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!