fix bugs for flake8

Author: wintermelon008

EduNLP/ModelZoo/quesnet/quesnet.py

@@ -160,7 +160,7 @@ def make_batch(self, data, device, pretrain=False):
 
             for i, fo in enumerate(q.false_options):
                 false_options[i].append([0] + fo)
-     
+
         lembs = SeqBatch(lembs, device=device)
         rembs = SeqBatch(rembs, device=device)
         embs = SeqBatch(embs, device=device)
@@ -195,24 +195,8 @@ def make_batch(self, data, device, pretrain=False):
 
         words = torch.cat(words, dim=0) if words else None
         ims = torch.cat(ims, dim=0) if ims else None
-        metas = torch.cat(metas, dim=0) if metas else None
-        
-        
-        # print("debug1")
-        # print(lembs)
-        # print(rembs)
-        # print(words)
-        # print(ims)
-        # print(metas)
-        # print(wmask)
-        # print(imask)
-        # print(mmask)
-        # print(embs)
-        # print(ans_input)
-        # print(ans_output)
-        # print(false_opt_input)
-        
-        
+        metas = torch.cat(metas, dim=0) if metas else None      
+
         if pretrain:
             return (
                 lembs, rembs, words, ims, metas, wmask, imask, mmask,
@@ -331,7 +315,7 @@ def forward(self, batch):
         h = outputs.hidden
 
         x = ans_input.packed()
-        
+
         y, _ = self.ans_decode(PackedSequence(self.quesnet.we(x[0].data), x.batch_sizes),
                                h.repeat(self.config.layers, 1, 1))
         floss = F.cross_entropy(self.ans_output(y.data),

EduNLP/ModelZoo/quesnet/util.py

@@ -23,7 +23,7 @@ def __init__(self, seqs, dtype=None, device=None):
         self._prefix = [0]
         self._index = {}
         c = 0
-        
+
         for i in range(self.lens[0]):
             for j in range(len(self.lens)):
                 if self.lens[j] <= i:
@@ -40,8 +40,9 @@ def packed(self):
 
     def padded(self, max_len=None, batch_first=False):
         if not self.seqs:
-            return torch.empty((0, 0), dtype=self.dtype, device=self.device), torch.empty((0, 0), dtype=torch.bool, device=self.device)
-    
+            return torch.empty((0, 0), dtype=self.dtype, device=self.device), \
+                torch.empty((0, 0), dtype=torch.bool, device=self.device)
+
         seqs = [torch.tensor(s, dtype=self.dtype, device=self.device)
                 if not isinstance(s, torch.Tensor) else s
                 for s in self.seqs]

EduNLP/Pretrain/quesnet_vec.py

@@ -39,6 +39,7 @@ def save_list(item2index, path):
 def clip(v, low, high):
     return max(low, min(v, high))
 
+
 # Basic unit of Dataset
 Question = namedtuple('Question',
                       ['id', 'content', 'answer', 'false_options', 'labels'])
@@ -334,7 +335,7 @@ def __init__(
         option_key=lambda x: x['ques_options'],
         pipeline=None,
         skip=0
-        ):
+    ):
 
         self.filename = filename
         self.skip = skip
@@ -349,7 +350,7 @@ def __init__(
             tokenizer = QuesNetTokenizer(
                 meta=['know_name'],
                 img_dir=img_dir
-                )
+            )
         self.tokenizer = tokenizer
         self.meta = meta if meta else tokenizer.meta
         self.load_data_lines()
@@ -358,16 +359,15 @@ def __init__(
             key=lambda x: x['ques_content'],
             trim_min_count=2,
             silent=False
-            )
+        )
         tokenizer.set_meta_vocab(self.lines, silent=False)
-
 
     def load_data_lines(self):
         '''Read data by row from a JSON file
-        
+
         Important: the data file is loaded during initialization.
         '''
-        
+
         # TODO: All data is read into memory without chunking.
         #       This may lead to low efficiency.
         data_dir = self.filename
@@ -402,7 +402,7 @@ def __getitem__(self, index):
             meta = token['meta_idx']
 
             if self.answer_key(line).isalpha() and len(self.answer_key(line)) == 1 \
-                and ord(self.answer_key(line)) < 128 and len(self.option_key(line)) > 0:
+                    and ord(self.answer_key(line)) < 128 and len(self.option_key(line)) > 0:
                 answer_idx = ord(self.answer_key(line).upper()) - ord('A')
                 options = self.option_key(line)
                 answer = self.tokenizer(options.pop(answer_idx), meta=self.meta)['seq_idx']
@@ -417,7 +417,7 @@ def __getitem__(self, index):
                 answer=answer,
                 false_options=false_options,
                 labels=meta
-                )
+            )
 
             if callable(self.pipeline):
                 qs = self.pipeline(qs)
@@ -556,17 +556,25 @@ def pretrain_embedding_layer(dataset: EmbeddingDataset, ae: AE, lr: float = 1e-3
 
 def optimizer(*models, **kwargs):
     _cur_optim = [
-        m.optim_cls(m.parameters(), **kwargs) 
-        if hasattr(m, 'optim_cls') 
+        m.optim_cls(m.parameters(), **kwargs)
+        if hasattr(m, 'optim_cls')
         else torch.optim.Adam(m.parameters(), **kwargs) for m in models
-        ]
+    ]
     if len(_cur_optim) == 1:
         return _cur_optim[0]
     else:
         return _cur_optim
 
-        
-def pretrain_quesnet(path, output_dir, pretrain_dir=None, img_dir=None, save_embs=False, load_embs=False, train_params=None):
+
+def pretrain_quesnet(
+    path,
+    output_dir,
+    pretrain_dir=None,
+    img_dir=None,
+    save_embs=False,
+    load_embs=False,
+    train_params=None
+):
     """ pretrain quesnet
 
     Parameters
@@ -672,7 +680,7 @@ def pretrain_quesnet(path, output_dir, pretrain_dir=None, img_dir=None, save_emb
             sentences=[[item] for item in emb_dict.keys()],
             min_count=1,
             vector_size=emb_size
-            )
+        )
         gensim_w2v.init_weights()
         gensim_w2v.train(corpus_iterable=w2v_corpus, total_examples=len(w2v_corpus), epochs=train_params['n_epochs'])
         w2v_emb = gensim_w2v.syn1neg
@@ -699,7 +707,7 @@ def pretrain_quesnet(path, output_dir, pretrain_dir=None, img_dir=None, save_emb
             batch_size=train_params['batch_size'],
             epochs=train_params['n_epochs'],
             device=device
-            )
+        )
         if save_embs:
             torch.save(trained_ie.state_dict(), os.path.join(output_dir, 'trained_ie.pt'))
         model.quesnet.load_img(trained_ie)
@@ -718,7 +726,7 @@ def pretrain_quesnet(path, output_dir, pretrain_dir=None, img_dir=None, save_emb
             batch_size=train_params['batch_size'],
             epochs=train_params['n_epochs'],
             device=device
-            )
+        )
         if save_embs:
             torch.save(trained_me.state_dict(), os.path.join(output_dir, 'trained_me.pt'))
         model.quesnet.load_meta(trained_me)