnn(weight=Tensor(embedding_table), vocab_size=embedding_table.shape[0],
                        cell=cfg.cell, batch_size=cfg.batch_size)
     loss = nn.SoftmaxCrossEntropyWithLogits(sparse=True)
     opt = nn.Momentum(network.trainable_params(), cfg.lr, cfg.momentum)

model_zoo/research/nlp/textrcnn/readme.md

@@ -74,7 +74,7 @@ DEVICE_ID=7 python train.py
 bash scripts/run_train.sh
 
 # run evaluating
-DEVICE_ID=7 python eval.py --ckpt_path ./ckpt/lstm-10_149.ckpt
+DEVICE_ID=7 python eval.py --ckpt_path {checkpoint path}
 # or you can use the shell script to evaluate in background
 bash scripts/run_eval.sh
 ```

model_zoo/research/nlp/textrcnn/src/dataset.py

@@ -21,6 +21,7 @@ import numpy as np
 from mindspore.mindrecord import FileWriter
 import mindspore.dataset as ds
 
+
 # preprocess part
 def encode_samples(tokenized_samples, word_to_idx):
     """ encode word to index """
@@ -78,7 +79,8 @@ def collect_weight(glove_path, vocab, word_to_idx, embed_size):
     # wvmodel = gensim.models.KeyedVectors.load_word2vec_format(os.path.join(glove_path, 'glove.6B.300d.txt'),
     #                                                           binary=False, encoding='utf-8')
     wvmodel = gensim.models.KeyedVectors.load_word2vec_format(os.path.join(glove_path, \
-    'GoogleNews-vectors-negative300.bin'), binary=True)
+                                                                           'GoogleNews-vectors-negative300.bin'),
+                                                              binary=True)
     weight_np = np.zeros((vocab_size + 1, embed_size)).astype(np.float32)
 
     idx_to_word = {i + 1: word for i, word in enumerate(vocab)}
@@ -140,7 +142,7 @@ def convert_to_mindrecord(embed_size, data_path, proprocess_path, glove_path):
         preprocess(data_path, glove_path, embed_size)
     np.savetxt(os.path.join(proprocess_path, 'weight.txt'), weight_np)
 
-    print("train_features.shape:", train_features.shape, "train_labels.shape:", train_labels.shape, "weight_np.shape:",\
+    print("train_features.shape:", train_features.shape, "train_labels.shape:", train_labels.shape, "weight_np.shape:",
           weight_np.shape, "type:", train_labels.dtype)
     # write mindrecord
     schema_json = {"id": {"type": "int32"},

model_zoo/research/nlp/textrcnn/src/textrcnn.py

@@ -22,8 +22,10 @@ from mindspore.common.parameter import Parameter
 from mindspore import Tensor
 from mindspore.common import dtype as mstype
 
+
 class textrcnn(nn.Cell):
     """class textrcnn"""
+
     def __init__(self, weight, vocab_size, cell, batch_size):
         super(textrcnn, self).__init__()
         self.num_hiddens = 512
@@ -89,7 +91,6 @@ class textrcnn(nn.Cell):
         self.tanh = P.Tanh()
         self.sigmoid = P.Sigmoid()
         self.slice = P.Slice()
-    # self.lstm = nn.LSTM(input_size=input_size,hidden_size=hidden_size,num_layers=num_layers,has_bias=has_bias, batch_first=batch_first, bidirectional=bidirectional, dropout=0.0)
 
     def construct(self, x):
         """class construction"""

model_zoo/research/nlp/textrcnn/train.py

@@ -31,7 +31,6 @@ from src.dataset import convert_to_mindrecord
 from src.textrcnn import textrcnn
 from src.utils import get_lr
 
-
 set_seed(2)
 
 if __name__ == '__main__':
@@ -56,7 +55,7 @@ if __name__ == '__main__':
 
     embedding_table = np.loadtxt(os.path.join(cfg.preprocess_path, "weight.txt")).astype(np.float32)
 
-    network = textrcnn(weight=Tensor(embedding_table), vocab_size=embedding_table.shape[0], \
+    network = textrcnn(weight=Tensor(embedding_table), vocab_size=embedding_table.shape[0],
                        cell=cfg.cell, batch_size=cfg.batch_size)
 
     ds_train = create_dataset(cfg.preprocess_path, cfg.batch_size, cfg.num_epochs, True)
@@ -74,7 +73,7 @@ if __name__ == '__main__':
     model = Model(network, loss, opt, {'acc': Accuracy()}, amp_level="O3")
 
     print("============== Starting Training ==============")
-    config_ck = CheckpointConfig(save_checkpoint_steps=cfg.save_checkpoint_steps, \
+    config_ck = CheckpointConfig(save_checkpoint_steps=cfg.save_checkpoint_steps,
                                  keep_checkpoint_max=cfg.keep_checkpoint_max)
     ckpoint_cb = ModelCheckpoint(prefix=cfg.cell, directory=cfg.ckpt_folder_path, config=config_ck)
     model.train(num_epochs, ds_train, callbacks=[ckpoint_cb, loss_cb])