ASTE batch_predict error

[Kensvin28]

PyABSA Version (Required)

pyabsa 2.3.1 transformers 4.29.1 torch 2.1.0.dev20230628+cu121

ABSADataset Version (Required if you use integrated datasets)

2.0

Code To Reproduce (Required)

from pyabsa import available_checkpoints from pyabsa import TaskCodeOption

ckpts = available_checkpoints( TaskCodeOption.Aspect_Sentiment_Triplet_Extraction, show_ckpts=True ) print(ckpts) triplet_extractor = ASTE.AspectSentimentTripletExtractor( checkpoint="english" ) triplet_extractor.batch_predict( target_file=ASTE.ASTEDatasetList.Laptop14, print_result=True, save_result=False, ignore_error=True, eval_batch_size=16, )

Full Console Output (Required)

---

TypeError Traceback (most recent call last) Cell In[7], line 1 ----> 1 triplet_extractor.batch_predict( 2 target_file=ASTE.ASTEDatasetList.Laptop14, 3 print_result=True, 4 save_result=False, 5 ignore_error=True, 6 eval_batch_size=16, 7 )

File c:\Users\PAVILION\AppData\Local\Programs\Python\Python310\lib\site-packages\pyabsa\tasks\AspectSentimentTripletExtraction\prediction\predictor.py:187, in AspectSentimentTripletExtractor.batch_predict(self, target_file, print_result, save_result, ignore_error, **kwargs) 183 raise FileNotFoundError("Can not find inference datasets!") 185 self.dataset.prepare_infer_dataset(target_file, ignore_error=ignore_error) --> 187 return self._run_prediction( 188 save_path=save_path if save_result else None, print_result=print_result 189 )

File c:\Users\PAVILION\AppData\Local\Programs\Python\Python310\lib\site-packages\pyabsa\tasks\AspectSentimentTripletExtraction\prediction\predictor.py:228, in AspectSentimentTripletExtractor._run_prediction(self, save_path, print_result, **kwargs) 225 all_results = [] 226 with torch.no_grad(): 227 data_loader = DataIterator( --> 228 self.dataset.convert_examples_to_features(), self.config 229 ) 230 if len(self.dataset) > 1: 231 it = tqdm(data_loader, desc="Predicting")

File c:\Users\PAVILION\AppData\Local\Programs\Python\Python310\lib\site-packages\pyabsa\tasks\AspectSentimentTripletExtraction\dataset_utils\data_utils_for_inference.py:155, in ASTEInferenceDataset.convert_examples_to_features(self, **kwargs) 153 for data in it: 154 try: --> 155 feat = Instance( 156 self.tokenizer, 157 data, 158 self.config.post_vocab, 159 self.config.deprel_vocab, 160 self.config.postag_vocab, 161 self.config.syn_post_vocab, 162 self.config, 163 ) 164 _data.append(feat) 165 except IndexError as e:

File c:\Users\PAVILION\AppData\Local\Programs\Python\Python310\lib\site-packages\pyabsa\tasks\AspectSentimentTripletExtraction\dataset_utils\aste_utils.py:246, in Instance.init(self, tokenizer, sentence_pack, post_vocab, deprel_vocab, postag_vocab, synpost_vocab, config) 244 for row in range(start, end + 1): 245 for col in range(s, e + 1): --> 246 self.word_pair_pos[row][col] = postag_vocab.stoi.get( 247 tuple(sorted([self.postag[i], self.postag[j]])) 248 ) 250 """4. generate synpost index of the word pair""" 251 self.word_pair_synpost = torch.zeros( 252 config.max_seq_len, config.max_seq_len 253 ).long()

TypeError: can't assign a NoneType to a torch.LongTensor

Describe the bug

Cannot run batch_predict on the Aspect_Sentiment_Triplet_Extraction.ipynb tutorial provided.

Expected behavior

Should be able to run batch_predict()

[yangheng95]

Sorry for late response, this is a confirmed bug in data preprocessing to be fixed in future, depending on my schedule.

[NareshNandakumarJHU]

Hello, I seem to have the same error. Has this been fixed yet?

Thanks!